I brought my 8-year-old son to soccer practice.

A field full of kids. One ball. And every single kid chasing it.

Same direction. Same spot. Same idea. Nobody passing, nobody in position, nobody looking at the open space. Loud, sweaty, adorable.. and pure chaos. 🐬

Now imagine the coach blows the whistle and explains a few basics. The playbook. The roles. A simple formation: you defend, you hold the middle, you stay wide and wait for the pass.

Very quickly it's a different game. The ball moves. Kids pass. Somebody is actually open. Ordered, effective, elegant.. and a lot more fun for everyone on the field. 🐯

I've watched the exact same thing happen inside companies.

Over our last 45 AI implementations with larger companies, the first months looked a lot like that soccer field. We started implementing, and I saw more chaos, not less. People started more things. They did not finish more things. Everybody was running to the ball. 🌴

Then we switched our approach. Module one became project management. Only after that did we put AI on top. And it became so much better.

In this article I'll unpack why AI produces more chaos in the first month, and what the whistle, the playbook and the formation look like for your company.

Here's the bitter part nobody puts in the demo. AI doesn't fix how your company runs. It copies it. Then it runs that copy faster, all day, without getting tired.

Clear workflows? AI multiplies clarity. Messy workflows? AI multiplies the mess.

Same tool. Opposite result.

You wanted your evenings back. You wanted to stop being the only one who knows where every deal really stands. Instead you end up managing the tools too.

So before you hand AI the keys, ask one question: do you trust your own workflows?

Trust is the first thing that breaks

Think about the last time something slipped.

A follow-up nobody sent. A proposal that sat in drafts. A promise from a call that lived only in someone's head.

That's rarely a people problem. It's a trust problem. Nobody fully trusts that what goes into the system comes out the other side.

So it stays in your head. You double-check. You become the backup system for everything, and you carry that weight into every weekend. The only closer, the only planner, the only one who knows what's going on. 🐯

Now add AI to that.

An agent can't read minds. It only sees what's written down. If half your real commitments live in your head, your inbox and a few chat threads, the agent works on the other half. Confidently. At speed.

That's how chaos scales.

Why the first month gets louder

Across those 45 implementations, the pattern kept repeating. One large institution in the UK made it impossible to miss.

After one month it was obvious. They were starting too many things and finishing too few.

More starts. Fewer finishes. Every team sprinting to the ball, now with faster legs. 🐯

AI made each person quicker at starting. Nobody had a position, so nobody knew which ball was theirs to finish.

So we changed the program. Module one stopped being about AI. It became about being organized: project management, built on three GTD principles. The whistle, the playbook and the formation.

1. Get current. Every open project and commitment on one list, up to date. (Where is every ball right now?)
2. Get clear. For each one: what does done look like, and what's the next action? (Who plays which position?)
3. Get creative. Only then: where could we do this better, faster, differently? (Now run the plays.)

After that, everybody started putting AI on their own processes.

And here's what happened. Every person who was well organized suddenly had six pairs of hands. They knew exactly what they wanted done, so they could hand it off. They got much more productive. 🌴

The kids stopped chasing the ball. They started passing it.

But only in that order.

The same lesson, inside our own company

I run Strategy Sprints with a team of people and a team of AI agents. They prospect, follow up, process my inbox, prepare calls and write drafts.

We learned the same lesson the hard way.

One day I opened my task inbox and found 28 to-dos that agents had created for me. I read them and said: "They obviously want me to do something, but what do they want me to do?"

Titles like "Review reply draft." Nothing I could act on in ten seconds.

When we checked them against reality, 16 of the 28 were already done or stale. I had already replied. The draft was gone. The deadline had passed.

The agents weren't broken. They were faithfully scaling a sloppy definition of "a task."

Another time an agent iterated on one community announcement by actually sending each version. Our community got the same message eight times in one afternoon. 🐯 Nobody meant harm. There was simply no rule that said: a list gets one finished message, not eight drafts.

Both problems had the same root, and it wasn't the AI. It was the missing basics underneath it. No playbook, no positions.

The basics have a name: GTD

GTD, Getting Things Done, the method David Allen built more than twenty years ago. An old method, quietly perfect for a brand new tool. 🐯

Think of it as an operating system for commitments. Five moves, and each one is something an agent needs from you:

1. Capture. Everything that has your attention goes into one trusted inbox. An agent can't see what's in your head.
2. Clarify. For each item: what is it, and what's the very next physical action? That's the instruction an agent can run.
3. Organize. Put it where it belongs: projects, next actions, waiting-for, calendar, or trash. That's where an agent looks.
4. Reflect. Review it weekly, so the system stays current. Stale lists make agents chase ghosts.
5. Engage. Do the work from the list, with a clear head. The judgment calls stay yours.

The magic word in GTD is trusted. David Allen's point: your mind only lets go of something when it trusts the system will hold it.

That same trust is exactly what an AI agent needs. An agent can only work from a system. If the system is complete and current, the agent is useful. If it isn't, the agent is just fast. 🐬

The second basic: Natural Planning

GTD handles the daily stream of inputs. David Allen's Natural Planning Model handles projects, before anyone (human or agent) touches them.

