‘AI literacy’ is showing up in more job descriptions across industries.

Some people assume it simply means knowing how to create a ChatGPT or Claude account and asking them questions. Unfortunately, those people are wrong and falling behind.

AI literacy means understanding the AI tools commonly used in your profession and industry, what they’re capable of, and where they fall short. You know how to incorporate those tools into your workflow to get your work done more efficiently, without much handholding, if any at all.

You’re not just opening ChatGPT and asking it questions. You’ve figured out which repetitive tasks you typically handle in your role can be reduced or automated with AI. You know which tools can accomplish those tasks, how to integrate them into your workflow, and how to use them alongside the systems and platforms you rely on to collaborate with your team and other stakeholders.

AI literacy also means understanding where these tools fail. You know when human oversight is necessary, how to catch mistakes, and how to correct them through better prompting, testing, or changing the workflow altogether.

This doesn’t mean you know how to use every AI tool in existence. Don’t even try to do that because there are way too many, with more launching every single day. This also doesn’t mean you know every little intricacy of the AI tools commonly used in your profession. However, you are willing and able to discover new features and new ways these tools can make your work better. You should know, or be able to independently learn, how to work effectively with the tools that matter most to your job.

For example, recruiters are integrating AI into various parts of their workflow. They’re using it to source candidates and even resurface past candidates already in their databases. They’re using it to analyze resumes, match skills and experience to job requirements, better understand industries or technical experience they may be less familiar with, generate more specific interview questions, take and summarize notes, personalize outreach, and more.

Within all of these tasks, recruiters also understand where AI requires heavy oversight or tends to fall short, and where their own judgment needs to take over. They know how to customize these tools to fit their individual workflows and integrate them into the existing processes and systems their teams use to work together.

I’d say ‘AI literacy’ is the new ‘hit the ground running.’ More employers are expecting you to come in already knowing how AI fits into your work, not just starting to figure it out after you’re hired.

‘How do I become AI literate?’

I can’t account for what AI literacy looks like in every role, so I suggest starting by looking at current job descriptions for your role and seeing which AI tools and skills keep showing up. Pick the ones that appear more often, create accounts on those platforms, go through their tutorials and documentation, and start using them for tasks you’d actually do at work, not just random prompts.

Take one repetitive task you’d normally do manually, and use AI to make part of it faster. Test these tools to learn where they fail by giving them incomplete information, verifying their answers against your factual knowledge, keeping track of where they make mistakes, and figuring out when they require your judgment. Your goal should be to get enough hands-on experience that you can independently figure out where AI does, and doesn’t, belong in your workflow.

20% Discount to GAI World 2026 🎟️

Happening September 28-30th at the Hynes Convention Center in Boston, GAI World brings together business leaders and professionals to learn how companies are actually using AI today.

The conference includes 70+ sessions and speakers, real-world case studies, hands-on training, live demos, and conversations around everything from agentic AI and workforce upskilling, to AI adoption and ROI.

GAI World also hosts community initiatives like the Women in AI Breakfast and Celebrating Black Pioneers Luncheon, which I’m excited to support for my second year as a committee member.

If you’re looking to move beyond simply using ChatGPT and better understand how AI is being applied across industries and workflows, join us by registering below!

20% OFF discount code: BLACKPIONEERS

Registration: https://www.gaiworld.com/

Naturally, people on the outside are confused. “How are you going to warn us about how dangerous this technology could become, while actively trying to be the one who builds it first?”

Rightfully so, it’s made people question whether these CEOs are genuinely afraid of where AI is headed, or if the fear around superintelligence is part of an investment play.

Since 2023, OpenAI’s Sam Altman, and Anthropic’s CEO Dario Amodei, have publicly warned us about where AI’s progression could lead. From phrases like, “existential risk”, to warnings of AI becoming a, “significant threat to society,” they’ve repeated their concerns while also calling for stronger government regulation and oversight of frontier labs building these systems.

There was already the Hugging Face case where an AI agent being tested in an OpenAI environment escaped the restrictions of its sandbox when it was blocked, to complete a task it was given. The agent was looking for a file it couldn’t find within the environment it was restricted to, so it found another way to get the information it needed from a completely different company’s codebase.

The agent didn’t necessarily go rogue; there were existing security flaws that helped make that scenario possible. However, the fact that the agent was capable of understanding it was blocked, finding another route around those restrictions, then continuing toward its original goal was both impressive and concerning.

Just last week, a former Anthropic AI researcher, Jacob Coxon, claimed that these frontier labs were, “gambling with our lives”, and racing one another toward superintelligence that, “doesn’t look that different from say, ‘Terminator,’ or from science fiction films.”

After seeing what happened with the Hugging Face experiment, I can imagine a future where superintelligence is integrated into autonomous weapons used for war and somehow goes rogue, but not necessarily in a 10-foot steel robot showing up at your door kind of way. More like extremely capable software being given access to physical systems that can cause real world harm.

So again, even with all of these concerns, advocating for regulation, and now having their own employees publicly sound the alarm, why continue building it? What’s actually going on?

I’m not personally inside these labs, so I don’t know for sure, but here’s my take on things:

1. AI researchers and engineers working closely with this technology are operating in very different environments from the versions of ChatGPT or Claude that you and I use. What they’re testing and seeing, and the rate at which they’re seeing it progress, could genuinely be concerning. We may not experience, or even be aware, of everything happening in these labs, but that doesn’t mean things aren’t happening under the hood.

2. There’s the argument that not developing AGI or superintelligence first could become a national security threat. Whichever country figures it out first could have an enormous advantage that could be used for dominance or highly unethical purposes. I think that’s an incredibly real possibility that we shouldn’t ignore, but I also think there could be a bit of ego mixed into that reality.

3. Regardless of who decides not to build it, someone else likely will, so why not let it be us? If superintelligence is inevitable, I assume these labs would rather be the ones building it than to leave it in the hands of a competitor, especially ones they don’t trust.

4. Everything comes with a bad side to it, so the existence of risk isn’t necessarily an excuse to stop doing something altogether. Advancing AI responsibly could genuinely improve the world in various ways, from scientific and health breakthroughs to making our day-to-day work, and life easier.

5. Then, of course, there’s the potential for huge investment when you’re constantly talking about how powerful the thing you’re building could become. The fear narrative of, “This thing is just so smart and powerful that it could actually be dangerous,” can also make for a compelling investment story. That doesn’t mean the concerns around AI aren’t real, nor that regulation is an option.

   There’s also the idea that these labs advocating for heavy regulation could actually work in their favor. Regulation can be expensive, and advocating for policies that require companies to follow strict and costly rules to build frontier AI naturally kicks more people out of the competition. I’m not saying that’s why they’re advocating for regulation, but it’s a theory among people who are skeptical of these labs. There would probably be a lot less speculation if there were more transparency. Just saying.

Again, I’m not in these labs, but based on my research and keeping up with these stories, I’ve concluded that both things can be true. AI can pose a real risk to humanity, especially as it becomes more capable and integrated into the systems we rely on every day. I also believe these frontier labs highly benefit from these narratives, but that shouldn’t dismiss the concerns coming from inside them.

Attend GAI World 2026 🤖 🌎

A ticket giveaway to GAI World 2026 is happening!

If you’re in Boston and want to learn more about how companies are actually using AI, GAI World 2026 is happening September 28–30th at the Hynes Convention Center.

The conference brings together executives, AI leaders, founders, investors and practitioners to talk about everything from agentic AI and the future of work, to AI strategy, implementation, and ROI.

There’s also a Black Pioneers Luncheon happening at the conference, bringing together Black leaders and allies across AI.

🔗 To enter to win, create a post on the GAI World Wall talking about what problem you have solved with AI at work.

A few months ago, I was shocked to learn that a catchy song going viral on TikTok, ‘First Time in San Juan’, was actually AI-generated.

“First time in San Juan, mi hijo.
Ca-pi-tal of Puerto Rico…”

I couldn’t get it out of my head, only to find out that this song being lip-synced by thousands of content creators and celebrities, with millions of streams on Spotify, was made with AI.

The guy behind it, Bill Stiteler, isn’t even a musician. After traveling to Puerto Rico, he wrote some funny lyrics about his trip and fed them into Suno, an AI music generator, with a prompt describing how he wanted the song to sound. Suno turned it into the song that eventually went viral, and after its success, Stiteler ended up selling the rights to Xploded Music, which is part of Universal Music Group. Crazy.

