#37: stop planning your AI strategy

I've been building an AI strategy playbook for my consultancy clients.

Figured I'd crack it open and share the core idea with you before it gets all polished and professional.

Because here's what I keep seeing: businesses treating AI like it's either a six-month planning exercise or random ChatGPT chaos.

Both are losing.

Every discovery call follows the same script.

Half the companies spent months building the perfect AI strategy. Beautiful decks. Workshop after workshop. Zero execution.

The other half? People using ChatGPT for emails. Nobody sharing what works. The company learning nothing as a unit.

Meanwhile their competitors ran three experiments in the same time period.

Here's the thing about AI: you can't plan your way there.

It's non-deterministic. You don't know what works until you try it. The timing will never be perfect.

But experimentation without direction is just chaos.

This is the unlock most people miss.

You'd never say "we're doing marketing in October, then we're done with marketing."

Marketing is a business function that's always running. You have projects within it (launch a website, run a campaign) but marketing itself? Always on.

AI works exactly the same way.

AI isn't a project with a start and end date. It's part of your business foundation. You build initiatives within it.

An initiative might be "automate quote follow-ups." That has a start, end, defined outcome.

But AI as a concept? That's just how you work now.

The cycle looks like this:

Identify where AI creates value → Implement and test → Scale what works → Start over

Again and again.

Most businesses aren't ready for AI.

Not because they lack vision or budget. Because their operations are held together with digital duct tape.

You can't automate chaos. An AI agent in a mess just creates more mess.

Quick diagnostic:

Customer support running from individual email accounts? Critical context only shared in meetings? Using memory as your CRM? Contracts scattered across random folders?

You need two things in place first:

Data lives somewhere predictable. Core processes are written down.

Not glamorous. But without this, AI just amplifies the chaos.

I tell clients: if you can't check 6 out of 8 boxes on basic data structure and process documentation, clean that up first. Takes 2-4 weeks. Then come back to AI.

Here's the framework I'm using with clients now.

Forget the 12-month master plan. Nobody follows those.

Instead: 90-day cycles.

Month 1: Implement one quick win
Month 2: Start the next one, build competence
Month 3: Measure, adjust, plan the next 90 days

Long enough to get results. Short enough to pivot fast.

The key is picking 2-3 quick wins to start. High value, low complexity.

Examples that work:

Start small. Build on success.

Two failure modes keep showing up.

Top-down only: Leader decides "we're doing AI" without the team coming along. Dies from lack of ownership.

Bottom-up only: Few enthusiasts trying things without leadership backing. Dies from lack of resources and prioritization.

You need both directions at once.

Leadership provides legitimacy, resources, decision-making when things hit walls.

The team provides concrete ideas, ownership, solutions that actually work in practice.

Weekly 15-minute shares. One person shows what they tested. What worked, what they learned. No fancy presentations.

Plus one place (Google Doc, Notion page, whatever) where the team collects prompts that work, useful tools, context docs.

That's it. That's the learning culture.

Start simple. Build complexity over time.

Level 1: Give people ChatGPT, Cursor, whatever tools fit their job. Train them. This is single-player AI.

Level 2: Automate one person's recurring process. Finance converting invoices monthly? Sales chasing quotes? That's still single-player but bigger impact.

Level 3: Rebuild an entire workflow within one function. Lead to sale. Content production. Multiplayer, single function.

Level 4: Automation across departments. Performance marketing and creative shipping new assets continuously. Together. Automated.

Most start at level 1. That's smart. Build experience there, then move up.

80% of what you test will fail.

That's not a bug. That's how learning works.

AI doesn't help if your problem is unclear processes (clean up first), messy data (structure first), or nobody knows how things should be done (document first).

The businesses winning the AI game aren't the ones with the best strategy.

They're the ones who started experimenting systematically while everyone else was still planning.

Until next week,

Martin

PS: feel free to connect with me on LinkedIn and say hi! It’s always fun to chat with new readers.

PS2: want to accelerate AI adoption in your business? Book a free 30-min discovery call here.