300+ Engagements. Here's What Separates the Companies That Actually Use AI from the Ones That Don't.

Most companies have the tools. Few have the outcomes.

I've been inside more than 300 client portals over the past two decades — sales orgs, marketing teams, RevOps functions, whole go-to-market motions. The pattern is hard to miss once you've seen it enough times. Two companies will buy the same tools, hire from the same talent pool, and spend roughly the same money. One of them turns AI into pipeline, faster cycles, and hours given back to the team. The other ends up with a pile of licenses, a Slack channel full of good intentions, and a quarterly report that says "exploring AI use cases." Same inputs. Opposite outcomes.

For a long time I assumed the difference was resources — that the winners simply had more budget, better engineers, or a sharper vendor. It isn't. I've watched lean teams run circles around enterprises with ten times the headcount. I've watched a company with one good operator and a modest stack out-execute a competitor with a dedicated AI team and a seven-figure tooling line. The thing that separates the companies that actually use AI from the ones that don't has almost nothing to do with what they bought.

It comes down to three operating choices most leaders never consciously make. None of them is technical. All of them are about how the company runs — who's accountable, how the work is shaped, and what gets measured. I've delivered more than 1,000 automated workflows across those 300-plus engagements, and the build was rarely the hard part. The hard part was always these three choices, and whether the company was willing to make them on purpose.

The 25% problem

Start with the number that should keep every executive up at night. The IBM Institute for Business Value CEO Study found that 85% of employees already have the skills to use AI. Actual utilization sits around 25%.

Read that again. The skills gap is mostly closed. Your people can use these tools. They're choosing not to — or more accurately, nothing in how the company operates is pulling them to. That's a 60-point gap between capability and behavior, and no amount of additional training or shinier software closes it, because training and software were never the constraint.

The same study found something else worth sitting with: 76% of CEOs now have a Chief AI Officer or an equivalent role, up from 26% just two years earlier. The org chart caught up fast. The outcomes didn't. Companies hired the title, bought the stack, sent everyone to the workshop — and still landed at 25%.

The skills gap is closed. The activation gap is wide open. Almost nobody is measuring the right one.

When a company tells me their AI initiative is stalling, they almost always describe it as a tooling problem or a talent problem. The instinct is to buy something or train someone. It's neither. It's an activation problem. And activation is a question of how you operate, not what you own.

I've sat in the meeting where a leadership team decides the answer to flat adoption is a second tool — a better copilot, a new platform, another pilot. It almost never works, because the first tool didn't fail on capability. It failed on activation, and the second one will fail the same way unless something about how the company operates changes first. The three choices below are that something.

Choice one: one accountable owner, not a committee

Here's the most reliable predictor I've found. Ask a company who owns AI outcomes. If the answer is a name, they're usually winning. If the answer is "the innovation committee" or "it's a shared priority across the leadership team" or — my favorite — "everyone owns it," they're stalled.

"Everyone owns it" means no one does. A committee can approve a budget and bless a roadmap, but a committee cannot be held to a number. It diffuses accountability across enough people that no single person's quarter depends on the outcome. So the work becomes a side project for everyone and the primary job of no one. It drifts.

The companies that move have a single accountable owner — one person whose job is to close the gap between AI capability and AI behavior, with the authority to redesign how work gets done and the mandate to report a real number every month. Sometimes that's a full-time Chief AI Officer. For most mid-market companies it doesn't need to be. It needs to be one person who owns the outcome, not a working group that owns the conversation.

This is the part leaders resist most, because naming an owner means someone can fail visibly. That's exactly the point. Accountability you can't see isn't accountability — it's a hope with a meeting invite. The committee survives because it protects everyone from owning the miss. The named owner is uncomfortable for the same reason it works: there's a person whose answer to "how's adoption" has to be a real one.

This is also why 76% of CEOs now have a Chief AI Officer equivalent and most still stall. The title alone doesn't fix it. I've seen the role created and then handed to someone with responsibility but no authority — they can convene and recommend, but they can't change how a single team actually works. That's a committee with one name on it. Real ownership means the authority to redesign the work, not just to report on it.

