Use AI to Build and Maintain a Skills Directory for Your Team

Why nobody can answer “who already knows how to do this”

Say a client services manager named Priya runs a team of nine. Her company just rolled out a new AI transcription tool, and she needs someone to turn raw call notes into client-ready summaries automatically, ideally by Friday. She posts in the team channel: does anyone already know how to set this up? Two people reply within the hour. Both have been doing something close to this for months, quietly, for their own accounts, because nobody had ever asked.

Priya isn’t disorganized. She runs weekly stand-ups, keeps a tidy project tracker, and knows her team’s workload cold. What she doesn’t have is a record of who can actually do what with AI, because that kind of capability rarely shows up in a job title, a performance review, or a project plan. It shows up in a Slack thread nobody searches, or in a habit someone picked up on their own time and never mentioned to anyone.

That’s the specific gap this article is about: not a scoring system for how good someone is at AI, but a living register of who on a team already has which AI skill, kept current enough that a manager can actually use it. Think of it the way a good IT department keeps an asset register. Nobody scores a laptop out of five. They just need to know it exists, who has it, and whether it still works.

A directory like that has one job: make invisible capability visible before it turns into an emergency.

There’s a reason to be specific about that. “AI skills” content could mean a scored proficiency matrix, a hiring rubric, or a plain list, and each of those is a different document built for a different decision. This piece picks one concrete version on purpose: a directory that answers “who can do this,” not “how good is everyone at this.” That choice matters because the two questions get answered by different documents, built and kept current in different ways, and trying to make one document answer both usually means it answers neither well.

Part of why the gap exists in the first place is that a lot of AI use at work is already happening off the books. In Microsoft and LinkedIn’s 2024 Work Trend Index, 78% of people who use generative AI at work say they’re bringing their own tools rather than ones the company issued, and that share climbs to 80% at small and mid-sized companies.^[1] Nobody is hiding anything maliciously. They just never had a reason to announce it, and no manager ever asked.

Skills directory vs. skills matrix: same word, two different jobs

Future Factors already has a companion piece on this site about building an AI skills matrix, and it’s worth being precise about how that differs from what’s covered here. The skills matrix scores each person against a proficiency scale, usually 1 to 5, across a defined list of skills and domains. It’s built to answer questions like who needs training first, and where the biggest gaps sit numerically across a department.

A skills directory answers a narrower, more practical question: who, right now, already knows how to do this specific thing. No scoring, no 1-to-5 scale, just a name, a described skill, and a way to reach that person. It reads closer to a company directory than a performance review.

A rough rule for which one you actually need:

• You’re deciding who to ask for help on a specific task this week: build a directory.
• You’re deciding where to invest a training budget across a whole department: build a matrix.
• You’re doing both, which is common: keep them as two separate documents that link to each other, rather than one document trying to do both jobs.

Plenty of teams try to force a single document to do both jobs, and it usually satisfies neither. A directory cluttered with proficiency scores starts feeling like a performance review, so people either inflate their own entries or quietly stop filling it in. A matrix padded out with contact details and daily-use notes gets too heavy to keep current, so it goes stale within a quarter. Keeping the two separate keeps each one usable on its own, rather than adding bureaucracy.

If you’re not sure which one your team needs first, start with the directory. It’s the cheaper document to build, it produces something useful within a day, and it tends to surface the exact gaps a matrix would later put a number on anyway. Teams that build a matrix first, before anyone has a working directory, often end up scoring people against skills nobody has actually mapped to a real person yet.

What actually goes in a skills directory

Strip away the spreadsheet formatting and a skills directory needs to answer four questions for any given skill: who has it, how they actually use it today, whether that’s still current, and who to contact first. Everything else is optional polish.

At minimum, each entry needs:

• The person’s name and role
• The specific AI skill or tool, described concretely (not “good at AI,” but “builds Zapier automations that pull CRM data into a weekly report”)
• Where they actually use it today, not just where they were trained on it
• When the entry was last confirmed as still accurate

Notice what’s missing: no score, no ranking, no comparison between Dana and Marcus. The table only has to be honest and current. That second part is the harder one, and it’s the subject of the rest of this piece.

A worked example: how a 12-person ops team built theirs in an afternoon

Picture a common one: a 12-person operations team at a mid-size logistics company decides to build a directory after a scheduling snag. A dispatcher who’d built a working AI routing script left for a new job, and it took three weeks for anyone to realize nobody else understood how it worked, let alone how to maintain it.

The team lead didn’t hire a consultant or buy software. She sent a five-question survey: what AI tools do you use for work, what do you actually use them for, how often, how confident are you that you could explain it to a colleague, and who else on the team do you think already does something similar. Eleven of twelve people answered within two days.

