


Cofounder and CEO @ScribeHow. Here to make AI agents work and help companies own their specialized intelligence. Alum @greylockvc @mckinsey @princeton


I noticed that folks with kids are better at using AI agents. Probably because they have lots of experience in instructing someone intelligent who looks for loopholes in order to claim that a task was accomplished.

you cannot automate what you do not understand and to understand, you cannot first automate slow down to go faster i guess

“Human-centric” and “AI-first” aren’t opposite ends of some spectrum of how companies value people. Working more efficiently, eliminating mundane work, and building products/services that solve people’s problems more reliably are human-centric ways to work.

I love how everyone is now targeting “systems thinking” as the most valuable skill in tech. Guys, that was always the answer

Giving everyone in your company LLM access won't make your company AI-native. Our latest Scribe data on non-developer AI usage shows that 18% of knowledge workers do 82% of the AI work. Companies give employees AI tools and hope everyone will figure out how to use them. Some will. But most won’t.

"The strategic question... is shifting from whether a company can build AI to what others cannot build without access to its unique data or capabilities."
Artificial intelligence is becoming easier for companies to access — but when competitors can also use many of the same foundation models and AI agents, simply adopting AI won’t necessarily set an organization apart. mitsloan.mit.edu/ideas-made-…

An AI assistant that learns to mimic an employee will learn their bad habits too. You can’t know which habits are worth copying until you look across everyone doing the same work and see what actually led to good outcomes.

When I talk to leaders about bringing AI to their company, they always ask about which model is best. Wrong question. It changes quickly and everyone’s going to converge on whichever one’s best anyway. It's not where you get an advantage. People should be asking what their company owns that nobody else will be able to rent.

An AI agent is like a smart new hire who’s only allowed to learn how to do their job by reading the company wiki. They'll understand the basics of how things work. But the way work actually happens lives in exceptions, workarounds, and judgment calls that almost never get written down. Models know a lot about how companies operate in general, but they don’t know how yours works. If you’re rolling out agents, that’s where I’d put effort first.

Daylight saving time is a policy that could only have been designed by people without young children. If your toddler also thought 4am was a perfectly reasonable wake-up time this fine Monday, I see you 👀

Scribe tested a frontier model on 30 enterprise tasks to see what context helped it most. With a company wiki: ~55% better than the model alone With step-by-step guides of how people do work: ~70% better With both: ~60% better. Worse than the guides alone. More context =/= better performance. It has to reflect how work actually gets done.

The stat going around today, that 10% of workers capture 90% of AI productivity gains, is a predictable reflection of how companies have been thinking about AI: as an individual productivity tool. Which means the gains accrue to individual employees and can leave with them. Treating it like a talent problem is not the corrective. Companies need to see that how work gets done is an asset (specialized intelligence) that they can own and benefit from beyond the sum of productivity gains. Their focus should be on capturing it.
Just spoke to the Head of AI at a huge company. The #1 thing they're struggling with is... what do I do with the AI laggards at our company? Like most companies, they’ve found that a Super Pareto Principle is true within their org. 10% of their employees are responsible for 90% of the uplift in productivity/token usage from software engineering to non-technical knowledge work. They’ve spent a lot of time & resources on AI enablement from LLM foundations to function-specific trainings and skill building. And yet, the power law of productivity increases hasn’t budged. So now they’re left with a few sobering thoughts and hard questions: 1) 90% of people are overpaid and 10% are underpaid 2) We’d be better off allocating tokens proportionally to the power law, but how do we do that without killing culture? 3) Significant layoffs are likely warranted and wouldn’t impact overall output, but cultural backlash is a big concern? 4) How do we hire more power law people as a non technology company? 5) Is all of our efforts on training and enablement ACTUALLY worth it? 6) How do we not experience a brain drain where the top 10% all leave because they want to be surrounded by other AI-native 10x’ers?

Scribe's Refresh is looking really good.

The more a company trains its AI on email and wikis, the more its agents learn the job from whoever writes the most. That’s a strange way to pick your process expert. The people who are great at their jobs rarely write down how they do it. Most vocal =/= most authoritative.

The best, most comprehensive take on AI I’ve read. After a year of confusion and fear about the future, I finished it feeling like a weight had physically lifted. It’s long. Make the time. @vgr is undefeated. contraptions.venkateshrao.co…