I sent my team this memo. You should read it.

****

Quick one this week.

Last week, I sent a memo to my entire team. I'm sharing it with you exactly as I wrote it.

The question at the end is one every leader should be asking about their business. Regularly.

Here it is.

Internal Memo:

“As I was working through my recent keynote, it struck me that we're in the fourth moment of AI

1. The Toy Era: We experimented. The potential was obvious, but the limitations demanded constant supervision.

2. The Reasoning Era: AI became a thinking partner. With good context and clear questions, it could help us analyze, challenge ideas and produce better work.

3. The Harness Era: We began organizing AI around outcomes. Give it the right information, tools, boundaries and ways to check its work, and it can carry substantial work through to completion.

4. The Personal Jarvis Era: We are beginning to give AI ongoing responsibilities. It retains context, monitors developments, advances work and brings us the decisions that require our judgment.

My concern is that much of the adoption I see is still in the second era. We ask questions, improve prompts, generate drafts, and then personally coordinate everything that happens next.

We can upgrade our AI without upgrading the way we work.

For example, there is a meaningful difference between:

• “Help me prepare for this client meeting.”

• “Review the client files, assess progress against commitments and prepare a verified briefing with the decisions we need to make.”

• “Keep this engagement moving. Track commitments, identify emerging issues and prepare the next actions for review.”

Each requires a different level of context, access and oversight. But each also changes how much work we must personally initiate and coordinate.

This applies to us twice: in how we run our own business and in what we help clients accomplish. If we teach better prompting while leaving the underlying work unchanged, we are stopping short of the opportunity.

Before our next AI Community of Practice meeting this Friday, I’d like each of us to identify one recurring workflow we could redesign.

• The outcome we need and how the work happens today.

• The steps AI could carry forward.

• The information, tools and boundaries it would need.

• Where human judgment and approval should remain.

• How we would measure whether the new approach is better.

Choose something meaningful enough to test a change in turnaround time, quality, capacity or client value. We can start with a bounded trial and expand responsibility as performance earns our confidence.

We ask our clients to reimagine how they work. What are we missing?”

The upgrade is the workflow. The model is just the engine.

Most of us are using a better tool to do the same work. That feels like progress. It isn't. Not yet.

The upgrade is the workflow. The model is just the engine.

So here's the same homework I gave my team. Pick one recurring workflow.

Map how it happens today. Decide what AI could carry, what it needs, and where your judgment stays. Run a bounded trial. Measure it.

Then expand only as it earns your confidence.

The technology is the easy part. The change is the work.

What's one workflow you're still personally coordinating?

Matt

• • • • • • • • • • •

P.S. If you're not doing a regular community of practice meeting with your team, you should be. We do ours every Friday morning.

P.P.S. State of AI is Tuesday, October 27, at 12 PM AT. Bring your toughest workflow and I’ll unpack it live. First session is a free trial. Please use it.

​[Save your seat]