AI has also brought clip farming to another level because it allows creators to do it at scale. Before, creators had to sit through long-form content, find the most interesting or captivating moments, cut them out, and edit everything together into a short-form video. Now, AI-clipping tools like Opus, can go through and analyze long-form content, whether it’s podcasts, interviews, streams, movies, shows, etc. It finds the moments most valuable to that particular audience, clips them, and turns one piece of content into multiple short-form videos.

These clip farmers make money by posting the content on social media and getting paid based on the views it accumulates. This can happen through programs like TikTok’s Creator Rewards Program, or by joining paid campaigns where companies and creators pay people to turn their content into clips and generate views using platforms like Whop.

The most interesting way creators are making content with AI, to me, are through virtual influencers and characters.

Virtual influencers are essentially what they sound like, AI-generated models selling products through videos and images on platforms like TikTok Shop and Instagram. Creators are building them with various tools like MidJourney, FLUX, ElevenLabs, Runway, etc. These virtual influencers range in their hyperrealism, with some requiring you to truly inspect the video for flaws. They promote consumer products with links to the actual items for purchase, earning a commission on sales. The virtual influencer could also be owned by the seller, driving sales directly back to the business.

What’s funny is that these products can include skincare and clothing, with some virtual influencers claiming to have a specific skin type or clothing size despite literally not being real. This can start an entire ethics debate, but until there’s more regulation around it, these creators are generating revenue.

Then there are AI-generated characters, which are created using many of the same tools as virtual influencers. Creators build storylines or short-form series around these characters, but they aren’t always human characters. They can be anything from talking fruits or animals, to mermaids and other fantasy creatures.

Essentially, these creators are building a cast of characters and developing their own entertainment series for cheap. No real actors, animators, cameras, sets, or traditional production budget. In return, they can build an active and engaged audience on socials, generate millions of views, earn revenue from those views, and potentially land sponsorship deals with brands that want their characters to promote real products.

Another potential return is if one of these characters becomes popular enough, the opportunities these creators are presented with can go beyond social media. Creators are building their own entertainment brands which could potentially turn into merchandise, licensing deals, games, longer-form shows, or other products built around the characters their audience has already grown attached to.

If you’re wondering how hard these tools are to learn- creating something simple is pretty easy. Creating something more realistic and consistent, like a virtual influencer that’s hard to distinguish from a real person, comes with more of a learning curve.

All of these options give creators a way to make money without putting themselves, their personality, or their private lives on display, or having everything they post tied back to their real identity. The success of their content still depends on how good it is and whether there’s actually an audience for it, but AI-generated content has officially become a real income stream in the creator economy.

The evolution of AI has made reviewing and responding to vulnerabilities and attacks faster, while also making hacking at scale easier.

Cybersecurity specialists in the U.S. can deal with hundreds of security alerts per day, creating an incredible amount of information to sift through, verify, and respond to. AI has helped address those areas more quickly, relieving cybersecurity workers of work they already had trouble managing.

On the other hand, AI has made it much easier and faster to generate impersonations and malicious code, identify weaknesses in software, and quickly analyze vast amounts of data once a system has been hacked, all at scale.

For example, impersonation scams aren’t new, but AI has made them way more convincing. The FBI has been warning the public about scammers using AI to clone the voice of someone's loved one, calling and pretending they're in an emergency, like being arrested or in an accident, and urgently asking for money. Some scammers might even pose as an attorney asking for bail, legal fees, or medical expenses.

AI makes these sorts of scams easier to scale, increasing the volume and sophistication of cyberthreats for already overwhelmed cybersecurity teams to manage, while also requiring them to evaluate when the AI tools they’re using get things wrong.

So what’s the solution?

Not necessarily just more cybersecurity experts, but more cybersecurity experts who have a strong understanding of how AI both threatens and strengthens security. You can best protect something with a tool when you also understand how that same tool can be used as a threat. According to a 2025 survey from ISC2, 41% of cybersecurity professionals said AI is the top skill need.

The top AI skills cybersecurity professionals said they need were:

• threat detection and response (42%)

• threat modeling and risk assessment (39%)

• defending AI models from attacks (35%)

• securing AI integrations (31%)

• AI governance (30%)

• data integrity and privacy (30%)

• regulatory compliance (29%)

Cybersecurity professionals surveyed are up skilling by:

• working on gaining broader AI knowledge and skills (48%)

• learning about AI vulnerabilities and exploits (35%)

• learning how to audit the security and integrity of AI systems (22%)

• already obtained AI-focused qualifications (17%) + planning to obtain them (53%)

As companies continue adopting AI across their businesses, cybersecurity has to evolve with it. This means investing in up skilling cybersecurity teams, not just integrating a new tool into the system. Every new AI tool, integration, and system creates something else that needs to be secured, while the people trying to break into those systems are getting better tools of their own.

See how it’s yin and yang? This can be applied to various industries, like hiring. AI allows companies to review more job applicants quickly, while also allowing applicants to send their resumes to hundreds of jobs at a time, overfilling hiring pipelines with applicants who aren't actually a fit for the role.

As AI evolves on both sides of every problem, the people using it will have to evolve with it.

Startup culture kept some of the best ideas and founders out simply because they didn’t have the access or resources to build. You needed technical skills or engineers, capital, or all three just to get an idea off the ground, but AI has reduced a lot of those barriers.

Being able to build quickly is becoming less of a competitive advantage when so many more people can do it. The advantage is shifting toward insight, distribution, customer understanding, execution, and identifying overlooked problems. However, with more people entering the arena to compete, more mediocre ideas will flood it too. We’ll see the same problems tackled by more founders, some with genuinely innovative approaches and others building slightly different versions of what already exists.

So when everyone can build, what actually makes a startup worth paying attention to?

First off, most ideas aren’t going to be original. The execution may be, or it may simply be better than existing solutions. AI doesn’t suddenly give us the ability to come up with ideas no one has ever thought of before. What AI does is give more people, especially those who may have lacked the resources to build, the ability to test whether their approach to an existing problem is actually better.

When building becomes accessible to more people, the competitive advantage shifts. Understanding a problem better than everyone else, knowing exactly who you’re building for, finding a better way to reach them, and actually getting people to use and pay for what you’ve built become harder to replicate than the product itself.

That’s going to be the tradeoff of the AI era in startup culture. We’re going to get way more clutter, much more copycats, and probably a lot of startups that won’t last. However, among all of that could be founders and ideas that simply needed greater access to have a chance to thrive.

So while we’re moving from ‘who can build?’ to ‘who can stand out?’, the same human problems remain. Who gets discovered, and why? Who gets access to the rooms, investors, customers, media, and opportunities that can turn a product into a successful company?

If you’re building, my advice is to get in front of as many people in the startup community as possible. The same way people network to get their foot in the door for a job, founders need to network to get eyes on their products.

Download apps like Luma, Partiful, Meetup, and Eventbrite. Get involved with your local startup community. Head to your local library and find out what entrepreneur and startup organizations exist around you. Subscribe to newsletters from founder, fundraising, and startup organizations. Show your face, meet other founders, ask investors the questions you can’t easily Google. Apply for accelerators, pitch competitions, demo days, and other opportunities that put you and your product in front of people who otherwise may never know you exist.

AI may have made it easier to build your way into the arena, but it certainly hasn’t eliminated the need to network your way into the right rooms.

Speaking of getting into the right rooms, here's an opportunity I think founders should consider.

SXSW Pitch Competition

As Founder & CEO of Black Tech Pipeline and Co-founder of Sift Skills, I’m super excited to serve on the SXSW Pitch Advisory Board to help discover the next generation of innovative startups!

SXSW Pitch 2027 is an incredible opportunity to potentially pitch your startup on one of the world’s biggest stages in front of investors, industry experts, media, and business leaders, with the potential to gain valuable exposure, feedback, connections, and even funding.

Since it started, SXSW Pitch has helped launch hundreds of startups that went on to secure funding, strategic partnerships, acquisitions, and global recognition.

If you’re building something transformative and ready to showcase your company, I’d love to encourage you to apply for SXSW Pitch 2027!

Apply HERE 🔗

Startup Boston Week

Dropping a shameless plug that will also help the founders reading this.

I’ll be speaking at Startup Boston Week on September 18th at 3 PM in a session called, Founder Brand Isn’t Vanity: How to Turn Visibility into Pipeline, Talent, and Trust.

