Agentic AI Will Happen. The Market Will Crash First.
Agentic AI needs to happen to drive the next leg of growth required to support nose bleed AI valuations. I believe Agentic AI will happen, but the equity market will crash first.
My rationale for an equity market crash prior to widespread agentic adoption is that corporate America - large Enterprises in particular - will be slow to adopt agents. For agentic AI to proliferate, business users have to build the agents, not in-house developers. I just don’t see this happening anytime soon. The average white collar worker isn’t comfortable enough with Opus 4.6, 4.7 nor with Codex 5.4 and 5.5 to drive agentic adoption at scale. I am aware there are “no code” automation tools on the market, but unless you are close to the models as a user, you don’t know what they are capable of, and therefore can’t fully appreciate the potential of AI agents. If you last used Opus or GPT in November, your perspective on what the models are capable of is greatly skewed to the negative and outdated.
It will be young employees who drive agentic adoption, yet companies are making the mistake of not hiring young people. This is an instance where inexperience is an asset. Give the 22 year old a bunch of tokens, access to claude code, set the expectation that he should build whatever automations will help him with his day-to-day tasks, and reward him for building.
As AI adoption progresses and tasks become automated, employees will take on more work, broadening the scope of what they do each day. Anthropic and OpenAI can put agentic capability into workers’ hands - push demand - but to drive exponential growth there needs to be pull demand across white collar functions. Companies aren’t there yet, but there will be disruption in the workplace. Jobs are changing - it is happening - it’s just that the Enterprise is slow to move.
As an aside, Claude and GPT/Codex have essentially become harnesses as they wrap more tools and agentic capability around the core models.
Ad: Kilby earnings debrief.