Five phases. I split the first one in two, because for AI it matters most:

1. Purpose. Why are we doing this? What happens if we don't?
2. Principles. What standards govern how we do it? What will we never do?
3. Vision. What does wildly successful look like, on a specific date?
4. Brainstorm. Every idea, unfiltered.
5. Organize. Keep, cut, sequence, give each piece an owner.
6. Next actions. The first physical move, with a name and a date on it.

Now map that onto AI.

Purpose and Principles are your guardrails. "A list never gets more than one message from us in three days." "Nothing goes out that wasn't approved." Write them down and an agent can follow them. Keep them in your head and an agent will cross them on day one.

Vision is the brief. Organize is the workflow. Next actions are the instructions the agent actually runs.

A lot of AI rollouts jump straight to step four, the brainstorm. Lots of output. No direction. Every kid on the ball again.

Right things. Wrong order. 🐯

How we run GTD at Strategy Sprints, with AI on top

Here's our setup today. Nothing fancy. GTD, with agents doing the clerical work.

Inbox zero, every morning, on the metro. My rowing session is on the Danube, an hour away by metro. That hour is when my inbox goes to zero. An agent does the clerical part: every email leaves with an owner (a reply draft, a next action for me, a task for the team, or the archive). I just make the calls. Thanks to GTD, that hour on the metro is a blessing, not a burden. 🌴

A star means "I have to do something." When I have no time to write a proper task, I hit the star. The agent turns each starred email into one clear to-do. Capture takes me one second. Clarify still happens. 🌴

Every to-do passes the ten-second test. After the 28-task mess, we wrote a standard. Verb first. The person's name. What exactly to do, with a direct link. Why it matters. How long it takes. If a message needs sending, the finished text is already in the note.

So "Review reply draft" turns into something like "Reply to Anna: yes to the call next week," with the reply already written underneath. (Anna is a made-up name. The format is real.)

Agents check reality before they create work. Before anything lands on my list, the agent checks the live source. Did I already reply? Is the draft still there? Is the date still ahead of us? If the work is already done, no task gets created.

Monday Strategy, Friday Learnings. Monday sets the week: three priorities, nothing else. Friday scores it: what worked, what to double down on, what to drop. That's the GTD weekly review, run as a business rhythm. The agents report against the same three priorities, so drift shows up within a day, not a quarter.

A weekly review I actually do. Honest moment: I used to skip mine. It felt like too much. Now it runs as an interview, one question at a time. I answer, it listens, and Monday starts fresh. The system adapts to the human, not the other way round. 🐬

Rules live in code, not in hope. After the eight-message afternoon, "don't do that again" wasn't enough. Now a code check blocks a fourth message to the same list within three days. Principles you write down can be enforced. Principles you only feel cannot.

None of this is clever AI. It's GTD. The AI just does the boring parts, every day, without forgetting.

And it's not only the company. I have three kids. My wife and I organize many, many things together (soccer practice included). Without GTD, I couldn't do that efficiently today. Same system, at work and at home. 🐬

The order: GTD first, AI second

About to add AI to your company? Here's the sequence I'd follow.

Step 1: Get current. Capture every open loop into one inbox. Calls, promises, ideas, the deal you keep meaning to follow up. Then one project list, up to date.

Step 2: Get clear. For every project: what does done look like? Then decide what a good next action looks like. Verb first. One person. One physical step. A link. If a human can't act on it in ten seconds, an agent can't either.

Step 3: Run one real weekly review. Projects, next actions, waiting-for. Current, complete, trusted. That's your baseline.

Step 4: Get creative, with Natural Planning. Pick the process that hurts most (often follow-ups or prospecting). Write its purpose, its principles and what "done" looks like.

Step 5: Hand the clerical parts to AI. Processing, drafting, checking, reminding. Judgment calls stay with people.

Step 6: Turn your principles into checks. Every rule that matters becomes something that blocks a bad action automatically.

AI shows up in step five. Not step one.

It feels slower at first. Then it's how you go fast without crashing. 🐯

Why this is a sales problem too

At Strategy Sprints we teach one thing above all: be in control of the sales conversation. No chasing. No begging. The right things in the right order.

You can't control a conversation you've lost track of.

A deal that goes quiet often didn't die on the call. It died afterwards, in the gap between "I'll send you something" and a follow-up that never became a next action.

GTD closes that gap. AI on top of GTD keeps it closed, even in the weeks you're busy closing.

And your new sales hire? They work from the same trusted system you do. They don't need you in their head. You stop being the only closer, and you become the founder who knows exactly where every deal stands. 🌴

That's when sales gets fun again. Your evenings come back. And dinner tastes better when your head isn't still at the office.

That's Sprintlife. Not more tools. More control, and the time freedom that comes with it.

Your move this week

Before you add another player to the field, blow the whistle. 🐯

Empty your head onto one list. Then rewrite your top five items as real next actions. Verb first. One step. Ten seconds to start.

If that feels hard, an agent won't make it easier. If it feels easy, you're ready for AI. 🐬

Want our tools and AI skills for exactly this, plus founders building it with you right now? Join the Sprint Club: https://www.strategysprints.com

Ready to accelerate now? Book a Discovery Call: https://calendly.com/strategysprint/discovery-call

Happy hunting.

Simon & The Sprinters 🐬⚡️🐆