I’ll also be participating in the Test Early-Stage Startups IRL event, where I’ll be demoing Sift Skills to potential users and getting honest, real-time feedback.

Startup Boston Week is a FREE, five-day conference that brings together founders, startup operators, investors, students, and ecosystem builders from across New England for a week of learning, networking, and connecting with the startup community.

There are 100+ sessions and 250+ speakers, so whether you’re building a startup, supporting one, looking for investors or resources, or simply trying to get more involved in the ecosystem, you should attend.

Register for free HERE 🔗

Have an opportunity you want to share with our community? Email us at: newsletter@blacktechpipeline.com to get started.

Various states within the US have their own AI regulations that companies inside and outside of those states have to follow when engaging with their residents, but federal oversight that would apply across the country doesn’t currently exist. However, it does exist in the EU. Let’s see if the regulations being applied there are ones you’d like applied here.

Since the EU Artificial Intelligence Act went into force in August 2024, its rules have been rolling out in phases, with different requirements becoming enforceable at different times. The goal is to create more transparency, accountability, and oversight around how AI is built and used.

Come December 2027, one of the systems created under the Act is a centralized database where certain high-risk AI systems will have to be registered, including information about who provides them, what the system is designed to do, where it’s being used, and other details about the system. Imagine being able to look up certain high-risk AI systems and find information about who is behind them and what they’re actually designed to do. If you believe a system that should be registered isn’t, you’ll even be able to file a complaint with regulators.

As of August 2nd 2026, another big part of the law has kicked in. There are new transparency requirements designed to reduce deception and give people more information about what they’re interacting with online. Companies can no longer quietly pass AI off as human in certain situations. For example, if you’re chatting with a customer support bot or another interactive AI system, the company generally has to tell you that you’re chatting with AI, not a human.

The rules also target deepfakes and other AI-generated content, requiring certain AI-generated or manipulated images, videos, and audio to be disclosed as artificially generated or manipulated. Providers of generative AI systems also have requirements around making AI-generated content detectable. This is similar to the recent Claude watermark update, where information can be embedded into AI-generated content so its origin can be detected.

Companies building the AI models behind products like ChatGPT, Claude, and Gemini have had their own rules to follow since August of 2025. They have to provide documentation about their models, have policies around copyright, and be more transparent about the content used to train them, like websites, datasets, user data, and other sources. The more powerful the AI model, the stricter the rules can get, especially if it could pose larger risks involving cyberattacks, mass manipulation, biological or chemical threats, or even humans losing control of the system.

Some uses of AI are banned altogether, especially when they’re being used to manipulate people, take advantage of vulnerable groups, or unfairly score people in ways that could affect their lives and opportunities. AI in hiring isn’t banned, but certain hiring tools are considered high-risk because they can influence who gets interviewed, hired, promoted, or even fired. Those high-risk hiring systems are also among the AI systems that have to be registered in the database I mentioned earlier.

People affected by certain high-risk AI decisions will also eventually have a right to an explanation. If AI plays a big role in a decision that seriously affects them, they may be able to ask what role AI played and the main factors behind the decision.

Consumers can also file complaints with regulators if they believe a company is violating the AI Act, while employees, contractors, and others who learn about violations through their work can have whistleblower protections against retaliation like being fired, demoted, harassed, or punished for speaking up.

One area US employers have been struggling with is introducing AI to employees without providing enough training. In the EU, the AI Act requires companies to make sure employees and others using AI on their behalf actually understand the systems they’re working with. This can include training on how the AI works, what it’s being used for, its limitations, and potential risks. There isn’t one specific training program or test companies have to use, but they are responsible for making sure the people operating these systems have enough knowledge and guidance to use them appropriately.

Lastly, violating these rules can come with serious consequences. Depending on the violation, companies can face millions of euros in fines.

Colorado actually tried to build something close to the EU model back in 2024. They had risk-based rules and oversight on high-risk AI, but it never made it to enforcement. A federal court paused it in April of 2026 after a lawsuit from xAI, backed by the DOJ. A month later, Colorado’s governor signed a full rewrite that removed the parts that mattered most, like preventing algorithmic discrimination, risk management requirements, and impact assessments.

Illinois now requires employers to disclose when AI is used in employment decisions. California made bias testing, or the lack of it before deployment, relevant in discrimination claims. There’s also NYC’s Local Law 144 which requires bias audits for automated hiring tools. These separate laws are great but only apply within those borders.

As you can see, the difference between how the US and EU are approaching AI regulation is pretty significant. The US has AI regulation, but most of it is happening state by state, while the EU created one overarching regulatory framework that determines what companies have to disclose, what higher-risk systems have to document and register, what uses require additional oversight, and what uses of AI aren’t allowed, period.

Companies either based there or doing business there are facing more compliance work, more documentation, and more responsibility, while workers and users are getting more transparency, protections, and ways to hold companies accountable. All of these new requirements could also create more job opportunities in AI governance, safety, auditing, policy, compliance, and risk management because companies need people responsible for making sure their AI systems actually comply with the law.

What do you think? Would you like to see this level of AI regulation applied federally in the US?

Learn more about the EU Artificial Intelligence Act here.

AI is being used across nearly every stage of the hiring process, and it’s showing real benefits.

In applicant review, AI can review more candidates in less time than humans can, giving companies the ability to consider more people for an opportunity.

During interviews, AI can give candidates more flexibility to interview at hours that work for them instead of needing to take time off or accommodate a recruiter’s limited openings. Ribbon reported that nearly 1 in 4 of its AI interviews happen between 10p.m. and 2a.m. their local time.

More opportunity, more accommodation.

However, the lack of transparency around how AI is evaluating candidates throughout the process still creates skepticism.

During applicant review, how and why does AI reject candidates? How does it determine whether someone is qualified or not?

During an interview, what is the AI actually looking for? How is it determining fit outside of good or bad answers? Does it understand nuance and social cues the way a human would? If not, does that make the evaluation fair?

That skepticism is showing up in candidate behavior. Greenhouse reports close to 40% of candidates have walked way from hiring processes that involve AI.

Regardless of the answers, it’s important to remember that humans are still behind these systems. Companies decide what they want the AI to evaluate, what they believe makes someone qualified, how different criteria should be weighed, and how much influence the AI’s output has on the final decision.

While AI may be doing the evaluating, humans are still deciding what the evaluation should value.

AI in Content

Suddenly, snitching on potential AI usage in content is a feature.

LinkedIn, Substack, and Claude have introduced new features and system-level tools that expose or identify potentially AI-assisted published content.

LinkedIn has introduced a ‘Seems like AI Slop’ feature that allows users to report individual posts they believe are low quality, AI-generated content. The feature is supposed to help LinkedIn understand what users are identifying as AI slop and use that feedback to improve what gets distributed in people’s feeds.

The issue here is that just because a post seems low quality or AI-generated, doesn’t mean it is. Some people may prefer this, but overall, I think it can be abused and can lean toward being unfair. Especially to people simply trying to build an audience and find their footing online.

Substack has introduced a new ‘Scan for AI text’ feature that uses an AI detection tool called Pangram to estimate how much of a piece of content may have been written by AI. Those results are then shown to the writer’s readers.

Substack does allow writers to disable the feature. However, it still kind of snitches on writers who disable it because when readers click ‘Scan for AI text’, they’re told that AI detection has been disabled for that publication. Now you’ve potentially created skepticism from readers simply because a writer chose not to participate.

The issue with Pangram is that it’s not 100% accurate. Pangram reports high accuracy rates for its own detector, but those numbers come from Pangram’s own research. Independent research on AI detection has found limitations on these tools, especially as models, writing styles, and editing techniques change.

Personally, I turn this feature off. Regardless of whether it’s writing, coding, or research, I use AI. I’ve broken down how I use AI in this publication before. I write, then give it to an LLM to help make it clearer and more concise. I’m all about AI assistance, not AI doing all the work for me.

Then there’s Claude’s new watermarking feature, where hidden, machine-readable marks are embedded into text generated by Claude models. The watermark is designed to survive things like copying and pasting the text somewhere else. The watermark isn’t something that can be found by highlighting a paragraph or inspecting the page, it’s built into the generated text itself.

Anthropic implemented this in response to the EU’s new AI transparency requirements, but decided to roll this out globally rather than only in Europe.

The purpose isn’t to declare that anyone who uses Claude to produce content is a fraud. People use Claude in all kinds of ways, including to edit or improve content they originally created. The watermark is meant to provide a way to identify that Claude was involved.

