A Baseline for AI Fluency: Your Business, Your Team, Yourself

When Mayank Gandhi and I wrote about moving from AI-curious to AI-native after the 2025 Skift Data + AI Summit, one question followed me into nearly every conversation afterward. How do we know where we actually are?

It sounds like a simple question. Watch what happens when an organization tries to answer it. In the past year I have seen a chief executive draft her investor update with an AI assistant at midnight while her team retyped meeting notes by hand the next morning. I have seen a business buy four hundred licenses and settle at sixty weekly active users. I have sat with a team that quietly automated half its workload while the board debated whether AI deserved a standing line on the agenda. Ask each of these organizations where it stands on AI and each will give you a single answer, one stage on one ladder. Each answer will be wrong the same way. It averages out the very gap that decides whether AI creates value.

The honest numbers.

It is worth being clear about what the adoption data says. McKinsey's 2025 State of AI research puts organizations using AI in at least one business function at 88 percent. Organizations attributing any earnings impact to AI sit at 39 percent. High performers, earning 5 percent or more of their results from AI, are roughly 6 percent. Cisco's AI Readiness Index finds 13 percent of organizations fully ready. Adoption is close to universal. Advantage is rare.

The number that explains the gap comes from BCG's AI at Work 2025 study. Around 80 percent of leaders now use generative AI regularly, against 51 percent of frontline employees, and the frontline figure has stopped moving.

The line between AI adoption and AI advantage does not run between companies. It runs through them.

Somewhere in your business, the leader is ahead of the team, or the platform is ahead of the people, or a fluent team is boxed inside a curious business. That gap is the diagnosis, and a single score hides it.

One position, three lenses.

The way through is to stop asking for one number and take three readings at once. The first is you, personally. Your knowledge, your practice, your judgment with AI, scored honestly and seen by nobody else. The second is your team, the people who report to you and what they can actually do, measured rather than assumed. The third is the business as a whole, how it operates in everything it does, at the level the P&L requires. Your real position is the intersection of the three. Each lens moves at its own speed, and the gaps between them decide what you feel every day. Friction, drag, or compounding.

I avoid the word maturity deliberately. Maturity implies a finish line and a judgment, and there is neither. Positions move in both directions, and the frontier moves faster than any of us. A continuum is the honest shape.

Five stages, one identity each.

The same spine applies to every lens. What changes is the evidence.

1. AI Curious. I follow it. Interest is real and regular, but structured knowledge is not yet there.
2. AI Literate. I understand it. Knowledge is in place, but application is not yet mapped to the business.
3. AI Aware. I see where it applies. Opportunity is mapped and prioritized, but sustained practice has not started.
4. AI Fluent. I work with it. Practice is sustained, judged, and measured on real work.
5. AI Native. It is how I work. AI-first is the default design of the work, and exceptions are logged.

Aware sits deliberately between Literate and Fluent. Knowing how AI works and knowing where it applies in your business are different capabilities, and the second is where most leaders stall. Naming the stage makes the stall visible.

Position is earned, not declared.

Self-assessment is where frameworks go to flatter, so a baseline has to run on evidence rather than opinion. In the working sessions I run with executive teams, every stage on every lens carries five capabilities, and each one is binary. The proof exists today or it does not. A document, a metric, a dated example, a named person, produced in under a minute. You hold a stage only when all five are met, together with every capability at every stage beneath it. Four of five holds you at the stage below.

Before any scoring, I ask each leader to mark an instinct. Where do you sit, where does your team sit, where does the business sit. Then comes the evidence pass. Most leaders land one stage lower than they assumed, and nobody in the room is exaggerating. Instinct averages. Evidence locates.

One example from a recent session. A leader cleared Curious, Literate, and Aware in full, then met four of five capabilities at Fluent. She holds AI Aware. The single unmet capability, in her case a standard for verifying AI output before it ships, became the first line of her ninety-day plan. Nothing else needed deciding.

The misses are the plan.

Two numbers, one picture.

Once the three lenses are scored, the picture collapses into two numbers. Your anchor is your lowest lens, the stage you actually operate at. A Fluent leader inside a Curious organization operates Curious. Your spread is the distance between your highest and lowest lens, counted in stages. Zero is aligned, one is workable, and two or more is live tension with a predictable failure mode.

Go back to the three companies at the top of this piece. The midnight chief executive is a leader ahead of both team and business, and her mandates will stall, because tools pushed onto a team without literacy burn credibility and budget. The four hundred licenses describe an organization ahead of its people. Spend without value, pilot purgatory, an investment case quietly eroding. The automating team is a surge from below, shadow AI spreading beneath a leader whose credibility gap grows by the week. And the most common pattern in the market right now, on the BCG numbers above, is the frontline gap, where leader and organization align on strategy while adoption thins exactly where the work is executed. None of these patterns is a verdict. Each is a forecast of the friction ahead, and the order in which to remove it.

Where to start.

None of this requires a program, a committee, or a new budget line. It requires one accountable leader and an honest hour.

1. Score yourself first. All five stages, evidence only, seen by nobody else. Landing lower than you assumed is the point, not the penalty.
2. Measure the team rather than polling it. If adoption has never been measured, the capability is not met, and running the measurement is itself the first move.
3. Score the business you have, not the program you announced. Announcements are not evidence.
4. Read the anchor and the spread, and name your pattern. Ten minutes, and the friction of the next two quarters becomes legible.
5. Let the misses write the plan. Give the first unmet capability an owner and a date, and re-score in ninety days.

For twenty years my work has come back to one conviction, that the power of data should be accessible to everyone rather than reserved for the largest players. AI raises that stake. The capability on offer has never been greater, and the distance between the companies that adopt and the companies that gain has never been wider. A baseline does not close that distance by itself, but I have never seen anyone close it without knowing exactly where they started.