Operators
Humble AI talk
This week we met with the President of an insurance company and the COO of an asset manager. One had just come back from a conference that helped executives compare notes on AI, and the other was trying to figure it out internally.
“We probably bored the hell out of our wives while we were at dinner just talking about [AI] stuff, because we were coming out of these really exciting sessions.”
- President, Insurance Company
“We have all of these Claude-enabled workflows that individuals have integrated into their specific tasks and workflows. But we don't have an architecture to the whole system... And I don't really know what that should look like or how to do that.”
- COO, Asset Management Company
Both of them have spent decades building and running successful businesses. Now the goalposts are shifting and they’re trying to figure out the new rules of the game. What does “good” AI look like? How much should we spend on tokens? Claude or ChatGPT or Copilot or all three? How many employees should focus on this?
The executive conference worked because people actually shared what was going wrong and their concerns. The key takeaway was that most of them were still drafting bullet-proof AI blueprints and really needed to be stacking short wins and losses to go for volume. The losses are the most interesting part here.
Pretty much everyone we meet tells us they’re falling behind their peers. After speaking with 100+ operators, we know that can’t be true. What is more likely is that people aren’t talking and everyone’s acting like ducks: appearing relaxed but vigorously treading water to figure out AI. Wins get passed around at conferences or LinkedIn, and the struggles seem to just vanish.
There are a few key points that seem to work well for productive AI conversations:
Focus on implementation: talking about the LLMs or tokens is just preamble. The meat of any AI story is figuring out how they deploy it.
Go for industry breadth: it’s a disservice to only talk to folks in the same industry. Often the most helpful ideas we pass along come from vastly different company types.
Celebrate the wins quickly: it’s natural for people to lead with their best stuff. Pivoting the conversation to the struggles is far more insightful.
People keep asking us to benchmark their AI against peers like it’s some sort of scoreboard. We’re happy to continue doing so, but here’s how you can do it yourself:
Pick five peers in five different industries (ideally people you don’t need to impress). Ask each of them what AI project isn't working and what they are trying to figure out. Gently steer them back when they drift to the AI wins (they will). See if you can help them with ideas and then flip the script and get their advice.
The key insight for this blog post isn’t life changing: go talk to more people. But it seems like the harder thing to do is have honest conversations about how hard it is to figure this whole AI thing out. The more discussions that start with that truth, the better.