Every Team Needs an AI Builder

A finance analyst on a planning team got tired of stitching together the weekly variance deck. Every Friday, he pulled numbers from three systems, dropped them into a template, wrote the commentary by hand, and chased two people for sign-off. It ate four hours.

One Friday, instead of doing the deck, he wrote a Claude skill. It pulled the numbers from the three systems, drafted commentary in last quarter's tone, flagged the rows that looked off, and posted the draft into Slack for sign-off.

By Monday, two of his teammates wanted one too. By the end of the month, the team had reclaimed fifteen hours a week and the variance deck was no longer the bottleneck. Nobody from the top asked for it. The CEO did not send a memo about variance decks.

This is what AI transformation looks like at the working level. One builder. One task. One team. Compounding.

Top-Down Mandates Do Not Always Work

The dominant playbook today is the top-down memo. Shopify CEO Tobi Lutke published an internal memo in April 2025 asking every employee to prove AI could not do a job before requesting more headcount. Meta tied AI use to performance reviews. Plenty of others followed.

Roughly 8 in 10 enterprise workers are avoiding or rejecting the AI tools their employers are pushing. The communication gap is even wider. 59% of executives say they have communicated a clear AI vision to their workforce. Only 8% of employees agree.

This is not a story of bad leaders or lazy employees. It is a structural mismatch. Mandates flow from the top down. The friction lives at the bottom.

The mandate has a place. It signals that AI matters. But on its own, it is just a memo.

The Friction Lives Close to the Work

Leaders see inefficiencies in aggregate. Teammates feel them in detail.

A finance team's pain is not a customer success team's pain. The annoying parts of a recruiter's week look nothing like the annoying parts of a security analyst's week. Most of the wasted time sits in team-specific friction that only shows up from the inside.

This kind of friction does not show up in a survey or a quarterly review. It shows up at standup, in side messages, and in the small workarounds people have stopped questioning. Only someone who sits with the team, doing the team's work, sees it clearly enough to fix it.

Central AI platforms are essential. They provide model access, the security boundary, the data plumbing, and the shared building blocks that make everything else possible. Without them, every team reinvents the basics and nothing scales.

But the platform alone cannot know that the planning team needs commentary in last quarter's tone, or that the support team needs triage that respects the unwritten rule about VIP customers. That knowledge lives with the team.

The best setups pair the two. A strong central platform supplies the foundation. Teams build on top of it for the details only they can see.

The unit of automation has to match the unit of context. The platform sets the floor. The team fills in the rest.

Champions Train. Builders Ship.

The closest thing the industry has tried is the AI Champion model. Citi built an internal network of 4,000 AI Accelerators supported by 25 to 30 AI Champions, embedded across 84 countries. They reached over 70% adoption of approved AI tools. The peer-led model worked where memos did not.

Champions are valuable. They host demos, answer questions, lower the activation cost of trying a new tool. But champions advocate.

Builders go out and build the thing, then hand it to the team.

The next layer is having a person on every team who builds the automations the team actually needs. Not a coach. Not an evangelist. A builder. Someone who sees the variance deck problem on Wednesday and ships the fix on Friday.

What an AI Builder Does on a Team

A day in the life looks ordinary.

They sit in the team's standup. They do the team's work. They overhear the things that never make it into a ticket. The complaint about the format that keeps changing. The workaround everyone shares but no one wrote down. The Friday ritual that eats four hours.

They know which customer the team treats differently and why. They know which spreadsheet is load-bearing. They know the unwritten rules. That context is the whole point of the role.

They go and build the smallest thing that removes the friction. A Claude skill. A Slack bot. A small agent. A short script. An MCP server wired to one internal tool. Whatever fits the problem.

They ship it for one person first. They watch how it gets used. They fix the rough edges. Then they share it with the team.

Then they move to the next friction.

Their toolkit compounds. The second automation reuses parts of the first. The third reuses parts of the second. By the tenth, they have a small library of patterns that fit how their team actually works. This is the same shape I described in The Rise of the AI Builder, now applied at the team level instead of the role level.

Inside the Team, Not Beside It

The shape of the role matters.

Put all the builders on a central platform team and they end up too far from the work. They do not sit close to any one team, so they ship slowly and end up serving whoever shouts loudest.

Ask every person on the team to be a builder and no one owns the team's friction. "Everyone should use AI" is just the mandate again.

The right number depends on the team's size and friction surface. The constant is that they sit inside the team, not outside it.

• Context. They live inside the team and feel the daily friction. They can think from first principles about why a particular issue even exists, and whether it can be redesigned from the ground up.
• Leverage. They ship instead of queueing.
• Compounding. They learn the team's workflows, quirks, and unwritten rules over months.

The Compounding Effect

Time math makes this concrete.

Week one. One annoying task automated. Thirty minutes a day saved for one person.

Month one. Five small automations live. The team stops asking "should we be using AI?" and starts asking "can we automate this part?"

Month three. The rhythm of the team is different. Reports draft themselves. Triage runs overnight. Onboarding docs update on their own when a process changes.

Other team members start picking up the toolkit. The builder stops being the only one who can do this and starts lifting the whole team. The team becomes AI-native at the working level.

This is the bottom-up half of AI transformation. It is not a press release. It is six months of one person quietly removing friction.

How to Spot or Develop One

The good news is they probably already exist on most teams.

They are the person who built the Slack script nobody asked for. The one who quietly automates their own work and never tells anyone. The teammate who shows up in standup with a new tool every two weeks.

At heart, an AI builder is a problem solver. They cannot let a broken process sit. A repetitive task itches at them until they fix it. They are the person who hears "we just do it that way every week" and immediately wants to know why. That instinct, paired with AI tools, is the entire job.

Protect their time. Give them the AI toolkit. Make it explicit that this is part of their job, not a side project they have to hide.

If the role is absent, you can develop one. Pick a strong domain expert who is curious about how the team works. Pair them with a few weeks of focused AI tooling time. The bar is no longer "can they write production code." With Claude Code and similar tools, the bar is now "can they describe what they want clearly and iterate."

You can also hire one. Look at what they have built more than where they have worked. LinkedIn's Associate Product Builder program asks for a 60 second demo of something the candidate shipped, not a resume. Apply the same lens.

One Builder, Every Team

Mandates push. Builders pull. The push has a place, but without a builder underneath, nothing actually moves.

The pattern that works does not require a transformation program or a new department. It requires one person on every team, embedded in the work, automating one task at a time. From there, the compounding starts.

The companies that win the next phase of AI will not be the ones with the loudest mandate. They will be the ones who quietly put a builder on every team, where the real change starts.