Todd Sawicki

As leaders in this industry, we have to think beyond just what’s possible with AI. We also have to think about what’s responsible. In this conversation, Justin Riley pointed out something that contractors need to be aware of. Consent for messaging does not automatically mean consent for AI voice outreach. That distinction matters. Especially as more businesses start implementing these tools. Watch the full episode: https://lnkd.in/gNEWjTCQ Listen to the full episode: https://lnkd.in/gih7aJAD

3

1 Comment

The SMS interactions between brand & consumer contain incredibly rich context for 1:1 personalized commerce experiences. Postscript is the platform enabling such experiences. On this week's episode, Rishabh and I sat down with Adam Turner to explore how SMS evolved from an abandoned-cart afterthought to a 1:1 sales & marketing channel. We covered: - Category creation and the rise of the “SMS marketer” - Compliance and quality as product design principles - Moving from broadcast to conversational journeys - Building distribution in the Shopify ecosystem, then scaling a sales-led motion - Where AI has augmented vs reinvented both marketing products and teams Link in comments

31

4 Comments

Excited to share that Swivel (formerly PilotDesk) has raised a $5.8M Series A, co-led by Tribeca Venture Partners and Ardent Venture Partners. Aperiam was an original Seed investor and participated in this round as well. Swivel is led by Joseph Hirsch and the team behind SpringServe, the CTV ad server acquired by Magnite. They’re now applying that experience to a massive and underserved opportunity: automating ad ops and account management using AI. This is one of the areas I am personally very excited about when it comes to the impact of AI. Swivel’s platform is already being used by leaders like LG Ads to streamline campaign workflows, reduce errors, and free up teams to focus on strategy - not spreadsheets. The rebrand from PilotDesk to Swivel reflects that vision: no more swivel-chair processes between disconnected systems. Just smarter, faster execution. Big congrats to Joe and the team. We’re proud to support them. Link to announcement below.

246

14 Comments

Childcare is a movable barrier and the care economy is workforce infrastructure. Siran Cao dropped so much truth on First Builders this week and shared why even capitalists in a vacuum should care.

8

Jeff Green / The Trade Desk getting into the CTV OS space is a smart move. There’s a lot of mid-size CTV providers globally that could benefit from a solution that isn’t from big tech, but is from a company that can max their inventory yield. If you’ve used any CTV app recently, you’ll know the UX has a lot of room for improvement, and if they have ads - they are often irrelevant.

2

#videogameUA is ripe for a discovery and an engagement quantum leap. Playcast is harnessing idle compute and game rewards to bring a streaming play across Meta Google TikTok Discord … unlocking instant play ads for #mobile, #PC and #UGC developers… try it now from our homepage Thanks for the fun chat Rick Howe !

14

2 Comments

Go-to-market isn't just a launch plan. It’s execution. It’s relationships. It’s staying close to the real problems operators face. The partnerships we've built are a great example of what happens when strategy and field expertise come together. This isn't just about delivering technology. It’s about delivering outcomes, for the railways, for safety, for the people who rely on them every day. Looking forward to building some new partnerships in the future. Thanks to Tim Flower and Petra Holečková for championing the charge. #Partnerships #GoToMarket #RailTech #AIinRail #SmartInfrastructure

16

Tech vs Service. Today, we manage two companies. Pontiac Intelligence: A Demand Side Platform to buy advertising space. A technology company. Clients pay a transparent licensing fee. Programmatic Mechanics: A trading desk model where we provide high quality traders and data scientists for organizations looking for personnel assistance in managing campaigns. Comes with dashboarding service across all channels. We spend a lot of time discussing the value of a service business versus technology. And, I just think the times are changing. It's not that technology is dead. It's just that really high quality service models might be able to retain clients for longer, and that might start to even the valuation playing field. Technology isn't dead. The ability to switch between technologies is easing. The moat is drying out. Every single technology companies moat is drying out.

30

4 Comments

California just handed cities a new enforcement tool to track your Airbnb. SB 346 took effect January 1, 2026. Now any California city can demand that Airbnb and Vrbo turn over detailed information about every listing: addresses, booking nights, and compliance status. This isn't theoretical. Cities like Los Angeles and San Francisco have been waiting for exactly this kind of authority. If you operate STRs in California, the compliance conversation starts with your registration. Is it current with each municipality where you operate? Does your listing display the correct permit number? Local occupancy limits vary significantly by city, so what flies in San Diego might get you fined in Santa Monica. Then there's the tax question. Platforms collect state lodging taxes automatically, but many local taxes are still your responsibility to remit. Documentation matters when audits come knocking. The broader trend is clear. 2025 saw more states tighten platform accountability. Louisiana now requires Airbnb and Vrbo to collect and remit sales tax as accommodation intermediaries. New York counties are building registries with detailed booking data requirements. Here's what I'm seeing with STR operators who stay ahead of this: they're using AI to handle the communication burden that compliance creates. Automated guest messaging about local quiet hours, occupancy limits, parking rules. Compliance gets you legal. Automation gets you sane. The operators who survive 2026 won't be the ones who ignored the regulatory shift. They'll be the ones who got compliant and then freed themselves from the operational weight of staying that way. Is your STR portfolio ready for January audits? And more importantly, is your communication system ready to support the guest experience while you focus on the business side? If you're curious what AI-assisted guest communication looks like for STR operators, I'm happy to walk through it. Drop a comment or send me a DM.