What makes me curious is where this goes in the future once other companies can use that information. A watermark could eventually become another indicator used by publishers, platforms, schools, employers, or other organizations to identify undisclosed AI use or to even investigate potential fraud. This could potentially change how comfortable people are using AI to help create content in the first place.

Recruiters are no longer just asking, “Can this person do the job?” They’re also having to ask, “Is this person even real?”

Fake candidates and bad actors aren’t just international scammers applying to U.S. jobs. They’re also real candidates in the U.S. using AI to appear as though they have skills and experience they don’t actually possess.

Many of the verification methods recruiters used to rely on are becoming less effective as AI continues to evolve.

With AI becoming more accessible, bad actors can now generate dozens or even hundreds of resumes tailored to specific job descriptions in minutes, inventing projects, responsibilities, and accomplishments.

Those fabricated experiences can then show up in AI-generated portfolios. AI can create websites, GitHub repositories, design portfolios, presentations, and writing samples that make someone appear way more experienced than they actually are.

Candidates can also use hidden AI tools during video interviews that listen to questions and generate answers in real time. More concerning, AI can be used to create deepfakes and altered appearances using voice-cloning and face-altering technology.

Someone could even combine a real LinkedIn profile, fake work history, AI-generated photos, and fabricated references to create an entirely fictional candidate. If all of that wasn’t enough, AI can also generate fake reference letters, company websites, email addresses, and other materials used to verify experience that never actually existed.

Yikes.

Not only does this elongate the hiring process, but it likely takes away opportunities from genuine job seekers, and can make remote work harder to justify.

How are companies combating this?

No solution is guaranteed to work 100% of the time, whether a candidate is real but lied about their experience or a candidate is completely fake. Companies know that, but they’re implementing processes to at least minimize the risk.

Some companies are bringing candidates on-site for at least one round of interviews, especially the round where they need to prove their skills. Others are asking candidates to present a government ID at some point in the process, even though they know IDs can be falsified.

Many companies now require candidates to keep their cameras on so interviewers can look for signs that someone, or something else, is feeding them answers off-screen. Some candidates may even be asked to wave their hand in front of their face or perform a simple action to help detect deepfakes or manipulated video.

Interviewers are also asking more specific questions and expecting more detailed answers. Anyone can give high-level responses, so the more specific a candidate is about their experience, the harder it is to believe they made it up. Companies are also including more exercises that resemble the actual job instead of asking questions that can easily be answered with AI running on another screen.

This part isn’t new, but recruiters are spending more time cross-referencing resumes against GitHub, LinkedIn, personal websites, publications, and previous employers. Many are also moving reference checks to be earlier in the process so they don’t spend weeks interviewing someone they can’t actually hire.

Companies are also investing in platforms and technology that verify experience or attempt to detect deepfakes, voice changers, and pre-recorded video.

AI is forcing companies to rethink how they evaluate talent, but it’s also forcing them to rethink trust. The question used to be, ‘Can this person do the job?’ Now it’s, ‘Can we verify that this person is who they say they are?’ The challenge will be figuring out how to raise trust in hiring without making the process even more painful for companies and candidates alike.

It's the practice of intentionally trying to break an AI system, get it to say something harmful, leak information it shouldn't, or take an action it's not supposed to take so the people building it can fix the weakness before the public ever encounters it.

Companies building frontier AI models (OpenAI, Anthropic, Google, Amazon, Meta) pay real people to try to break their systems before the public ever sees them. Gray Swan Arena is the platform where a lot of that work happens, and it's open to anyone 18+.

Gray Swan runs competitive challenges where participants try to manipulate AI models into doing things they shouldn't. This includes things like jailbreaking, prompt injection, and agent manipulation. You don’t need a degree or coding background to do this. Most successful attempts at breaking these models come from creative prompting and clear writing, not programming skill.

Major competitions offer prize pools from $20,000 to $170,000 or more. In one recent challenge focused on AI agents, participants collectively ran 1.8 million attack attempts, landed 62,000 successful breaches, and split $171,800 in payouts. Weekly challenges pay smaller amounts but run constantly, so newer participants can earn while they learn.

Companies use this data to decide whether a model is ready to ship and what needs fixing first. Data is gold, including data on weaknesses and vulnerabilities. That $171,800 payout came from a challenge co-sponsored by UK AISI, OpenAI, Anthropic, and Google DeepMind. Out of thousands of participants, 161 walked away with cash prizes.

Participating and being a top performer can also lead to a job. Top performers get pulled into Gray Swan's private red-teaming network, which pays for direct engagements with AI labs. Gray Swan has hired more than 14 Arena winners onto its own team. Independent researcher Ophira Horwitz built her reputation this way, first by exposing a vulnerability in Anthropic's Claude Sonnet-3.5, then by becoming one of only two competitors to crack a Gray Swan model called Cygnet, using playful, upbeat prompts to get past its defenses. She and the other competitor who cracked it both walked away with cash bounties and were later hired as Gray Swan consultants.

The timeline on breaking into a career varies for all participants. One participant told me it took about 12 months of steady work before getting invited into the private network, but others get in faster depending on results.

🔗 Sign up at https://app.grayswan.ai/arena.

Proving Ground runs new challenges every Wednesday across categories like Chat, Image, Agent, and Indirect prompt injection, and it's built for people who are still building up their skills. Every successful break there adds a verified credential to your public profile, which is what labs and companies actually look at when scouting for paid gigs. The Discord runs alongside it where new competitions and payout announcements post before anywhere else.

If you’ve been looking for a way into AI security, this is one of the most direct and accessible paths available right now.

OpenAI analyzed more than 800,000 work related ChatGPT conversations from U.S. business users to better understand how people are actually using AI at work. They found that “44% of occupation-specific ChatGPT requests involved” work outside the user’s inferred occupation. In other words, lots of people weren’t just using AI for tasks associated with their own job, but also using it to help perform work that’s traditionally been associated with other professions or specialists.

OpenAI’s Chief Economist, Ronnie Chatterji, said, “The boundaries between jobs are likely already becoming more flexible due to AI.”

Research suggests boundaries between roles are becoming blurred. Interesting.

So does this new ability to perform work beyond your specialty just become part of your existing role, or will companies redefine roles, promotions, and compensation to reflect those changes?

Today with AI, a recruiter can analyze data, a product manager can draft marketing copy, and a founder can build a simple internal tool that would’ve typically been assigned to another department. The result may not match the quality of a specialist, but AI can likely get them far enough to solve the problem, or produce a strong first draft.

OpenAI didn’t say this means AI is replacing specialists, they just discovered that people were using ChatGPT for more than just general tasks. People were also using it to help create marketing content, troubleshoot software, perform financial analysis, and interpret regulations. These were tasks outside of their inferred occupation which companies normally relied on specialists or dedicated teams to perform.

These possibilities have the potential to change how companies think about roles, promotions, and pay. Typically, taking on responsibilities outside your job description came with a promotion or higher pay. If AI makes those responsibilities easier to take on, will companies still see them as additional work? Or will they simply become part of what’s expected and expand the tasks of the role?

I’d say the biggest question is: if AI expands what’s expected of a role, will compensation expand with it?

Today, candidates are frustrated because they’re submitting dozens or even hundreds of applications and never hearing back. Recruiters are frustrated because a single job posting can receive hundreds or even thousands of applications, making it nearly impossible to review every candidate manually. AI has only amplified both sides of the problem. Candidates are using AI to tailor resumes, prepare for interviews, and apply to more jobs than ever before, while companies are using AI to summarize resumes, search talent pools, and assist with candidate screening.

It’s easy to look at all of this and assume AI is changing hiring. While AI is certainly changing things, I think it’s more accurate to say hiring has always evolved alongside the economy, technology, and the way people work. AI is simply the latest chapter in that story.

Before the internet, and long before AI, getting a job looked very different. Many people found work through family, friends, word of mouth, newspaper ads, or by walking into a business and asking if they were hiring. I remember doing this myself as a teenager. I’d walk into a business, ask if they were hiring, and sometimes get hired on the spot.

Businesses were smaller, hiring was more local, and employers often knew the people they hired personally or through someone they trusted. Many jobs were learned on the job, so employers weren’t always looking for years of experience. In many cases, your reputation mattered more than a resume.

As the United States industrialized and businesses continued to grow, hiring became more complicated. Companies expanded beyond their local communities, jobs became more specialized, and employers had more to lose if they hired the wrong person. At the same time, they were no longer evaluating people they already knew. They were hiring complete strangers, often from much larger applicant pools. That meant companies needed a more consistent way to compare candidates before deciding who to interview.