The answers surprised her. Three people were already using AI to draft client emails, one was using it to build spreadsheet formulas nobody had documented, and the router builder’s usual backup, it turned out, already understood most of that script because the two of them had debugged it together once, months earlier. None of that had ever been written down anywhere a manager could see.

The directory itself took less time to build than the survey did to write.

What made it stick wasn’t the format. It was that the team lead treated the first draft as provisional, something to be corrected in public rather than filed away as finished. She posted it in the team channel and asked people to fix their own entries, which caught two small mistakes and surfaced one skill nobody had claimed for themselves but everyone agreed belonged to a specific person.

None of this required a special tool. The survey went out through whatever form software the team already used, the draft lived in a shared spreadsheet, and the only new habit was a five-minute agenda item at the next stand-up to walk through it out loud. The hardest part wasn’t building the thing. It was deciding, on the spot, who would be the one checking it next month.

Keeping it alive: ownership and the refresh habit

Every directory like this starts accurate. The failure mode tends to look the same: it goes stale, quietly, and nobody notices until it hands someone bad information at exactly the wrong moment.

That’s really the whole argument of this article. A skills directory only earns the name if someone is accountable for keeping it current. Without that, it’s an archive of who could do what six months ago, dressed up to look like a live system.

Who owns that upkeep depends on where you sit. A direct manager is usually the right owner for a single team’s directory: they already run the stand-ups where this information surfaces naturally, and they’re the one who needs it most urgently when staffing a request. An individual contributor’s job is simpler: keep your own entry honest, and mention it when someone else’s skill has quietly expanded or gone stale. A department head shouldn’t personally own any one team’s directory, but should own the standard that every team keeps one, and ask to see it during quarterly planning instead of leaving it to chance.

A few signs a directory has quietly stopped being one:

• Nobody has edited an entry in more than a quarter
• New hires never get added
• The person people actually go to for help isn’t the person listed

Where AI actually helps you build and maintain it

There’s a reasonable question buried in the idea of using AI to build a skills directory: what part of this should AI actually be doing? Not scoring people, and not making the staffing call. Where it helps is turning slow, uneven inputs into a usable first draft.

A short survey like the one in the earlier example produces messy, inconsistent answers: some people write two words, others write three paragraphs. AI is genuinely useful for turning that mess into a consistent first-pass entry someone can then correct. It’s a drafting tool for the directory itself, not a judge of what belongs in it.

None of this requires special software. A shared spreadsheet and a survey tool most teams already have will get a first version built in an afternoon. Where structured facilitation earns its cost is when the directory needs to scale past one team, when categories need to stay consistent across a whole department, or when building it is also meant to surface where formal training would close a gap the exercise just revealed. Teams can absolutely start this on their own. Expert guidance tends to shorten the distance between a rough first draft and something leadership actually trusts.

What a live directory catches that a headcount plan never will

A directory earns its keep in the moments a headcount plan can’t help with: the Tuesday afternoon when a client asks for something nobody budgeted time to learn, and someone on the team already half-knows how to do it.

It also catches the reverse problem: skills that look covered on paper but aren’t, in practice, current. Someone who took a course eight months ago and hasn’t touched what they learned since isn’t in the same position as someone using a tool daily. A live directory, updated on the cadence described earlier, tells the difference. A training completion record generally can’t.

This gap between training completed and skill actually retained is a real pattern, not a hunch specific to one team. In the same 2024 Work Trend Index, only 39% of people globally who use AI at work say they’ve received any AI training from their employer.^[1] Most of the AI capability already sitting inside a team grew informally, which is exactly the kind of thing a training log can’t show a manager, and a directory can.

If you’re going to check whether the directory itself is working, resist the urge to measure it by whether people filled it in once. Usage alone is a weak signal: track whether entries keep getting updated past the first month, which is persistence, and whether staffing decisions or gap-closing training actually changed because of what the directory showed, which is impact. A directory that got used once and sat untouched after that mostly failed, even if it looked complete on launch day.

When nobody consults a directory at the exact moment a real decision comes up, that’s rarely a failure of the format. Most often it means the ownership step got skipped.

If your team doesn’t have one of these yet, don’t start with a template. Start with three questions sent in a normal channel this week: what AI tools do you already use for work, what do you actually use them for, and who else do you think does something similar. Whoever reads those answers first should also be the name attached to keeping them updated going forward. That’s the whole first step.

Sana Mian, Co-Founder of Future Factors AI

Sana is an AI educator and learning designer specialising in making complex ideas stick for non-technical professionals. She has trained 2,000+ learners across corporate teams, bootcamps, and keynote stages. Future Factors offers AI Bootcamps, Corporate Workshops, and Speaking & Consulting for businesses ready to adopt AI without the overwhelm.

More about Sana →