It's no secret that companies in this type of spot struggle. They tend to cut R&D, raise prices, and eventually, if things can't be overcome, they begin to lose their strongest engineers and salespeople.

At a time in software where I'd argue that adapting matters as much as a moat, or more, these companies will have an additional challenge as their ability to adapt becomes limited.

So, I'd agree with Lemkin's advice: take some time to list out and consider the customers and top talent of your competitors bought by PE firms around 2021.

There may be opportunities there in the near future.

Gemini's tags look gone for good, so expect more AI visits to show up as Direct in GA4. This means continue the basics of asking prospects how they found you and, if it was via AI, what they were asking when you were recommended.

This isn't necessarily a new trend or surprising. I include this more as a point of something we're watching: what is the rate of consolidation, is it increasing, and if so, why?

PitchBook notes that LPs short on cash are re-upping with the firms that paid them back, which makes sense.

What I wonder is how are funds using the additional capital to build unfair advantages that increase the likelihood of their next fund's success?

Brand is an obvious advantage for them, but their build-out of platform teams is another area we're watching. It'll be interesting to see where this goes.

Bain is republishing some additional graphics from their July study.

What shocked me was the percentage of buyers that are not only using AI for initial research, but throughout the buying process. Seems increasingly important to understand how AI is perceiving you.

I suggest growth teams use any of the many open-source projects to track this. Ours is Overheard AI.

It's good to see open models putting pressure on the frontier labs.

The very important nuance here, though, that people forget, is that AI companies often show prices as cost per token. But different models, and even different model generations, can have significantly different amounts of tokens required per task.

Be sure to review your cost per task, and also to run your own tests, and consider what models you need for what aspects of which tasks to balance performance and price.