Contact Us

The Training Multiplier: Why 90% of Businesses See Zero from AI

Two groups of business professionals, one with a structured workflow, one still experimenting, the gap between AI activity and AI capability.

The Training Multiplier: Why 90% of Businesses See Zero from AI

The research on what separates the businesses getting results from the ones getting nothing is unusually consistent. Here’s what it says.

A few weeks ago I shared a statistic on LinkedIn that generated more response than most things I post.

A study of approximately 6,000 executives, published by the National Bureau of Economic Research in February 2026, found that 90% of companies report zero measurable productivity improvement from AI. Not marginal improvement. Zero.

Some people pushed back. Others said it matched exactly what they were seeing. The most useful responses asked a simple question: what are the other 10% doing differently?

That question has a clearer answer than most people expect.

What the research actually says

90%

Of companies report zero productivity gains from AI NBER, ~6,000 executives, Feb 2026.

5%

Of workers are considered genuinely AI fluent Google/Ipsos, 4,464 workers, Feb 2026.

14%

Of workers have been offered structured AI training Google/Ipsos, 4,464 workers, Feb 2026.

The NBER study found that 69% of businesses are already using AI, but average use sits at just 1.5 hours per week. A quarter of respondents reported zero AI use at all. Executives in the same survey forecast a 1.4% productivity increase over the next three years — a number that, if accurate, would make AI one of the most underwhelming technology investments in recent memory. (NBER via Fortune, February 2026)
 
Google and Ipsos research adds another layer. Of 4,464 US workers surveyed, only 14% had been offered any structured AI training by their organisation. Workers who received both tools and guidance were 4.5 times more likely to become genuinely fluent than those given tools alone. (Google/Ipsos via Fortune, February 2026)
 
Research into structured AI training investment consistently points in the same direction: organisations that invest deliberately in training see meaningful productivity increases, while those that rely on organic adoption alone see very little. These results hold across multiple studies, the specific returns vary, but the direction is consistent.

 
Two confirmed studies, plus a consistent body of training research. The conclusion is the same.
“The difference between the companies seeing results and the ones seeing nothing isn’t the tools. It’s structured training, executive alignment, and a repeatable framework.” Dovetail Digital
One AI-fluent worker in a workplace where most colleagues haven't yet built the same skills, the fluency gap most organisations don't see until it becomes a dependency problem.

Why AI activity and AI results are different things

When I look at the research alongside the businesses we work with, the gap isn’t surprising. It’s just rarely named clearly.
 
Most organisations introduced AI the same way they introduce most new tools. Someone champions it. A few people start using it. Some good results emerge: faster reports, cleaner summaries, better first drafts. Those results get shared at a team meeting. Everyone nods.

And then nothing changes systematically.

The person who got the results keeps getting them. Everyone else continues as before. The organisation has AI activity. It doesn’t have AI capability.

I’ve seen this in professional services firms, in hospitality and club management, in marketing and operations teams. The tools are there. The willingness is usually there. What’s missing is the shared structure that converts individual wins into consistent organisational practice.

The isolation trap When one person builds fluency and others don’t, the organisation becomes dependent on that individual. When they’re away, on leave, or move on, the capability goes with them. This is the most common and most costly form of AI adoption failure and it looks like success right up until it isn’t.

We’ve written about this dynamic in detail in Why AI Adoption Feels Busy but Doesn’t Stick.

What the 10% are doing differently

The research is specific about the solution. The businesses seeing measurable gains from AI share three characteristics. None are complicated. All require deliberate leadership decisions, which is why most organisations skip them.
 
1. They treat training as infrastructure, not an event
The common approach: a one-off session. A vendor demo. A half-day workshop ticked off the quarterly plan.

The businesses seeing compounding returns treat training differently. Structured prompts the whole team uses. Regular practice built into existing workflows. Clear standards for what good AI-assisted output looks like.
 
The result isn’t just that people know how to use the tools. It’s that everyone works from the same foundation. Quality becomes consistent. Results become replicable. The gains compound rather than staying with one person.

A senior manager at a recent Golf Management Australia leadership session illustrated this clearly. His board reports went from an hour to roughly seven minutes, not because he found a better tool, but because he’d already invested time in understanding what a good report looked like, what his board needed, and what tone carried authority. Structured practice preceded the time saving.
 
2. They align leadership before they scale tools
One pattern we see repeatedly: AI tools deployed broadly before leadership has agreed on the basics. What’s approved. What information can and can’t go in. Who reviews outputs before they leave.

When those questions haven’t been answered, staff either avoid AI to be safe, or use it without guardrails to get things done. Neither produces the results the business is looking for.

The organisations seeing consistent gains have had those conversations firs, not as a compliance exercise, but as a practical operational decision. Three clear answers to three clear questions is usually enough to get started.