2

The "Earn-to-Own" transition is a common theme in emerging studios. Many venture studios get their start as service businesses. Ad agencies, dev shops, engineering firms. It makes sense when you think about it - they've played a key role in the venture-building process, but favored the consistent cash flow in lieu of owning the assets. Then, they have the epiphany: their business already has many of the elements needed to build a venture studio, so why not make the move? Here's the thing: navigating this transition is harder than it sounds. Many studios begin with a plan to offset capital needs with services work. When it's time to fund raise for the studio, a clear conflict emerges - and it's often too late to do anything about it. Investors want to know that the studio's resources are going toward venture-building and not client work. Without a clean separation, the studio never grows up. It's a bad look, and a waste of precious time. Transitioning from a services firm to a venture building operation is hard - but not impossible. If you're thinking about such a move, and need resources, we cover this in our latest paper, "the GP Dilemma." DM me if you'd like a copy!

27

8 Comments

Ride the fractious horse 🐎 Had a blast talking with Harrison Chase from LangChain the other day to a great group of founders, builders and investors about building agentic systems. One learning I shared: our team builds effective agentic systems by spending way more time doing than speculating. 🏇 In other words, we "ride the fractious horse": New models drop daily. New frameworks emerge constantly. Best practices get declared and discarded overnight. By the time you survey the landscape and theorized about how today's reasoning models can serve your system, both the landscape and capabilities have shifted. So we write playbooks instead of adopting them. The only way to develop those playbooks? Build and experiment at the fastest possible rate. ✈️ Wilbur Wright laid a great analogy back in 1901: He'd just realized that existing "standard" measurements of lift and drag were wrong, and that solving flight meant spending serious time actually trying to fly (previous "pilots" had logged minutes, at most, of cumulative airtime across all experiments). From his Chicago speech (this guy was such a clear writer btw): "There are two ways of learning how to ride a fractious horse: one is to get on him and learn by actual practice how each motion and trick may be best met; the other is to sit on a fence and watch the beast awhile, and then retire to the house and at leisure figure out the best way of overcoming his jumps and kicks." "The latter system is the safest; but the former, on the whole, turns out the larger proportion of good riders." As for planes: "It is very much the same in learning to ride a flying machine; if you are looking for perfect safety you will do well to sit on a fence and watch the birds; but if you really wish to learn you must mount a machine and become acquainted with its tricks by actual trial." We're not inventing flight, but constantly evolving LLMs and probabilistic workflows won't behave as you hope on first contact. If you want to figure out how to build effective products around those fractious beasts, get on the horse (keyboard) and start riding.

72

5 Comments

Most launches flop, not from bad products, but from ‘ta-da!’ rollouts with zero warm-up. Headed to a fall conference to break some big news? My latest MediaPost column provides a 6-week blueprint to turn the booth into a momentum machine. Details and article link in comments ↓ #gtm #marketing

9

1 Comment

Back in 2015, I was a software engineer building the invisible infrastructure that helped publishers monetize with ads. Not the polished front end, but the backend stack: ad serving, bidders, exchange auction logic, data enrichment, attribution, verification. Every time you loaded a page, we ran an auction in milliseconds to match an impression with an ad, ideally relevant, but always revenue generating. That system was complex, but it monetized the open web and kept most of the internet free for decades. That model is now changing. AI labs are cutting direct deals with publishers like Axel Springer and platforms like Reddit. Users spend less time clicking through Google results and more time asking ChatGPT, Claude, or Perplexity. Google was about sifting through citations. Chatbots are about generating synthesized answers for almost any question you ask. This shift puts pressure on the open web. Volunteer communities are thinning out and moving to walled ones (e.g. Discord and Slack). High-quality content is moving behind paywalls. The rest is getting noisier as AI-generated filler proliferates. The traditional journey of search → click → impressions → ads is changing. Large language models are becoming the new distribution layer for content. Publishers once optimized for Google’s crawlers; now they need to understand how they show up in model responses. Search intent was powerful; chat intent is even more precise. Search revealed what you wanted, chat reveals why you want it. Full post in the comments.

