"What should I automate first?" is the wrong question. Here's a better one.

Welcome to another edition of Leading Momentum, where we cover on-the-ground insights leaders need to know about getting AI operations to work in their organization.

Today I’m going to share an easier way to think about what to automate, by applying one of our protocols we’ve published: Progressive Delegation.

I keep having the same conversation with leaders.

Their team asks: "what should we automate first?" And the leader doesn't have a great answer. Because honestly, that question sounds reasonable. But it's the wrong starting point, and it's the reason most teams get stuck.

They try to jump straight to automation, hit a wall of complexity, and then either come back asking for more direction or quietly stop trying. And the leader is stuck in the middle of it. You told everyone to use AI. Maybe you bought the tools. But now people keep looking to you and you don't have time to hand-hold every person through it.

I get it. I really do. Because we went through this ourselves.

We have this process in the Bootcamp where we give personalized feedback to every learner on the playbook they're developing. It draws from everything we know about them plus a huge library of playbooks we've built over the years. It's detailed, specific work.

At first, we wrote every piece of feedback by hand. And that was fine for a while. But we started noticing patterns in what we were saying, how we were coaching, what we kept correcting. So we thought about it like... if we had to teach a really sharp intern to do this, what would we tell them?

That became the first version of the playbook. And it was rough. Our team had to review every single output, override things, fix things, keep tuning the instructions. We were in the weeds.

But here's the thing. Every correction we made went back into the playbook. Not into the AI's memory. Into the actual instructions. So each round got a little better.

Over time, we started removing review steps. One by one. Until we were only looking at the final output. Now when someone submits a playbook, AI produces the feedback and our team reviews it before sending. Each week the team logs every issue they fixed and we triage those together.

The next step might be making it fully automatic. We might get there. We might not. Either way, it's already saved us a massive amount of expert time.

I've started calling this progressive delegation. And it's become the framework I give every team that asks me "what should we automate?" We just published it as a full protocol on our site. It goes deeper into each stage and what earning the next one actually looks like. Review the Progressive Delegation protocol

But here's the short version. You don't go from zero to fully automated. You earn each stage.

Think about self-driving cars. Nobody went from "human drives everything" to "car handles it all" in one step. First the car assists. Then it handles more while the human stays ready. Then maybe, eventually, full autonomy. Each level was earned by proving the last one worked.

Your team's AI work follows the same stages. And the answer to "what should we automate?" is: don't. Not yet. Start at stage one: be the expert in the loop.

That means your people write down the process, build a simple first version with AI doing the work, and review everything. They're not checking boxes. They're accountable to the quality. And every correction improves the playbook for next time.

The part that trips people up is they correct the AI in conversation and expect it to remember. That's not a feedback loop. That's just having the same conversation over and over. The corrections have to go into the instructions.

Once the playbook gets good enough that your team is mostly just approving outputs, you've earned stage two. And from there, you keep removing yourself until the workflow runs on its own, or until you find the level where the quality stays high and you stop. Both are wins.

The reason this matters for you as a leader: it replaces "what should we automate?" with a better question.

Instead: "Where can we get AI to help and stay as an expert in the loop, for now. Then automate more later?"

That's a question your team can answer without you. They just start at stage one with whatever process they already own, and earn the next stage.

That's it. That's the framework.

What's the one process on your team that you wish could just run itself? I'd love to hear what's on your mind.

Read more about the protocol here: https://runonamp.com/protocols/progressive-delegation

Thanks for reading. It's a privilege that you spend time with these each week.

Whenever you're ready, here's how we can help:

Cheers,
Rachel