What I've Learned Watching Organizations Adopt AI

AI · January 22, 2026

Over the past couple of years I've worked with a number of different organizations — mostly in the software sector: ISVs, SaaS providers, companies that consider themselves agile and pride themselves on adopting new technology quickly, experimenting fast, and failing fast. This post isn't meant to be prescriptive guidance. It's a collection of what I've observed along the way — patterns that seem to repeat across very different companies.

Why Organizations Adopt AI

The reasons vary, but a few come up again and again: competitive pressure, the need to ship more features faster, and growing demand from the market itself. None of this is particularly surprising — but it's worth saying that AI itself isn't new. It's been around the corner for years. What changed everything was transformers.

The Gardener and the Carpenter

There's a framing I keep coming back to, borrowed from psychologist Alison Gopnik's book The Gardener and the Carpenter (2016). Her argument — about parenting, not technology — is that a "gardener" creates the right conditions and then lets a child grow in their own direction, while a "carpenter" works to a fixed blueprint and tries to shape the outcome precisely. I think it maps remarkably well onto how organizations approach AI adoption.

The gardener's approach means paying attention to where AI adoption is already taking root organically — which teams or individuals are quietly experimenting and getting real results — and then giving those efforts more light, water, and space to spread. The questions worth asking are simple: where is this already working, and who's behind it?

The carpenter's approach is the top-down rollout: a detailed master plan for how AI will be used across the organization, decided in advance and pushed downward. The problem is timing — by the time such a plan is finalized, the technology has usually moved on. Leaders who insist on specifying exactly how AI should be used everywhere often end up rolling out a solution built for a problem that no longer matches today's reality.

Almost everything else in this post is really just an elaboration of the gardener's side of that distinction.

Culture Is the Real Driver

If there's one pattern that stands out above everything else, it's this: AI adoption has a direct relationship with a company's culture.

People embrace change when they feel incentivized — and that incentive is rarely financial. It's far more often social: recognition, a sense of progress, being part of something that's moving forward. And culture itself is usually set by middle management and senior leads, whether they realize it or not.

This is also where things go wrong. Organizations that try to mandate the use of AI tools often end up in a difficult spot — unable to justify the cost of usage, and losing track of whether it's actually improving productivity. Mandates create the appearance of adoption without the substance of it.

What Healthy Adoption Looks Like

My view is that AI adoption should grow naturally — and that happens when people operate in a safe environment, where a culture of innovation is genuinely embraced, and where people are empowered to share what they learn and get stronger every day. From there, it's up to leadership how they choose to treat it.

Organizations that embrace this well tend to do two things:

Give people access to the right tools.
Give them real guidance on how to get the most out of those tools — for example, how to configure a CLAUDE.md file so that Claude Code behaves the way your team needs it to, or how to use skills to automate the repetitive work that sits right at your fingertips.

On the other hand, the people who believe AI will replace them are usually the ones who resist the change the hardest. And ultimately, a lot of this comes down to listening — to people, and to their day-to-day work. Organizations that don't track how their people actually use these tools have no visibility into what's working, and they can't meaningfully celebrate adoption because there's no backbone behind the story.

Be Data-Driven

The most successful organizations I've seen pick the right metrics for each specific function and team, rather than relying on one blanket number. A few examples:

How many active users does the SRE team have on a given tool?
How much have we reduced incident resolution time?
How are we improving our runbooks, and are they genuinely available to engineers when they need them?

These are the organizations that can show success with data — not just anecdotes. Being data-driven isn't a nice-to-have here; it's the difference between a story you can tell with confidence and one you can't.

Don't Plan Too Far Ahead

One trap I keep seeing: planning in a very rigid, long-term way effectively means using yesterday's tool to solve tomorrow's problem — and that doesn't work in the current landscape. New models, new architectures, and new ways for agents to collaborate are emerging constantly. Plans need to stay adaptable.

Key Takeaways

A few things consistently make a difference:

Have an AI champion — someone who listens to their teams, works backwards from repetitive tasks and pain points, and is focused on helping people rather than showcasing themselves to leadership.
Team culture matters. The teams that succeed are the ones who understand that IT knowledge is always changing, who know that standing still means becoming rusty, and who see AI as a significant help and assistant rather than a threat.
Org culture, especially how middle management treats this, matters just as much. Agile organizations tend to find success here more easily.
Don't mandate usage — set realistic, clear goals instead. Something like "how can we increase our release frequency?" gives people a real target without forcing a tool on them.

It's also worth keeping an eye on initiatives from the wider industry. AWS, for example, has been pushing an approach called AI-DLC (AI-centric Development Life Cycle), built around two ideas:

AI-powered execution with human oversight — AI creates detailed work plans, actively seeks clarification, and defers critical decisions to humans, since only people have the contextual and business understanding needed to make those calls.
Dynamic team collaboration — as AI takes on routine tasks, teams come together in collaborative spaces for real-time problem solving, creative thinking, and faster decision-making, shifting away from isolated work toward high-energy teamwork.

None of this is a finished playbook — it's just what I've learned so far. If anything here resonates with what you're seeing in your own organization, I'd love to hear about it in the comments.