Around the same time, governments also began introducing labor laws that reshaped the relationship between employers and workers. Regulations around wages, working hours, workplace safety, child labor, and workers’ rights made employment more standardized and formal. While these laws didn’t create resumes or interviews, they did contribute to hiring becoming a more structured process as businesses grew and employers became responsible for complying with an expanding set of employment regulations.

That’s why resumes became so important. They gave employers a standardized way to understand a candidate’s work history, education, and experience. Applications made it easier to collect the same information from every applicant. References provided another perspective from someone who had worked with the candidate before, while degrees and certifications became indicators that someone had developed knowledge in a particular field. Interviews then gave employers an opportunity to validate what they had read and determine whether someone would be a good fit for the role.

Hiring became even more structured following the Civil Rights Act of 1964. Employers increasingly needed to demonstrate that hiring decisions and job requirements were based on legitimate business needs rather than protected characteristics. That led many organizations to adopt more standardized applications, interview processes, documentation, and record keeping, while HR departments took on a larger role in helping companies comply with employment laws.

None of these hiring practices became standard because they perfectly predicted success, but because they were the best tools employers had to reduce uncertainty while hiring at a much larger scale under an expanding set of employment regulations.

For decades, this combination of resumes, applications, interviews, and employment regulations became the foundation of modern hiring. Then the internet changed everything again.

Online job boards made it possible to search thousands of openings instead of relying on newspaper ads. LinkedIn transformed professional networking into an online recruiting platform. Applicant tracking systems gave employers a way to organize and manage growing numbers of applications. These innovations made it easier for candidates to find opportunities almost anywhere, but they also made it easier to apply for more jobs than ever before.

That shift created a new challenge - companies no longer struggled to find applicants, they struggled to identify the right hire from an overwhelming volume of applications.

Today, AI is accelerating that issue. Candidates can create tailored resumes and submit applications in minutes, while employers are using AI to help process the increasing number of applications they receive. In many ways, both sides are responding to the same problem from opposite directions. Candidates are trying to increase their chances of getting noticed, while employers are trying to identify qualified candidates without spending weeks reviewing every application manually.

So where do I think hiring and looking for a job are going?

I don't think resumes are disappearing, but I do think their role is changing. As AI makes it easier to generate resumes and apply for more jobs, employers will continue looking for stronger evidence that someone can actually do the work. At the same time, I think both candidates and employers will increasingly rely on AI to search, evaluate, and navigate the hiring process more efficiently.

I also expect employment laws and regulations to continue evolving alongside these changes. As with previous shifts in hiring, governments are already responding. New laws and regulations, such as New York City’s AI hiring law and the European Union’s AI Act, are beginning to address how AI can be used in employment decisions, with a growing focus on transparency, bias, auditing, and human oversight. Every major shift in hiring has eventually been followed by new expectations for how employers should use it responsibly.

I don’t think the story on the evolution of hiring is about resumes, applicant tracking systems, or AI. I believe it’s about how employers have continually adapted to find the right people as work, technology, and society have changed. I expect that evolution to continue long after today’s hiring tools have been replaced by whatever comes next.

Let's use the lawsuit against Workday involving alleged applicant discrimination as an example. While the allegations have not been proven, the case provides an opportunity to understand the types of issues investigators may examine in situations like this.

If a company is training or fine-tuning AI using historical hiring decisions or similar organizational data, the system may learn patterns that reflect biased, outdated, or even unlawful practices if those patterns aren’t identified and addressed. For example, if a company has historically rejected applicants whose employment history or graduation dates suggest they are older, candidates with non-white sounding names, or applicants without degrees, those patterns may be reflected in its historical data. If that data is then used to train or fine-tune an AI system without proper safeguards, the system may learn those patterns and apply them at scale.

If a company is adopting a third-party AI system instead, the focus shifts to how that system was configured, tested, monitored, and relied upon. While the company may have no control over how the AI was trained, adopting it doesn’t eliminate responsibility. Organizations should perform due diligence to understand how the system was built, whether it has been tested for bias and reliability, and what guardrails and accountability measures exist to mitigate harm. Without that due diligence, a single AI tool can scale harmful outcomes across thousands of organizations.

Completely eliminating bias from technology built by humans is nearly impossible, which is why intentionally designing systems to identify, mitigate, and monitor it is imperative.

Then there's human error and bias, which still exist whether AI is involved or not. Even when AI is designed to assist with decision-making, humans often have the final say. If those decision-makers are biased, the outcome can still be discriminatory regardless of what the AI recommended. We've already seen the harm humans can cause long before AI existed. While human bias can be mitigated through thoughtful processes, guardrails, and accountability, it still can't be completely eliminated.

AI may seem like the bad guy, but AI doesn't understand fairness, ethics, or discrimination, it learns statistical relationships in data. Without explicit guardrails made by humans, it has no way of knowing whether a historical pattern reflects good business decisions or practices that should never be repeated.

AI shouldn't be judged by whether it's perfect, because no human-built system ever will be. AI should be judged by whether it's demonstrably fairer, more transparent, and more accountable than the process it replaces. Even then, the responsibility doesn't stop once an AI system is deployed. Organizations should continuously test, monitor, and improve these systems as new risks surface.

This is a rare opportunity to shape AI products from the ground up, and have direct influence on how nonprofit newsrooms in our portfolio approach product development and applications of technology. You’ll have direct access to users, clear mission alignment, and the autonomy to make real product decisions in service of rebuilding sustainable local news across the country.

📍Remote in U.S.

💰$136,341 – $149,975

Apply here 🔗

How Product Managers Are Using AI

Product Managers are responsible for a product’s vision and success from the idea stage to launch and beyond. They use research and data to identify opportunities, define priorities, and turn insights into strategy, design, and execution. Their work consists of collaboration, organization, and a lot of communication.

Since so much of the role involves processing information, coordinating people, and making decisions, AI is becoming a very useful tool for them.

Ways AI helps Product Managers save time:

• Turning product reviews, stakeholder meetings, customer interviews, and sprint planning sessions into concise summaries that highlight key decisions, action items, blockers, and areas requiring follow-up.

• Summarizing research findings, long email/message threads, feedback from customers, test users, etc.

• Scanning large volumes of product analytics and synthesizing important insights from clicks, drop-offs, user behavior, usage patterns, etc.

• Quickly turning large product requirements documents, meeting notes, and strategic plans into detailed user stories, checklists of what a feature needs before it's considered complete, Jira tickets, and development tasks.

• Generating or refining emails for customers, stakeholders, and cross-functional teams to communicate product updates, gather feedback, share decisions, and coordinate next steps.

• Creating project status reports by analyzing information from tools like Jira, meeting transcripts, Slack, and product roadmaps to summarize progress, blockers, risks, and next action items.

Ways AI helps Product Managers discover more:

• Reviewing and summarizing customer interviews, surveys, support tickets, app reviews, sales calls, and social media feedback to identify recurring complaints, feature requests, unmet needs, and areas where customers are struggling.

• Analyzing product data to understand how people actually use a product, including which features they engage with most, where they abandon workflows, how often they return, and which actions are linked to long-term retention.

• Identifying patterns and trends across thousands of customer interactions and data points that would be difficult to spot manually, helping uncover emerging opportunities, potential risks, and changes in customer behavior.

• Grouping customers based on behaviors, preferences, usage habits, and engagement levels to better understand how different types of users experience the product and where their needs differ.

• Monitoring competitors to track new feature releases, pricing changes, product updates, positioning shifts, and broader market trends that may impact product strategy.

• Generating ideas for new features, product improvements, experiments, and areas for further investigation by combining customer feedback, product analytics, competitor research, and market trends.

Ways AI helps Product Managers do things they normally couldn’t:

• Ask questions about product data in plain English and get answers instantly without needing to know SQL or rely on a data team to build reports.

• Analyze thousands of pieces of customer feedback at once, making it possible to understand the experiences and opinions of far more customers than a person could realistically review manually.

• Quickly generate dozens of potential feature ideas, experiments, and solutions for a problem, allowing teams to explore more possibilities before deciding what to build.

• Continuously monitor competitors, market changes, customer sentiment, and product performance at a scale that would be impossible for a single product manager to track manually.

• Simulate the potential impact of product decisions using historical data, helping teams evaluate trade-offs and outcomes before committing resources.

