- The Augmented Advantage
- Posts
- AI Carpenters vs. AI Gardeners
AI Carpenters vs. AI Gardeners
And why typical AI maturity models fall flat
How do you prepare your organization for the AI age?
I still see a lot of organizations trying to get this "right" at first bat. Don't mess it up. Make sure you pick the right tools. The right use cases. Don't risk vendor lock-in.
While all of these are valid concerns, in reality they don't seem to work as expected.
91% of employees are already using AI tools, but only 1% of organizations consider themselves to have achieved AI maturity. These figures stem from recent McKinsey research, which inspired today's post because it finally put words to something I've been feeling for a long time.
We got the concept of "AI maturity" all wrong.
Let's dive in!
Why Classical Maturity Models Don’t Work for AI
Bring in any AI consultant and they'll show you their "AI Maturity Framework." The specifics vary, but the steps always look familiar:
Crawl: Identify use cases, run pilots
Walk: Build governance, create centers of excellence
Run: Scale across departments, integrate into core processes
It's neat. It's logical.
And it rarely works.
80%+ of AI projects die during prototyping. An MIT research shows that only 7% of organizations get AI mature despite following these "proven" approaches.
The problem isn't that companies are bad at following the steps. It's that AI doesn't work like previous technology waves.
With ERP or CRM, you could plan orderly rollouts because the technology did predictable things. You knew what success looked like at each phase.
But AI value emerges unpredictably. The "no-brainer" use cases hit unexpected integration obstacles. Meanwhile, value appears where you least expect it – like the legal team happily using ChatGPT to review contracts, saving hundreds of (super expensive) hours every week while your official AI Center of Excellence struggles with its 12-month roadmap. (And often wanting to shut down the legal use case because it makes them look bad – why didn't you come up with that obvious case, stupid!?)
Classical maturity models can't handle this chaos. They want orderly progression, measurable milestones, centralized control. So companies spend months building governance frameworks instead of nurturing what's already working.
I've felt this disconnect for the past two years but couldn't articulate it. Until I came across this McKinsey report that describes a concept from child psychology called "The Gardener and the Carpenter."
It explains AI adoption better than any maturity model I've seen.

The Carpenter vs. The Gardener
Back in 2016, developmental psychologist Alison Gopnik found that parents should allow children to develop according to their natural tendencies rather than predetermined constructs. Don't force "ideal paths." Instead, "nurture what grows" to develop healthy and successful people.
She calls this "a gardener's mindset."
It's easiest to understand by contrasting it with a carpenter's approach.
Carpenters plan everything meticulously. Measure twice, cut once. They have to. Cutting expensive tile wrong means throwing away the whole piece. Poor planning means ordering too little material and facing angry clients. As the old saying goes, "Give me six hours to chop down a tree and I will spend the first four sharpening the ax."
(Turns out, for AI this mindset doesn’t work.)
Gardeners work differently. They know they can't control everything (who knows next week's weather?). So they manage risk by diversifying – not sure if strawberries will grow? Plant some tomatoes too. Different outcomes, but both can make a great meal.
Most importantly, gardeners observe constantly. Where are the green shoots appearing? Let's water those.
AI has been rolled out using the carpenter's mindset – carefully planning every detail from the top down, often manifested in six-figure strategy decks (we paid so much, it has to be good!). Hence the maturity model obsession: Stage 1, Stage 2, Stage 3.
But what if we tried the gardener approach instead?
After working with 50+ organizations, I've seen that curiosity and adaptability – the gardener's core traits – are what separate AI success stories from the 80%+ that fail.
The Jobs of an AI Gardener
As an AI Gardener, you have three core responsibilities:
1. Know And Cultivate Your Ecosystem
When you're in the desert, your strategy for growing tomatoes must be wildly different than in Italy. Maybe it’s impossible altogether. Gardeners know their constraints and possibilities. They can't control the weather, but they can build a greenhouse.
Same with your organization. What's your "AI climate"? A highly regulated bank operates differently than a scrappy startup. Legacy systems, risk tolerance, cultural norms. These shape what AI can grow where.
2. Plant and Nurture Selectively
Not every plant belongs in your garden. Some might be hazardous to the whole ecosystem. You have to rip those out.
It's fine to shut down AI initiatives that violate ethical or governance standards or use tools that don't have any perspective of scaling beyond the proof-of-concept phase.
But the real gardener work is on the flip side – planting the seeds and helping the good sprouts flourish – from Prototype to Production.

“Gardening” AI use cases from prototyper to production
Instead of building an "AI strategy" top-down, you define a high-level ambition level, identify the people already making AI work (often without official permission) and help them scale. Or figure out why promising experiments are failing.
Maybe that finance team's forecasting experiment needs better data access. Perhaps the customer service ChatGPT usage reveals how your entire support model could be reimagined.
3. Observe Everything
Every gardener watches new seeds carefully. But they also scan for unexpected sprouts. Plants they didn't plant but are thriving anyway. And how they affect the garden.
In organizations, this means asking two questions:
Where should AI be creating value? Based on your competitive position, which business domains seem ripe for AI transformation? Customer acquisition? Product development? Supply chain? These are your "perfect apple-growing conditions"—you'd be surprised if AI wasn't sprouting there.
Where is AI already sprouting organically? The HR team using Claude for job descriptions. Engineers using Copilot for code reviews. Instead of shutting these down, study what makes them work.
The magic happens when these perspectives intersect. The finance team's forecasting experiments might reveal your entire budgeting process is ready for systematic transformation.
The key is having a theme rather than a rigid plan. Are you building for moonshot innovation? ROI-focused efficiency? A strategic mix?
This leaves the notion of "maturity" behind entirely. Your garden next year might look completely different if conditions change. Who would say they have a "mature" garden anyway?
How Gardeners Solve The 2-Level Problem of AI Adoption
The gardener approach can actually solve the gap between individual AI productivity and real system transformation
Level 1: Individual Productivity - People using ChatGPT, Claude, or Copilot for personal tasks. This is where most organizations are stuck. Having people use AI for email drafts, or report summaries.
Level 2: System Transformation - Redesigning processes, redefining roles, and reimagining how work actually gets done. Very few unlock this level. But it's where the real competitive advantage lives.
The carpenter approach tries to jump from Level 1 to Level 2 through planning. The gardener recognizes that Level 1 teaches you where Level 2 transformation should happen.
But only if you're paying attention.
How do you measure gardening success?
Forget ROI for a moment. ROI measures single use cases, not your overall approach. To evaluate if you're a good gardener, look at four key metrics:
Domain Integration Depth: How many critical business domains are you actively "gardening"? Not just using AI tools, but fundamentally rethinking how work gets done.
Organic Innovation Rate: How many AI experiments sprout naturally from your teams each month? High rates suggest good experimentation culture.
Competitive Differentiation: Is your AI creating unique advantages competitors can't easily copy? Or are you just using the same tools as everyone else?
Process Redesign Velocity: How fast can you go from idea to execution? Can you make the jump from simple assistants to more complex autopilots or even agents?
These aren't maturity levels, but health indicators for your innovation ecosystem.
Conclusion
Most leaders will probably read this and nod along, then go back to building their six-figure roadmaps. They'll keep hiring consultants to create maturity assessments and governance frameworks while their competitors are already harvesting what they planted months ago.
But if you're ready to ditch the maturity trap, start asking different questions:
How should my AI ecosystem look like?
Where is AI already sprouting in your organization?
Which experiments are showing unexpected promise?
What can I do to nurture the right growth?
Get your hands dirty. That’s how gardeners are getting ahead.
And becoming ultra competitive.
See you next Friday!
Tobias
Reply