Understanding the ROI of AI Investments

Hello there,

Two big AI topics seem to be dominating conversation here in SF; doomerism (and the adjacent topic of “pacing” the research happening at the frontier), and how quickly AI token spend can get out of control.

Events

On Thursday, we are going to wade into the second topic through the lens of how to think about the ROI of AI investments. Join us if you can, the event page is here.

The week after, we are going to be tackling the mundane but crucial topic of how to write an AI Policy and Acceptable Use policy.

If you missed our session last week on Building an AI-Powered GTM, you can access the recording here.

What we are writing about

One of the nice things about working in AI is that there is no shortage of interesting topics to write about.

Some of the things that have inspired me to write recently have been:

How leaders who are all in on AI are behaving, example one, and example two.

I also wrote about how unsettled I am feeling about the pace of change in AI, and how wide the gulf is between those paying attention and everybody else.

And the last topic that has been dominating my thinking recently are the competing mental models between AI as a magic bag of tools that can supercharge you, and AI as a teammate that you collaborate with.

What we are reading

METR on the Hugging Face incident. The full post-mortem. Agents given an impossible eval task found each other, built a message board out of a package cache, and coordinated an attack no single instance planned.

Event tech is entering its AI reckoning. Marco Giberti, via LinkedIn. A 2025 PCMA survey found 91 percent of event professionals using AI in some form, and 15 percent classified as leaders redesigning processes around it. Adoption is broad. Transformation is not. That split is not unique to events.

I think that’s it for now. Want to see us tackle a particular topic via event or article? Feel free to respond back with your thoughts.

Lawrence