We’ve written about the governance side of this in AI Isn’t Moving Too Fast — Organisations Are Just Struggling to Absorb It.

3. They use a repeatable framework, not ad hoc experimentation
Ad hoc experimentation produces early wins. Someone finds a prompt that works. A workflow gets faster. The business celebrates.

But without a framework, a structured way of capturing what works, sharing it across the team, and building on it, those wins stay isolated. The organisation resets to baseline every time someone leaves or a workflow changes.

A repeatable framework means the learning compounds. What one person figures out, everyone uses. What works in one function gets adapted for others.

This is the core of how we structure AI capability work, using the INGRAIN methodology as the operating backbone. Get Clear on intent and governance. Get Fluent through structured training and measured pilots. Get Consistent by embedding rhythm and decision rights. Get Growing as internal ownership takes hold. Each stage builds on the last. Skip one, and the next doesn’t stick.
Stick Figures - Collection of Images for website, blogs & newsletters (1)

What this means for an Australian service business

The research is primarily US and European. But this holds locally.

Across the Australian service businesses we work with, professional services firms, hospitality operators, golf clubs and member organisations, the dynamic is consistent. AI tools are being tried. Some individuals are getting results. The organisation as a whole is not building capability.
Stage What it looks like What it means
AI activity A few individuals getting faster results Results disappear when that person is absent
Early capability Shared prompts and workflows the team uses consistently Results are replicable but still fragile
Embedded capability AI is part of how work gets done — not an add-on Results compound as more people build fluency
Most organisations sit at stage one. A few are at stage two. The ones at stage three are typically the ones who started with a framework, not a tool.

The uncomfortable leadership question

AI-fluent workers are 4.5 times more likely to report higher wages and four times more likely to report a promotion. Fluency is already a career differentiator and right now only 5% of the workforce has it.

If your team isn’t building it, who is? And what happens when those who are, in other organisations, at other businesses, start to pull further ahead?

This isn’t a reason to panic. It’s a reason to be deliberate. The investment required to close the gap is modest relative to the return. Structured AI training consistently outperforms unstructured tool deployment across every study that has examined the question.

The businesses that act on this deliberately over the next twelve to eighteen months will have a meaningful advantage. Not because they moved fast. Because they moved with a plan.
“Training isn’t just about skills. It’s about creating the conditions for AI to be used reliably, not just experimentally.”  Dovetail Digital

Where to start

If you’re in the 90%, and most businesses are, the entry point doesn’t need to be complicated.
 
  1. Get an honest read on where you sit. What tools are being used? By whom? How consistently? Are results replicable, or individual?
  2.  Have the leadership conversation first. Agree on three governance basics before expanding anything: what’s approved, what goes in, who checks output.
  3. Invest in shared training not a single session, but a structured programme that gives the whole team the same foundation.
  4. Capture what works. Document it, share it, build on it. That’s how individual wins become organisational capability.
 
That’s the pattern the 10% are following. If you want to run the numbers on what a more structured training investment might return for your business specifically, the AI Training ROI Calculator does that in about two minutes.

Want to know where your AI capability actually stands? The AI Impact Report is a short structured assessment that shows where your AI activity is building genuine capability, and where the gaps are. Around ten minutes. Specific, prioritised recommendations.

Find out where you stand →

Frequently Asked Questions

Q1. Why do most businesses see zero results from AI despite using it?

Research from NBER (2026, ~6,000 executives) found that 90% of companies report zero measurable productivity impact from AI. The main reason is that most organisations deploy AI tools without building shared fluency, governance, or repeatable workflows. Without those elements, gains stay with individuals and don’t compound across the organisation.

Research into structured AI training investment consistently shows that organisations investing deliberately in AI training see significantly higher productivity returns than those relying on organic adoption alone. Tools without training tend to produce minimal results, while tools combined with structured guidance produce compounding gains.

According to Google and Ipsos research published in February 2026 (4,464 US workers surveyed), only 5% of workers are considered genuinely AI fluent. Only 14% have been offered any AI training by their organisation. Workers who receive both tools and structured guidance are 4.5 times more likely to become fluent than those given tools alone.

AI activity means individual staff are using tools and getting occasional results, but those results are not consistent or replicable. AI capability means the organisation has shared workflows, governance, and training that allow AI use to produce consistent results across teams, regardless of which individual is working.

Still feeling stuck? You’re not alone but you don’t have to figure it out solo.

Our DMA helps you cut through the noise and focus on what matters most.

Get Started With Our Digital Maturity Assessment Today!

Share this post with your friends.

Facebook-f Linkedin-in Instagram
Agency Arcade, About Us - Agency Arcade, Contact Us - Agency Arcade, Our Services - Agency Arcade