Focused on what AI means for legal, finance, accounting & IT. For the people closest to the work, not the loudest voices talking about it. Building modInsight.
San Francisco
Joined May 2012
- Tweets139
- Following183
- Followers24
- Likes49
10 live interviews this year and we're just getting started. The industry is unfolding and the lessons are being immediately absorbed - join us.
‘We were wrong on nearly everything a year ago’ is exactly the right way to think about the pace of things. Every morning is a new inflection we have to reconcile @trueventures
The InfoSec/GRC seat is a single point of failure most companies don't staff redundantly, and when it breaks, it shows up as a sales problem, not a compliance problem.
Budget season question nobody has a clean answer for yet.
Where does AI token spend actually belong on the P&L?
COGS. OpEx. R&D.
Most companies are guessing right now.
Zoom out further. Legal, accounting, consulting, healthcare, the professions where expertise is the product, are moving slowest on this. Not because the tools aren't there.
The ROI conversation simply hasn't caught up yet.
There's an existing answer for how this gets classified. I don't think it holds up anymore.
We used to just store this stuff.
Sat in a database, dormant, done. Now the thing sitting in that same budget line is doing the actual work, constantly. Different category of spend.
Simple in theory. Almost nobody splits it this way today.
The real problem isn't the definition, but the attribution of token spend. Nobody's engineering team can tell you at month end what percent was shipping versus experimenting.
It's all hitting one number.
Here's the test I keep coming back to.
If the AI is doing the actual work that delivers your product, that's a cost of delivering your product. Put it in COGS.
If your engineers are spending tokens tweaking a model or testing something new, that's R&D.
What's the diff between NYC and SF founders? Something we are seeing in law / accounting: SF founders are building deeper vertical tech plays while NYC founders are building native AI firms that do the work.
Intelligence is COGS for expert roles, not an IT expense. Presenting AI spend as a % of payroll is now outdated.
We'd never hire associates we had to re-brief on every assignment and couldn't tell if they were learning the work. Why do we accept this from AI?
One billion is a threshold number for machines / physical AI.
But being 'bigger than LLMs' must mean there is more value in aggregating the value of general intelligence into what we touch every day.
Applied Intuition co-founder Qasar Younis on physical AI:
"It seems weird to say right now but I think it will be bigger than LLMs."
"Yes, writing code is important, but when you sit in an airport in Detriot, or Munich, or Turin, or Seoul, or Tokyo, where a lot of these companies that make these big machines are, the top thing on the mind of everybody in that gate is not, 'I wish my agent would work better.'"
"They're truck drivers. They're working in the real world."
@qasar at the 2026 RAISE Summit