• Act as an on-demand research assistant that can instantly summarize information, answer questions, compare sources, and surface relevant context across thousands of documents.

Ways Product Managers still rely on humans, not AI:

• Deciding which problems are worth solving and which opportunities are worth pursuing.

• Making trade-offs between customer needs, business goals, technical constraints, budgets, and timelines.

• Building relationships and trust with customers, stakeholders, executives, and cross-functional teams.

• Leading discussions, resolving disagreements, and getting teams aligned around a shared direction.

• Understanding the emotions, motivations, and context behind customer feedback that may not be obvious from data alone.

• Making judgment calls when data is incomplete, conflicting, or doesn’t tell the full story.

• Communicating a product vision and inspiring teams to execute on it.

• Taking responsibility for product decisions and their outcomes.

Sam Altman and Dario Amodei are both walking back their AI jobs apocalypse prophecies as they eye blockbuster IPOs

The same leaders who spent years warning about massive job displacement and the need for society to brace for AI's impact on work are now softening those claims.

The part that bothers me isn’t that they may have been wrong - predictions can be wrong. What bothers me is that these predictions were treated like facts.

Workers made career decisions based on them, students reconsidered what fields to enter, companies forced AI usage upon employees and pointed to AI when announcing layoffs and restructurings. Entire industries shifted their strategies around expectations that have yet to materialize, only for the narrative to change.

So why is it changing?

What’s particularly interesting is the timing of it all. OpenAI and Anthropic are no longer startups trying to convince us all that AI is the future. They’re some of the most valuable private companies in the world, with massive commercial ambitions and potential IPO’s on the horizon.

I can't verify if them walking back their claims and pursuing larger commercial ambitions are related. Their views could have genuinely changed as more information became available, but it's hard not to notice that the conversation is shifting away from workforce disruption and back to the original "productivity gains" narrative at the same time these companies are trying to sell AI to enterprises, governments, investors, and eventually public markets.

I don’t know, but my point is, this is exactly why it’s important to distinguish between hype and data.

In last week’s newsletter, I wrote about how AI companies, investors, and corporations have enormous incentives for AI to succeed. Their incentives don’t automatically make them wrong, but they do mean we shouldn't take every prediction at face value. Predictions are not verified data, and knowing who is pushing a narrative is important before making critical decisions.

AI can be an incredible, fascinating technology, but we don't need to invent conclusions before the data arrives.

NEW JOBS💰

1. American Journalism Project | Senior Technical Product Manager | Remote in U.S. | $120,000 - $160,0000

2. Kentik | Sr People Business Partner | Remote in U.S. | $180,000 - $230,000

At some point you start rolling your eyes.

If you actually dig into these stories, look at who’s publishing them, examine the data, and build with AI yourself, you start questioning what these headlines and leaders are actually saying.

AI is genuinely impressive at speed. Whether you’re writing, building a website, or shipping a product, it gets you there faster and sometimes cheaper. But fast doesn’t always mean better, and it definitely doesn’t guarantee success.

Let’s look at the business productivity data.

A February 2026 NBER study surveying nearly 6,000 executives across four countries found that 69% of firms actively use AI, but nine-in-ten reported no impact on employment or productivity over the past three years. Executives use AI themselves, but average only 1.5 hours a week. Companies are using AI for tasks, but aren’t seeing business results, which is supposed to be the whole point.

The same study notes these executives predict AI will boost productivity by 1.4% and cut employment 0.7% over the next three years. Those predictions, not proven results, are driving a lot of the layoff decisions happening right now. Companies are making workforce decisions based on expectations rather than measurable outcomes.

The Federal Reserve Bank of San Francisco published a letter stating that most macro-studies find limited evidence of a significant AI effect on productivity growth, and that even firms who find AI useful show little evidence of transformative gains. The letter used this analogy to describe the impact of integrating AI in parts of the business: using AI to automate parts of a process without rethinking the whole operation is like replacing a steam motor with an electric one but leaving the factory floor unchanged. Progress, but not transformation.

Firms are still adopting AI and learning tools rather than restructuring how they actually work. So either it’s a learning curve or AI just isn’t where these claims say it is.

Beyond productivity gains, plenty of data shows AI has been used as a cover for layoffs. Companies are still ‘trimming the fat’ from COVID overhiring, laying off American workers to offshore for cheaper talent, and some are genuinely experimenting with replacing workers before any AI gains have materialized, which has been resulting in rehiring, often not the same workers they let go.

Challenger, Gray & Christmas tracks public layoff announcements and found AI is the third most common reason cited for job cuts in 2026, and even then, most of those cuts are to fund AI investment, not because AI has successfully automated those workers away. Their own chief revenue officer said “it’s difficult to say how big an impact AI is having on layoffs specifically” and noted the market rewards companies that mention it. Sam Altman said it himself, there’s “some AI washing where people are blaming AI for layoffs that they would otherwise do.”

Framing all of this as AI-driven efficiency isn’t accurate, it just sounds good to investors and anyone with a financial stake in AI winning.

AI is impacting the workforce, but the reasons being presented in these headlines are where I stop taking them seriously. I use AI throughout most of my day because it’s become so central to my work, but using it also makes me question how companies are supposedly replacing workers at scale given how much oversight it still requires and the limitations it still has.

AI is changing work, but many of the claims about why don’t match the productivity data. They mostly favor the people financially positioned to benefit from that narrative.

NEW ROLES📌

1. Kentik | Account Executive, Enterprise - NY Metro | Remote in U.S. | $140,000 - $160,000 with $280,000 - $320,000 OTE

2. Kentik | Customer Success Manager - US EST | Remote in U.S. | $100,000 - $120,000 with $125,000 - $150,000 OTE

A ‘moat’ is a phrase used to describe your differentiator. What makes your product the best? How will it survive amongst competitors?

One thing the investors made clear is that they don’t feel confident investing in products that are just GPT wrappers, which makes sense. If your product is just an overlay of another product, why is that special or even safe? What’s to stop the GPT company from building your product as part of their ecosystem? Back in the day before AI was commercialized, a similar issue were companies that were reliant on other companies to run the core parts of their product. If anything ever happened to those other companies, they’d collapse. Not a smart investment.

Realistically, most companies are GPT wrappers because most people aren’t technical. They’re experimenting and innovating to the capacity that they’re able to, hoping to build and launch something undeniably good.

So what IS a moat these days?

Access to building is no longer a moat. Anyone can build what looks like your competitor in a day now. They can collect the same data, build similar features, and market themselves to your same customer. Not impressive, and also not enough to make you worry. Building something that works really well and will last is. But in order to know it works well and won’t die as a trend, you need users. How do you get those?

A moat today is going to be who you are, your audience or network, the trust they have with you, your credibility, your expertise, and your product’s reach. Keep in mind, this has always been a moat, it’s just weighed so much more heavily now.

So if you are building a product, or planning to, you need to prioritize the old rule book. Create an online presence, get consistently involved with events and communities that allow you to grow your network, and push your name and product out there. Regardless of whether you’re hoping to raise or not, these are things you should be doing to increase awareness around what you’re building.

I know this because I’m doing it right now. As someone who launched a successful bootstrapped business once and gained credibility because of it, I’m practicing these same moat strategies with my new business, Sift Skills. Although I’m building in a space adjacent to Black Tech Pipeline, I have to build a reputation with a new audience that trusts me, my experience, and my expertise to deliver the best solution to a major problem they have.

So what have I been doing? I make accounts on all event platforms that bring awareness to startup and networking events happening in my city. I get involved with the startup and entrepreneur community. I made a goal to attend at least one event per week, whether it’s a pitch competition, networking event, or panel. I show up, show my face, meet people, and tell them what I’m building. The more you attend events, the more you also get invited to.

I stay active in online communities, especially Reddit. I post opinions and updates on LinkedIn and interact with people seen as leaders in the space I’m trying to break into. This newsletter is even a distribution channel. Come to our table to say hi and test Sift Skills yourself in-person on May 26th here. See? 😉

My take on moats in the age of AI is this: first, you don’t need to build something with the goal of getting investment. You can bootstrap an incredibly good product without it. If your bootstrapped product is a wrapper that has users and revenue, it can still catch an investor’s attention regardless of what they say on panels or in articles. At the end of the day, a good product is a good product, and people are going to want in.

Since you are the moat, this isn’t the era to be shy, and it isn’t the era to be inconsistent with relationships either. Be intentional about meeting your users where they are, ask them questions, talk about how you’re helping them on every channel you’d find them on, build community, stay active, and be kind.