Choice two: redesign the workflow, don't bolt the tool on

The second choice is where most of the 25% actually gets lost.

The default move when a company adopts an AI tool is to drop it into the existing process and expect lift. The sales team still works the same sequence — they just have a tool that drafts the email now. The support team still follows the same script — there's a copilot suggesting responses they mostly ignore. The tool gets bolted onto a workflow that was designed for a world without it. Adoption stays optional, friction stays high, and within a quarter the license is shelfware.

The companies that get outcomes do the harder thing. They redesign the workflow around what the tool makes possible, then rebuild the process so the AI step isn't optional — it's load-bearing.

A concrete version. A Series B SaaS company I worked with was drowning in inbound lead follow-up. Leads sat for hours, sometimes days, and the good ones went cold before a rep got to them. The instinct — theirs, and almost everyone's — was to bolt an AI assistant onto the existing process so reps could draft replies faster.

I didn't build that. I redesigned the intake itself. Enrichment, qualification, and routing happened before a human ever touched the lead. By the time a rep saw it, the research was done, the fit was scored, and a first message was drafted. The rep's first action wasn't staring at a blank screen and a name — it was a decision on a pre-researched, pre-drafted opportunity. The AI wasn't a faster way to do the old job. It changed what the job was. Follow-up time collapsed because the slow steps no longer existed, not because someone was typing quicker.

And here's the part that matters for adoption: nobody had to choose to use the AI. It wasn't a button a rep could skip on a busy day. It was the path the work took before it reached them. That's what load-bearing means — remove the AI step and the workflow stops, so it never quietly becomes optional.

That's the difference between automating a task and redesigning a workflow. One saves a few minutes. The other removes the steps entirely. You can't get there by adding a tool to a process you refuse to change — and most stalled AI initiatives are exactly that refusal, dressed up as a rollout.

Choice three: measure adoption as a number, not a vibe

The third choice is the simplest to describe and the rarest to find. The companies that win measure adoption as a number. The ones that stall measure it as a feeling.

Ask a stalled company how adoption is going and you'll hear "really well," "the team's excited," "we're seeing great engagement." None of that is a measurement. It's a vibe. And a vibe can stay positive indefinitely while utilization sits at 25%, because nobody's looking at the actual rate.

The companies that move treat adoption like any other operating metric. What percentage of eligible workflows actually run through the AI step. How that percentage is trending month over month. Where it's high, where it's near zero, and why. When you measure it, two things happen. You see the truth — which is almost always lower and patchier than the vibe suggested. And you create the feedback loop that lets you fix the specific workflows where adoption died, instead of relaunching the whole program and hoping.

It doesn't have to be elaborate. One number, tracked monthly, broken down by workflow. The first time a company actually instruments it, the reaction is almost always the same — a quiet pause, because the real rate is so much lower than the story everyone had been telling. That pause is the most useful moment in the whole engagement. It's where the vibe dies and the work begins.

You can't manage what you won't measure, and you can't measure a feeling. The 25% number exists because most companies never built the instrument to see their own.

Why this is the AI Adoption Framework™

These three choices aren't a list I assembled to be tidy. They're what I kept finding underneath every successful AI rollout and missing from every stalled one — so I named the pattern. The AI Adoption Framework™ is how I take a company from "we own the tools" to "we operate on them": name the accountable owner, redesign the workflows so the AI step is load-bearing instead of optional, and measure adoption as a number you watch every month.

Notice what's not on the list. Not a bigger budget. Not a better model. Not more headcount or another round of training. I've seen companies with all four stall at 25%, and I've seen companies with none of them pull ahead — because they made the three operating choices and the others didn't. The gap between capability and outcome was never about capability. The tools are good enough. They've been good enough for a while.

The 85% is real. Your people can do this. Whether they actually do comes down to whether someone owns the outcome, whether the work is built around the tool instead of beside it, and whether you're honest enough to measure what's really happening. That's not a technology decision. It's a leadership one — and it's the one almost nobody is making on purpose.

If you want the version of this that goes deeper — the specific metrics I track, the workflows that tend to stall first, and how I'd diagnose your own activation gap — I write about it every week. That's where the practical detail lives.