All insights
Perspective

The AI value gap

Most companies now use AI. Far fewer have changed how the work gets done. That difference, more than budget or size, is what separates the ones profiting from it.

AXXEM AI ·

Using AI is no longer the question. Profiting from it is. And the distance between those two things is now large enough to measure.

Almost everyone uses it. Almost no one profits.

Three recent studies, three different methods, one striking pattern:

  • A study of 1,250 executives: about 5% of firms are "future-built," capturing real value from AI at scale.1
  • A survey of 1,993 respondents at all levels: 88% report regular AI use in at least one function, but only 39% report EBIT impact at the enterprise level, and only ~6% are "high performers" attributing 5%+ of profit to it.2
  • A preliminary industry analysis: roughly 5% of custom enterprise gen-AI tools reach production with measurable P&L impact, though general-purpose tools like ChatGPT do far better.3

The three measure different things (maturity, profit attribution, pilot success) yet they converge on the same shape: a small minority captures nearly all the value.

A newer study reframes the bar more loosely and still finds 20% of firms capturing 74% of AI's economic value, a different threshold telling the same story.4

Worth attaching one honesty note: the widely repeated "95% of pilots fail" figure comes from that same preliminary, contested analysis. It belongs alongside this convergence, never as a standalone stat.

The gap is deliberateness, not budget

For a small business, the finding that matters is this. What best predicts AI-driven profit is whether you redesign how the work is actually done. Spend and headcount matter far less.

Among those 1,993 respondents, the high performers were about 2.8× more likely to have redesigned their workflows (55% vs 20%), while lower performers bolted AI onto unchanged processes.

BCG's "10-20-70" guideline points the same way, as a rule for where to put your effort: about 10% goes to the algorithm, 20% to the surrounding technology and data, and 70% to the people and process changes that make it stick.5

If the value lives in judgment and process more than in budget and scale, then the gap isn't rigged against small businesses. The lever is one you actually control.

What the winners actually do

  • They redesign one workflow deeply, while everyone else sprinkles AI thinly across everything.
  • They buy and integrate the foundation and build only the thin, business-specific layer on top of it. That is our engineering stance, and it cuts against the current: a tooling vendor's 2026 survey of 817 people who build internal software found a surge in custom builds, much of it ungoverned "shadow IT."6
  • They measure the value and invest in adoption, the people part that makes it stick.

What this means for your business

The AI value gap runs between the deliberate and the passive. Size has very little to do with it.

The underlying tiering makes the point better than a binary does: roughly 5% are "future-built," but another 35% are scaling and starting to generate value, and about 60% are not yet getting material value from it. The dividing line is the redesign itself.

For a small or mid-sized business, that is the entire opportunity: you can be deliberate without being big.

See where you stand — take the free 2-minute AI Readiness Snapshot, or book a 45-minute call.


  1. BCG, The Widening AI Value Gap: Build for the Future 2025 (Sept 2025; n=1,250 executives). 

  2. McKinsey, The State of AI in 2025 (Nov 2025; n=1,993 participants across 105 nations, fielded Jun–Jul 2025). n counts respondents, not distinct organizations. 

  3. MIT NANDA, The GenAI Divide: State of AI in Business 2025 (preliminary v0.1, Jul 2025). 

  4. PwC, 2026 AI Performance Study (Apr 2026; n=1,217 senior executives, primarily at large, publicly listed companies across 25 sectors). This is not a small-business sample. 

  5. BCG, Scaling AI Requires New Processes, Not Just New Tools (2026). BCG frames 10-20-70 as a practitioner guideline for where to put scaling effort, not as an empirical finding. 

  6. Retool, The Build vs. Buy Shift (Feb 2026; n=817 "builders" including Retool's own customers, of whom only 36% are software engineers or developers). Published by a platform vendor surveying its own orbit.