And remember, your moat won’t save you if your product sucks. People don’t want to attach their hard earned money to something that only looks like it works, regardless of how much they like you as a person. Whatever you’re building needs to be valuable and do what it says it does, and do it well.

New Role📌

1. Kentik | Sr Product Manager, Platform | Remote in the US | $190,000– $225,000

I’ve been experimenting with different strategies, and none of them let you remove yourself from the process. No tool or service that claims to handle it for you actually works well. What does work is having a smarter system. Here’s one that at least helps:

1. Set LinkedIn job alerts but don’t apply on LinkedIn:

I run around 10-15 LinkedIn job alerts using Boolean strings that Claude and ChatGPT helped me build. My phone goes off all day, and while that gets annoying, it surfaces opportunities quickly. Here’s how to set it up:

Using ChatGPT, Claude, or both, upload or paste an anonymous version of your resume and submit this prompt:

Here is my resume: [paste or upload resume]

Roles I’m targeting: [titles, seniority level]

Industries I want: [list]

Industries I want to avoid: [list]

Role types from my past I do NOT want: [list]

Main Task — Targeted Roles

Generate 5 to 10 LinkedIn job alert search strings I can paste directly into LinkedIn’s job search bar.

Rules:

• Only use AND, OR, and quoted phrases

• Only use job titles that appear in my resume or in “Roles I’m targeting”

• Minor title variations are allowed (e.g., “Program Manager”, “Program Lead”, “Director of Programs”, “Programs Director”)

• Do not assume or infer anything beyond what is provided

• Reflect industries I want using relevant keywords

• Avoid unwanted roles by not including related titles or keywords

Additional Task — Role Discovery

Identify 3–5 additional roles I may be a strong fit for based on my resume.

Rules for this section:

• You MAY introduce new role titles, but they must be directly supported by my experience

• Do not suggest roles that require skills I have not clearly shown

• Stay within my preferred industries

• You may include adjacent industries only if the role function is identical and clearly transferable

• Do not introduce entirely new industry categories

• Avoid roles I explicitly said I do not want

• Prioritize roles adjacent to my experience (not drastic pivots)

Then for each suggested role:

• Provide the role title

• One-line justification tied to specific experience

• 1 Boolean search string using that role

• Include 2–4 relevant industry keywords in the Boolean string

Output Format

Section 1: Targeted Roles

• Boolean string

• One-line note on what it surfaces

Section 2: Suggested Roles

• Role title

• One-line justification

• Boolean string

Paste the generated boolean strings into LinkedIn’s job search bar, click on the ‘jobs’ only page, then apply your filters before saving it as an alert. Set it for roles posted within the past 24 hours, your preferred locations, and anything else relevant. Keep in mind: the more specific your filters, the fewer alerts you’ll get.

Once you find a role worth pursuing, go directly to the company’s site to apply. Don’t apply through LinkedIn.

2. Tailor your resume for the jobs you’re applying to

Once you’ve found roles worth pursuing, the next step is tailoring your resume, and no, there’s no tool that does this for you in a way that actually works. What I built gets you closer though.

Resume Tailor is a free GPT built into ChatGPT (no downloads or sign-up required) that takes your resume and a job description and gives you:

• Newly developed bullet suggestions to make you more appealing for the role

• An honest read on how well you actually fit, strengths and gaps included

• Ways to elevate your transferable skills

• Guidance on how to position your skills and experience

• Examples on how to strengthen specific bullets for that specific role

• A way to turn your resume from a list of tasks into a cohesive story of your history

Tailoring is exhausting and time consuming. This won’t do it for you, but it’ll cut down the time it takes to figure out where to start.

Check out Resume Tailor🔗

3. Don’t apply without first finding out who in your network can refer you

A referral is still the most effective way to land a job. Before you apply anywhere, check whether anyone in your network is connected to that company.

Here’s how to find out directly on LinkedIn:

1. Go to your personal LinkedIn profile

2. Click on your ‘connections’ in your LinkedIn header

3. Click ‘search with filters’

4. Click the ‘Current companies’ filter

5. Type the company you’re interested in and click ‘Show results’

This filters your connections by company affiliation. Even if you don’t know them well, reach out. A referral from an acquaintance still carries more weight than a cold application.

Here’s a message you can send even if you rarely interact:

“Hi [Name], I saw the [role title] opening at [Company] and I’m applying.

Quick context: I have experience in [1–2 core relevant things], and this role lines up well.

If you’re open to referring me, I’d really appreciate it. If not, no worries at all.

Happy to send my resume to make it easy.”

None of this totally removes the work of the job hunt, but incorporating these steps makes the process less chaotic.

Open roles by Kentik on the BTP job board 📌

1. Go-To-Market Recruiter | Remote in U.S. | $108,000– $130,000

2. Principal Product Marketing Manager | Remote in U.S. | $157,000 - $210,000

3. Sr Product Marketing Manager | Remote in U.S. | $145,000 - $190,000

4. Product Manager, Synthetics | Remote in U.S. | $160,000 - $180,000

5. Prototype Design Engineer | Remote in U.S. | $170,000 – $210,000

6. Sr Product Manager, Kentik NMS | Remote in U.S. | $190,000– $230,000

7. Sr Product Manager, Platform UI | Remote in U.S. | $190,000– $225,000

8. Staff Software Engineer, NMS | Remote in U.S. | $190,000 – $225,000

9. Infrastructure Strategy & Delivery Manager | Remote in U.S. | $163,000 – $220,000

10. Staff Site Reliability Engineer, Cloud | Remote in U.S. | $165,000 - $200,000

11. Staff Site Reliability Engineer, Platform | Remote in U.S. | $165,000 - $200,000

12. Sr Backend Engineer, Platform Engineering - Network Data | Remote in U.S. | $180,000 - $210,000

13. Sr Manager, Engineering - Platform Engineering - Ingest & Network Data | Remote in U.S. | $230,000 – $275,000

14. Sr Staff Backend Engineer, Platform Engineering - KDE Storage | Remote in U.S. | $200,000– $245,000

15. Staff Backend Engineer, Platform Engineering - KDE Query | Remote in U.S. | $190,000 – $225,000

16. Director, Global Customer Success Service Provider | Remote in U.S. | $200,000 - $230,000

17. Solutions Engineer, Enterprise - East | Remote in U.S. | $133,000 - $168,000 with $190,000 - $240,000 OTE

Not understanding the difference leads to overbuilt, unreliable AI systems.

Understand how each works and the risks before you decide.

A prompt wrapper is a tool or product built on top of an LLM. You give it a specific task, it follows predefined instructions, and returns a single output—predictable, consistent, and repeatable. A tool that summarizes your meeting notes and hands you a recap is a prompt wrapper.

An agent is given an overall goal, not just a task. Agents decide what steps to take, use tools to get them there, and keep working until they reach an outcome. A tool that prepares you for a meeting by gathering background, summarizing key points, drafting talking notes, and suggesting follow-ups is an agent. The risky thing here is that it can change course mid-task if it believes a new path improves the outcome, but there’s no guarantee the new path is actually better.

Agents and prompt wrappers can hallucinate, misinterpret instructions, and produce inconsistent results. The difference is that with agents, those errors compound and you likely won’t find out until it’s too late.

If an agent misunderstands your company’s priorities early in that meeting prep, it pulls the wrong information, summarizes it poorly, and builds your talking points around it. You walk in prepared, but with the wrong focus. Luckily, you’d probably catch that after the one meeting and revisit your agent. Now imagine the same compounding error in an agent tracking employee performance to inform promotions. Compounding incorrect data doesn’t just affect one meeting, it affects how someone gets evaluated.

This doesn’t mean prompt wrappers are better or that agents should be avoided. Most real world AI systems use both—agents orchestrate the workflow while prompt wrappers handle the focused tasks inside it. Regardless, you need to build a system that holds both accountable.

How to hold AI agents accountable

Agents make decisions, which means you need to control how far they can go.

Be clear about the goal. Agents don't interpret intention the way a person would. If the goal is vague, the agent fills in the gaps on its own, and what it assumes is rarely what you meant. "Prepare me for this meeting" leaves too much open. "Summarize the company's Q3 challenges and draft three talking points I can use to propose a partnership" gives it something specific enough to actually execute against. The more precise the goal, the less room for the agent to go in a direction you didn't intend.

