The AI maturity ladder.

← Back to all articles

The shopping receipt problem

Three Claude licences, a Notion AI seat and two custom GPTs don't make a mature stack. They make a shopping receipt.

This is the most common state I find when auditing marketing teams: real budget spent, real tools deployed, and a maturity self-assessment that has very little to do with how the team actually works. The licences are evidence of intent. They are not evidence of maturity. Buying an enterprise AI licence is not an AI strategy, and a stack of subscriptions is not an operating model.

AI maturity is measured in six things, none of which appear on an invoice:

• Workflows that run end-to-end through AI, not isolated prompts inside individual tasks.
• Context the AI actually knows about your company: brand, ICP, positioning, ways of working, encoded in a form a model can use.
• IP that's versioned instead of trapped in one person's prompt history.
• Measurement you can point to: one metric, tracked before and after.
• Bus-factor: what survives when your strongest AI user is on holiday.
• Cross-team reach: whether the maturity extends past your function's borders.

That last one is the trap, and we'll come back to it. First, the ladder.

The four levels

Level	Name	One-line test
Level 0	Prompt-curious	People use AI. The company does not yet.
Level 1	Tooled	The licence arrived. The operating model did not.
Level 2	Systematic	AI has a repeatable, owned, measured way of working.
Level 3	Compounding	The system gets better every time work moves through it.

Level 0: Prompt-curious

People use AI. The company does not yet. The work depends on individual talent, private prompts and personal accounts. Output quality varies wildly because the standard lives in people's heads, not in the system.

The markers: individual ChatGPT accounts, no central prompt library, AI treated as "a tool people use" rather than a way the team works. In my diagnostic work, somewhere between 40 and 60% of marketing teams are still here in 2026, whatever their pitch decks say.

The diagnostic: if your strongest AI user left tomorrow, would the work break? At Level 0, the honest answer is yes.

The signal to move up: codify what the best user already knows. Their prompts, their context, their judgment about what good output looks like. That knowledge is currently an employment risk; turn it into an asset.

Level 1: Tooled

The licence arrived. The operating model did not. The company has bought access: enterprise seats, stacked tools, perhaps a custom GPT or two. Nobody owns the standard of use. Tool access is treated as strategy.

Most companies stop here and think they are done. The procurement cycle ends, the line item appears in the budget, and the maturity conversation closes. This is the shopping-receipt stage.

The diagnostic: you can buy your way into Level 1. You cannot buy your way out. The exit requires a decision no vendor sells: appointing an owner for how AI gets used, not just who gets access.

The signal to move up: that ownership decision, made and named. One person accountable for the standard, the context, and the first measured workflow.

Level 2: Systematic

Now AI has a repeatable way of working. The difference from Level 1 is not prompt volume. It is shared context, workflow ownership, and one measured before-and-after result.

The markers: a shared GPT or project used by three or more people weekly. Brand and ICP encoded in shared context, not pasted fresh into every conversation. One workflow that runs end-to-end. One metric tracked before and after.

A caution from the field: a Notion page of prompts is not Level 2. It is Level 1 with documentation. The test is whether the system runs without its author, not whether the prompts are written down.

The starter system: one campaign brief in, one first draft out, one cycle-time number tracked. Brief-to-first-draft cycle time is the strongest starter metric I know, because it is visible to everyone and hard to argue with.

Level 3: Compounding

The system gets better every time work moves through it. This is where AI stops being a productivity hack and starts becoming operating IP: versioned, measured, and woven into the seams of the stack where HubSpot meets GA4 meets the CMS meets Slack.

The markers: versioned IP. AI working at the stack seams, not just inside documents. Cycle time tracked as a trend, not a one-off. New hires productive by day three because the context is in the system, not in someone's head. And the big one: AI that crosses from marketing into sales.

The reality check: marketing output scales without marketing headcount. That is the observable, board-level symptom of Level 3.

The save-this test: does the system remember, measure and transfer the work better next time? If yes, it compounds. If no, it is a faster version of last year.

The cross-team trap

Here is the failure mode that caps more stacks than any tooling gap: marketing runs on AI, and sales still receives leads as a CSV and re-keys them into Excel. The enrichment, the scoring, the context marketing built, all of it dies at the handoff.

If that is your situation, you are not at Level 3. You are at Level 2 with a marketing silo. The value of everything upstream is being written off at the team boundary, every single day, quietly.

AI compounds across teams or it doesn't compound at all. A system that gets better every cycle needs the whole cycle, and in B2B the cycle does not end where the marketing org chart does. This seam, the marketing-to-sales handoff, is where I now start almost every maturity diagnostic, because it is the single most expensive place for value to leak. It is also the subject of Off The Deck Issue 2: sales and marketing were always misaligned, AI just made it expensive.

Where teams actually land

Most marketing teams I audit self-rate two levels above where they actually are. The Level 1 team with great licences self-describes as Systematic. The Level 2 team with one working pipeline self-describes as Compounding.

The gap is rarely the tools. It's the layer around them: shared context, integrated workflows, cross-team reach and IP that does not walk out the door at the end of a contract.

If you want your real level, not the flattering one, run the three questions that are hardest to wriggle out of:

1. If your strongest AI user left tomorrow, what survives? (Level 0 test)
2. Who owns how AI gets used, by name? (Level 1 exit test)
3. Does sales receive what marketing's AI produces, in a system, without re-keying? (Level 3 entry test)

You can buy your way into Level 1. You cannot buy your way out. Everything above it is built, owned and measured. That is the whole ladder.

Frequently asked questions

What is an AI maturity model for marketing teams?

An AI maturity model for marketing teams measures how deeply AI is embedded in how the team works, not how many tools it owns. The Focus4ward AI maturity ladder has four levels: Level 0 Prompt-curious (individuals use AI, the company does not), Level 1 Tooled (licences purchased, no operating model), Level 2 Systematic (shared context, owned workflows, one measured result), and Level 3 Compounding (the system improves every cycle and crosses team boundaries).

Why do most marketing teams overestimate their AI maturity?

Because they count tools instead of measuring the layer around them. A team with enterprise licences, several custom GPTs and an AI notetaker looks mature on paper. The maturity test is different: does the AI know the company's context, do workflows run end-to-end, is the IP versioned, is there a before-and-after metric, and does the capability survive when the strongest AI user is on holiday? In Focus4ward audits, most marketing teams self-rate two levels above where they actually sit.

What is the difference between buying AI tools and having an AI strategy?

Buying AI tools is procurement: licences, seats and subscriptions. An AI strategy is an operating model: who owns the standard of use, what context the AI has access to, which workflows run through it end-to-end, and how results are measured. Tool access without an operating model is Level 1 on the AI maturity ladder. You can buy your way into Level 1. You cannot buy your way out of it.

How does a marketing team move from Tooled to Systematic?

Three moves. First, appoint an owner for how AI gets used, not just who gets access. Second, encode shared context: brand, ICP and ways of working written down in a form a model can use, not trapped in one person's prompt history. Third, pick one workflow, run it end-to-end through AI, and track one metric before and after. Brief-to-first-draft cycle time is the strongest starter metric.