14

2 Comments

Build AI like people matter, or you will build precarity at scale. Chávez’s reminder cuts through metrics and margins: the point of labor rights is people. If algorithms set the pace, assign routes, and rate work, justice depends on how we train, audit, and contest them. People-first means workers define success before code ships, and safety anchors optimization. Bias and burden audits track overtime spikes, heat exposure, and missed breaks, and automatically throttle when thresholds are tripped. Guarantee explanation and appeal that are fast, low-friction, and safe from retaliation, then share gains through bonuses and schedule flexibility. At Global Orbital Imaging AI (GOiAI), field MLOps treated development as labor: forward nodes ran INT8 shadow models on ARM NPUs to A/B score proposed routing and shift policies; workers submitted tap and voice critiques that auto-tagged errors; the edge packaged 160 KB model-card diffs, threshold proposals, and error tiles over a 0.8 Mbps SATCOM link to ATAK/CoT, updating the battalion COP in about 25 seconds so stewards and S2/S3 could approve or roll back with signed rationale. Labor Day should feel like reliability: predictable hours, safe shifts, and decision systems that respect human judgment. Reliability is engineered by contracts, councils, audits, and the courage to pause a rollout when harm shows up. Build like you believe it. Two-step challenge: name one distribution gap you will address this week, then post your first step in the comments for accountability. Optionally tag a teammate and add a 7-day check-in. #GhostSignal #GOAIST8 #Innovation #AI #EdgeAI #Leadership #ProductManagement

2

It's such a privilege to talk to builders like 2x exited founder and Hulken CRO Lee A. Rotenberg. There were so many threads to pull on in this brand story but this one about speed really resonated. Get my full conversation with Lee and Hulken Co-Founder, Alex Schinasi, below.

46

5 Comments

Performance agencies sometimes think they need to differentiate by building proprietary tech. I’ve seen that story many times — and I disagree. I recently had a check-in with two C-level execs at a performance agency. We’d been delivering exceptional results for their clients, including a $33 CPA using our machine learning optimization engine, compared to $120 manually. But performance wasn’t their focus. They'd decided to build new tech in-house. “If other agencies can use tvScientific,” they said, “we’re not unique.” It brought to mind the New York Times’ decision to build its own ad server — a costly, years-long detour that ended with a move back to DFP. The problem was a lack of clarity about what business they were really in and what they were positioned to excel at. The same applies to performance agencies. They’re not in the business of building ad tech (which is a constantly evolving, capital intensive business and is an entire business in its own right). They’re in the business of delivering outcomes. That means using the best tools available on the market and winning on superior client service. Not rebuilding tools from scratch. The edge comes from execution — knowing how to evaluate, adopt, and apply best-in-class technology to drive better results. Trying to be both a tech company and a service business is a recipe for subpar results on both fronts. Instead, invest in identifying cutting-edge technology, use it better than your competitors, and deliver both superior results and superior client service. That’s the path to agency differentiation. Just my .02. More detailed thoughts on this topic in my weekly newsletter (see comments).

193

17 Comments

🧀 𝐎𝐧𝐥𝐲 𝐨𝐧𝐞 𝐦𝐚𝐢𝐧 𝐟𝐫𝐚𝐮𝐝 service?  That's like locking your front door and leaving every window wide open. 𝐌𝐞𝐞𝐭 𝐒𝐥𝐢𝐜𝐤 — 𝐚 𝐟𝐫𝐚𝐮𝐝𝐬𝐭𝐞𝐫 𝐰𝐡𝐨 𝐬𝐥𝐢𝐩𝐬 𝐭𝐡𝐫𝐨𝐮𝐠𝐡 𝐬𝐢𝐧𝐠𝐥𝐞-𝐥𝐚𝐲𝐞𝐫 𝐝𝐞𝐟𝐞𝐧𝐬𝐞𝐬 𝐥𝐢𝐤𝐞 𝐛𝐮𝐭𝐭𝐞𝐫 𝐭𝐡𝐫𝐨𝐮𝐠𝐡 𝐒𝐰𝐢𝐬𝐬 𝐜𝐡𝐞𝐞𝐬𝐞 𝐡𝐨𝐥𝐞𝐬. In Part 1 of our 3-part series, Clutch breaks down why relying on one single fraud check / vendor is a recipe for disaster. 👇 🔹 Synthetic identities? He's got those. 🔹 Device fingerprinting? He rotates like burner phones. 🔹 Knowledge-based auth? He bought your mom's maiden name for $3. No single layer is perfect. But what happens when you start stacking them? 🤔 Drop a comment if this hits home — and stay tuned for Parts 2 & 3 this week. #FraudPrevention #Fintech #SwissCheeseModel #RiskManagement #Clutch

18

1 Comment