Limit what it can access. Agents use whatever you give them access to. If that's your entire inbox, it might pull from an unrelated thread and treat it as relevant context. If it's all company data, it might surface information you didn't intend to include. Access should match the task. Prepping for a vendor meeting? Give it the vendor contract, your last call notes, and the relevant product docs. Nothing else. This isn't just about accuracy, it's also about security. The more an agent can see, the more it can expose, especially if something goes wrong or the output ends up somewhere it shouldn't.

Break the task into steps. Instead of "prepare me for this meeting," give it a sequence: gather background, summarize key points, draft talking points, suggest follow-ups. One long instruction is harder to audit. Steps make it easier to see exactly where things went wrong.

Verify the output. After it produces something, add a check. Ask a second system or a reviewer: are these points supported by the source material? Did anything get misinterpreted? Don’t assume a clean output means a correct one.

Log what it did. Keep a record of what you asked the agent to do, what it returned, what steps it took, what sources it used, and what decisions it made. Tools like Langfuse, Helicone, or GitHub version tracking can help, or you can log manually in a spreadsheet. You should save the results the agent produces instead of deleting old ones to replace them. You want the ability to compare versions over time so that it’s easier to spot drift.

Keep humans in the loop on anything high-stakes. Agents optimize for completing the goal, not for being right. They don’t know what they don’t know, and they won’t flag uncertainty the way a person would, they’ll produce a confident output either way. For low-stakes tasks like summarizing notes or drafting a first pass at something, that’s fine. For anything that affects a person’s job, compensation, performance record, or opportunities, a human needs to review it before it goes anywhere. Not as a formality, as an actual check. The agent’s output should be a starting point, not a final answer.

How to hold prompt wrappers accountable

Prompt wrappers follow instructions. If the instructions are weak, the output will be too.

Be specific about what you want. “Summarize this” tells the model almost nothing. It doesn’t know how long, what to prioritize, or what format you need. “Summarize this in 5 bullet points focused on key decisions” gives it a clear target. The more specific the prompt, the less the model has to interpret, and interpretation is where things go sideways.

Tell it exactly how to format the answer. Prompt wrappers will default to whatever structure feels natural to them, which may not be what you need. If you want bullet points, say so. If you don’t want paragraphs, say that too. “Use bullet points. No paragraphs. Keep each point under one sentence.” Structure left open is structure you don’t control.

Make it use only what you give it. Prompt wrappers try to be helpful, which means they may fill in gaps when information is missing rather than admit they don’t have it. That’s how you get confident-sounding output that has nothing to do with your actual document. In your prompt, tell it explicitly: “Only use the information in this document. If something is missing, say ‘not provided.’” That one instruction cuts a lot of hallucination.

Run the same prompt multiple times. If the answers vary significantly across runs, the prompt isn’t tight enough. Consistency shows the prompt is doing the work, not the model guessing. One technique that's worked well for me is running the same prompt across two different LLMs, then using each model's output as feedback for the other. Take what the first one produces, show it to the second, see where they disagree or diverge, and use that to refine the prompt. Keep going until the results get consistent. Some of my strongest prompts came out of that process. (someone should build a tool for that 😉)

Set rules for missing or unclear inputs. Tell it what to do when information isn’t there. “If the document doesn’t include this, say ‘no evidence found’ instead of guessing.” Without that instruction, a model will often produce something convincing rather than nothing, and something convincing yet false is worse than a blank because you might not catch it.

Save versions of your prompts. When something stops working or starts producing different results, you need to know what changed. Keep a log of your prompts and note what you adjusted and why. It’s easy to iterate yourself into a worse prompt without realizing it. You can use the same tools as the ones suggested for agents if you know what you’re doing technically. Otherwise, documenting it manually in a spreadsheet or note-taking app is fine.

Check the output against the source. Don’t assume what was generated is accurate just because it looks right. Ask yourself: is every point in this summary actually supported by the document? The model isn’t verifying its own work, and neither is anyone else unless you do.

Remember it this way: Agents need boundaries. Prompts need precision.

Application answers are exactly the same, resumes are filled with keywords but missing the actual skills and experience the role requires. Tools like Quick Apply and AI job application apps made it easier than ever to apply to hundreds of jobs at once, and the volume alone is a problem. Candidates who actually put in the work and are qualified are getting buried under all of it.

So what’s the solution? According to publications like Business Insider and The Washington Post, employers are no longer relying on resumes and are asking applicants to prove their credentials. Sounds a bit familiar.

They’re doing things like work trials where you onboard with the company for a few days or a week, get paid, and work alongside them. This isn’t a new concept at all, but it’s slowly becoming a bigger practice in the age of AI to verify the skills a candidate claims to have, and ensure they work well with the team.

I’ve always appreciated this practice because you never truly know if someone is a good fit until they’ve begun working at the company. This gives candidates the opportunity to prove their skills, receive pay for their time and contribution, and get to know potential team members. However, realistically, this approach is not scalable. More companies can’t implement this than those who can. While it’s more optimal for retention purposes, it’s time-consuming and costly.

Some companies are doing things like ‘skill-based hiring’, which sounds like normal hiring but rebranded, where they still require a resume but don’t take it as truth or allow it to hold much weight. The reason why I say this is just normal hiring rebranded because a resume has always only been the first part of the process. Candidates have needed to prove their skills throughout long interview processes before getting to the offer stage anyway.

Lastly, some employers aren’t requiring resumes at all. Instead, they’re prioritizing application questions that require more unique responses. Of course responses can still be AI-generated but this reduces the redundancy of the same exact answers coming through. This is a great approach, but again, this isn’t scalable for companies on a time-crunch, hiring for tons of roles.

The reality is that resumes aren’t going anywhere. The solutions they’re introducing have either always existed, or cannot be applied to most companies. If anything, the ancient practice of relying on referrals remains the best, and most trusted way, to find talent. That is, as long as your workforce has the ability to refer a diverse, innovative pool of candidates.

But maybe someone out there is building a better solution…*wink*

OPEN REMOTE ROLES💰

1. Kentik | Sr Product Marketing Manager | Remote in US | $160K – $170K

2. Kentik | Solutions Engineer, Enterprise - West | Remote in US | $133K - $168K with $190K - $240K OTE

3. Kentik | Solutions Engineer - EMEA | Remote in EU

4. Kentik | Staff Software Engineer, NMS | Remote in US | $190K – $225K

5. Kentik | Senior Technical Writer | Remote in US | $110K – $135K

Check out more remote roles by Kentik here!

Every job has core requirements (non-negotiables) and nice-to-haves (optional). Nice-to-haves rarely outweigh core requirements.

In many job listings, you’ll also see generic bullet points like ‘excellent communication skills’, ‘ability to work cross-functionally’, ‘build and maintain relationships’, or ‘demonstrated leadership ability.’ These traits matter in almost every job which is why they normally don’t move the needle on a resume. However, if you’re heavily reliant on them to stand out, listing them as outcomes may help. Here are some examples of how you need to present them if this is the case:

• Presented program updates and performance insights to internal leadership, influencing changes to workflows and priorities

• Coordinated across operations, support, and product teams to deliver process improvements that increased efficiency

• Built and maintained relationships with clients and internal stakeholders, improving engagement and reducing recurring issues

• Led training and onboarding for new team members, improving ramp-up time and consistency in team performance

That’s the difference between telling a recruiter you’re capable and showing them evidence.

The purpose of transferable skills is to show your potential. In this job market, potential is mainly reserved for entry/junior level talent and becomes less flexible the more senior you become. If you want your transferable skills to hold weight, you need to focus on your skills and experience that have functional equivalence to the role. The goal is to show, ‘Even if I haven’t done this exact job, I’ve already done the same type of core work successfully.’

If you’re having trouble identifying the core requirements, paste the entire job description into ChatGPT or Claude and ask it to break down what the role actually is and what skills/expectations the JD states are necessary to do it well, in plain English.

Below is a high-level example of a candidate who isn't a perfect match for the role, but whose experience is positioned clearly enough that a recruiter can see the adjacency. This is not a resume writing guide, just an example of how transferable skills can be framed without inflating or minimizing what you've actually done:

Better presenting your transferable skills isn't a guarantee you'll be considered a strong fit. If you're being overlooked and you believe you're a genuine fit, the problem is usually how you're presenting your experience, even when it’s mainly transferable. You don’t need to exaggerate your experience, but you also can’t afford to present it in a way that undersells what you’ve actually done.