Sudheer Maremanda

Voice AI is evaluated on your worst instance. Text lets you iterate quietly. Voice doesn't. The call is live, emotional, and full of interruptions. I keep hearing this from teams building in the space. Heard it reinforced hard at Thursday night's Fin x Cartesia event. Peter Bar @ Intercom and Israel (Iz) Shalom @ Cartesia are building at the edge of this.

23

4 Comments

Meeting users where they are; very well done by OpenAI For Clinicians: (UI)ChartGPT - Inside Epic (UI)ChatGPT/Codex - With Epic Plugin For Consumers: ChatGPT - Connecting with myChart, Function Health etc. #AIinhealthcare #patientaccess

3

🚀 PIVOT Pathways: Future-Proofing Leadership & Innovation Rapid technological shifts and the integration of AI tools are fundamentally altering how tech teams build, innovate, and lead. Surviving and thriving in this new environment isnt about competing with technology—its about combining AI tools with human judgment, creating psychological safety for experimentation, and adapting leadership styles to modern team needs To help engineers, managers, and product leaders navigate this shift, Raja Saravanan and I have curated "PIVOT Pathways: Future-Proofing Leadership & Innovation" — a 3-part audio track focused on turning technological change into a strategic advantage. 🎧 What’s in this Playlist? **** Co-existing with AI https://lnkd.in/gi84v8QM From AI coding assistants to automated copilots, AI is rapidly reshaping software development workflows and democratising technical knowledge. Explore how to thrive alongside AI by blending automated tools with deep expertise, sound judgment, and ethical accountability. **** Building a Culture of Innovation https://lnkd.in/g8pmqGhq True innovation doesnt happen by putting buzzwords on a slide—it thrives when teams feel safe enough to challenge the status quo and experiment. Discover how to celebrate intelligent failure, use "What if?" questions to unlock creativity, and leverage AI as an accelerator for everyday experimentation. **** Leadership Styles https://lnkd.in/g49u5DQZ Leadership isnt one-size-fits-all, and modern software teams require adaptive management. Learn when to pivot across directive, coaching, supportive, and delegative leadership styles based on situational urgency, team maturity, and the rise of AI. 💡 Why Listen? Whether you are managing engineers or navigating an AI-driven workflow yourself, these episodes provide practical frameworks to lead with agility, encourage creative problem-solving, and build future-ready tech organizations. #PIVOT #FutureOfWork #AILeadership #EngineeringLeadership #TechInnovation #PIVOTPathways

11

Lately, I've been trying a new pattern: putting my TODOs directly in the UI as visible, in-context instructions. It's been surprisingly effective in this early phase, where rapid dogfooding and idea incubation matter more than polished flows. I've never been a fan of TODOs. They often become feel-good bandages that hide future land mines. "We'll get to that when…" often becomes "never." And whether they sit in code, backlogs, wikis, or READMEs, they don't hit the same way as seeing them in the product every day. Early software is full of semi-perfect flows. There's always a balance between refinement and time-to-market. In my case, I'm still moving from manual user creation to automation. I'm not ready to build the automated flow yet, but I can't afford to lose track of what needs to happen when a new user is added. So instead of burying TODOs in documentation that will quickly go stale, I'm putting the necessary steps directly in the UI as temporary onboarding instructions. It eliminates waste, keeps the work visible, and creates a natural placeholder for the future user guide and FAQ. This approach combines incremental development, edge-case handling, and user communication, and it forces clearer thinking about what users actually need. So far, it's been a surprisingly powerful pattern. Ymmv?

6

Been thinking a whole lot about how systems thinking is more relevant now than ever before in this wild and woolly age of AI. So much so that @joeltosi and I decided to do a Maven lightning talk on the topic. Join us? https://lnkd.in/gRVRy9bW

