Craig Horton Advisory
The Transformation Brief
Craig Horton
The Transformation Brief

Good morning. This week's issue we cover where the AI dividend falls, JPMorgan reducing some teams by 30 to 40% while telling investors not to expect fatter margins, Anthropic putting $1.5 billion behind an implementation firm, and a radar where OpenAI has put a physical keypad on the desk for steering AI agents.

Top signals
1

The frontier model companies are moving into the transformation business

Anthropic, Blackstone and Hellman & Friedman launched Ode with Anthropic this week, a $1.5 billion enterprise AI services firm built on the Fractional AI team, two weeks after Microsoft committed $2.5 billion to its Frontier Company. The frontier model companies now want to run the deployment, and the deployment is where your operating model gets decided.

Business Wire, 15th July 2026
2

Supervising AI agents now comes with hardware

OpenAI released its first hardware product this week, the Codex Micro, a $230 desktop keypad built with Work Louder for monitoring and steering AI coding agents, with a push-to-talk key and a dial that adjusts how hard the model thinks. When supervising agents gets its own control surface on the desk, the supervision role in your operating model has become a real job.

TechCrunch, 15th July 2026
3

Cost discipline has reached the prompt

JPMorgan is telling staff not to point expensive frontier models at easy tasks, with CFO Jeremy Barnum urging teams to use the right model for the right purpose. A $20 billion technology budget has started managing AI cost at the level of the individual task, which tells you AI spend is now a normal operating line.

Bloomberg, 14th July 2026
Comment
Craig Horton: Know where the dividend goes before you promise it

Every AI business case I read ends the same way, a savings line that lands neatly in next year's P&L. I see executive teams approve these numbers without asking the one question that matters: who actually keeps the saving?

This week the biggest bank in the world answered it, and the answer was mostly not us.

Where does the AI dividend go?

What's the latest?

On Tuesday's Q2 earnings call, Jamie Dimon put numbers on what AI is doing inside JPMorgan: close to a thousand use cases in production, and "discrete areas where we did reduce jobs by 30% or 40%", with most of those people redeployed into other roles.

Then he told investors to be careful what they bank: "You don't uniquely benefit from AI. In a competitive, capitalist world, we all will use AI to do a better job for the customers."

A question came to mind: if the work is changing that much, and the margin is barely moving, where is all that value going?

Why does it matter?

It matters because the answer decides how you should write the business case. The dividend goes to three places. Your customers take the largest share, because every competitor has the same tools and the gain gets competed into price and service. The operating model takes the next share, the redesigned roles, the redeployed people, the new work. The margin gets what is left, and it is the smallest of the three.

So when the next initiative arrives with a savings line, ask:

The AI dividend readout, JPMorgan Q2 2026: the gain lands with the customer first, the workflow second, the margin last
The AI dividend readout from JPMorgan's Q2. The saving is real, the question is where it lands.
My reframe question
Before you approve an AI business case on a savings line, can you show how much of that saving reaches the P&L after competition takes its share and the redeployment is paid for, and who signed for that number?

Once again, I appreciate your interest and time to read this brief, and as always, look forward to your feedback and comments. Enjoy your weekend, Craig

The Transformation Brief  ·  brief.craighortonadvisory.com  ·  Archive