30

3 Comments

As someone who has built and worked on AI/ML products for the better part of the last decade, with a philosophy background, I want to share where I land on the ethics of AI. The loudest version of this conversation treats the technology itself as the moral object. That AI is dangerous and that building it is the choice being judged. I do not think that holds. But I want to give the other side its strongest form first. Tools are not perfectly neutral. What a technology makes easy, people will do more of. Design and implementation choices carry moral weight before anyone uses the tech. That is a real argument and I take it seriously. But it is an argument about design, not about existence. It says build carefully. It does not say do not build. And I think the case for building is strong. Yes, every meaningful expansion of what humans can understand and do has arrived with some misuse alongside it. But it has also brought enormous increases in quality of life. We did not stop at the printing press or the internet, and the world is better for it. I stand by the notion that powerful tools in the hands of people trying to do good will outproduce the harm. That belief is why I got into this line of work and why I am still in it. The question worth arguing about is not whether AI should exist. It is who decides what these systems are allowed to do, and whether the people building them are making those calls deliberately. I would rather be in the room for that conversation than outside it. #AIEthics #AIProductManagement #ProductLeadership

7

HIMSS is almost here, and I’m looking forward to catching up with so many of you. I'm especially looking forward to our session with UC San Diego Health. Karandeep Singh and Jeffrey Pan will showcase how they're using Notable's Platform and AI Agents to act as a critical workforce multiplier, transforming patient access, care operations, and RCM across the enterprise. Learn more about everything we have going on at HIMSS here: https://lnkd.in/gQifSd2A. And definitely come say hello if you're attending!

29

2 Comments

Intelligence is becoming a commodity. Gemini 3 Deep Think just scored 84.6% on ARC-AGI-2. Visual reasoning puzzles. Easy for humans, hard for AI. Average human scores 60%. Strong abstract thinkers, much higher. So the model wins. Case closed. Except it used 138,000 reasoning tokens per task. Tasks that strong abstract thinkers solve in seconds. Here's the thing. Same benchmark. Same task. Gemini Pro, direct mode: 96 tokens. Deep Think on the same task: 138,000. That's a 1,400x cost gap. The variable isn't the model. It's who's pointing it. The person who knows which 3 hypotheses to test instead of searching 10,000? Most expensive component in the stack. Differential thinking just got a 1,400x multiplier.

5

Arvind Jain sees context graphs emerging from enterprise search and observability. I agree on the destination and disagree on the starting point. Arvind - a fellow IITD alum who's devoted his career to enterprise search - has built a tremendous business at Glean, and his "capture the how, learn the why over time" framing is sharp. But the foundation needs to be different: 1 - Search sits in the read path, not the write path. By the time activity data reaches the index, the decision context - why an exception was granted, what inputs were weighed, who approved - has already been flattened. 2 - Inferring intent from observed patterns is fundamentally different from capturing decision context in the operational flow. One reconstructs; the other records. 3 - The ~80% accuracy Arvind cites on task understanding is impressive, but when agents affect customers, contracts, and compliance, you need an authoritative record, not a probabilistic inference. (The math of compounding errors is brutal - I’ve argued for 99%+ accuracy in past editions of my Substack.) -- We believe orchestration-layer startups - where the context graph is built by actually executing work - are better positioned: 1 - They sit at the point of decision. When an agent triages an escalation or approves a discount, it pulls from multiple systems and applies policies in real time. That's the moment to capture the decision trace, not reconstruct it later from activity signals. 2 - They create decision traces as first-class artifacts. The orchestration layer captures the full picture: what inputs were gathered, what policies applied, what exceptions were granted, and what state existed at the moment of decision. That's not inferred; it's captured - which means enterprises can answer "why did we do that?" definitively. 3- They can learn across customers. Arvind notes enterprise data can't be aggregated for privacy reasons. But orchestration-layer startups focus on bounded workflows - which means they can refine ontologies within a customer and across deployments without ever sharing raw data. -- PlayerZero, Maximor, and Oliv are great examples of this. PlayerZero builds the context graph by automating L2/L3 support - sitting at the intersection where code, config, infrastructure, and customer behavior collide. Maximor AI captures decision lineage by orchestrating finance workflows where reconciliation logic and exceptions actually live. Oliv AI builds it for sales - starting as a co-pilot that performs specific tasks (update CRM, send follow-up) and capturing the decision traces along the way. There will be multiple context graphs within each org, and enterprise search can certainly be a starting point for one of them. Glean's approach creates value for knowledge retrieval and understanding how work flows. The question is whether the authoritative record of decisions will emerge from observability or from orchestration. We're betting on orchestration.

221

48 Comments