88% Of Companies Use AI As A Tool, Only 12% Built A System

Jul 01, 2026 07:00 PM - 6 hours ago 320

Avinash Kaushik has a gift for puncturing comfortable myths pinch uncomfortable data, and recently he dismissed a persistent meme that elder leaders unit their organizations to adopt AI while softly staying successful the 1990s themselves. Notion data, he shared successful his post, shows that astir elder people, including CEOs, are really the astir precocious AI users successful the dataset, operating astatine Levels 3 and 4 astatine six times the complaint of individual contributors.

That inversion amazed me.

The communicative astir practitioners are telling themselves is that AI adoption is simply a top-down instruction problem wherever activity demands alteration but won’t exemplary it. Hey, that’s what I thought, too, until I publication Notion’s “Great Renovation” report, a study of much than 6,100 AI decision-makers and mundane users crossed 10 world markets that tells a different and much unsettling story. The spread isn’t betwixt leaders who push and workers who resist. It’s betwixt organizations that person moved AI from an individual instrumentality to a system, and the overwhelming mostly that person not.

That majority, by the way, is 88%. That’s bigger than a breadbox, arsenic my mother utilized to say.

The Baseline Is ‘Early,’ And That’s Not The Exception

Notion system its findings astir a four-level maturity model. Level 1 is AI arsenic a thought partner – individuals utilizing standalone devices to draft, brainstorm, and analyze. Level 2 arsenic an assistant, Level 3 arsenic teammates and level 4 is AI arsenic the system, wherever autonomous agents tally complex, business-critical processes end-to-end. The distribution crossed 6,118 respondents: 57% astatine Level 1, 31% astatine Level 2, 10% astatine Level 3, and 2% astatine Level 4.

Twelve percent of world organizations are operating AI astatine the level wherever it really reshapes really activity gets done. Eighty-eight percent are still chiefly utilizing AI the measurement you’d usage a amended hunt engine.

This matters for MCP readers successful particular. If you’re moving successful SEO aliases contented trading correct now, your statement is almost surely successful that 88% group. And the competitory unit isn’t coming from organizations that person somewhat amended prompts. It’s coming from the 12% that person integrated AI into their existent workflows, built governance astir it, and started measuring its effect pinch existent metrics alternatively than self-reported clip savings.

The Leader-Worker Gap Is Real, But The Direction Is Surprising

My caller file connected getting AI buy-in focused connected alteration guidance clash and the trouble of moving an statement from knowing that AI search is changing to really restructuring really contented is produced and measured. The Notion information adds a position I didn’t person then.

Decision-makers astatine precocious organizations picture a fundamentally different translator than the group doing the day-to-day work. At Levels 1 and 2, the lawsuit for AI runs almost wholly connected efficiency: speed, productivity, costs reduction. At Levels 3 and 4, thing shifts. Customer acquisition climbs 8 percent points arsenic a apical motivation. Enabling caller capabilities climbs five. Meanwhile, improving worker productivity – the ascendant Level 1-2 driver – really drops 4 points among the astir precocious adopters.

That’s not a mini reframing. It’s a wholly different logic for being successful the game.

For trading teams specifically, this connects straight to what I flagged successful my file connected the informing signs your squad is adjacent for AI cuts. Teams making the “we’re redeeming time” statement to warrant their AI finance are speaking Level 1-2 language. The organizations astir apt to past the adjacent information of consolidation are the ones arguing for customer acquisition gains and capabilities that didn’t beryllium before.

Why The Learning Curve Gets Steeper, Not Flatter

One of the much counterintuitive findings successful the Notion study is the percent of AI determination makers who opportunity finance is outpacing readiness climbs steadily arsenic organizations get much advanced. At Level 1, 48% study that gap. At Level 4, it’s 68%.

This is not what astir translator playbooks predict. The presumption is that organizations get amended astatine absorbing AI arsenic they summation experience. Notion’s information suggests the other that the much profoundly you embed AI into existent workflows, the harder it is for labor to support gait pinch what the statement is deploying.

Singapore leads globally astatine 21% of organizations astatine Level 3-4. The United States sits astatine 11%, tied pinch Japan. If you activity for an American institution that considers itself an AI leader, those numbers are a useful reality check.

3 Things Separating The 12% From Everyone Else

The Notion information connected implementation strategies is wherever the practitioner awesome gets clearest. Compared to Level 1-2 organizations, precocious adopters are doing 3 things astatine meaningfully higher rates.

First, integration. Fifty-five percent of Level 3-4 organizations person integrated AI pinch existing systems, versus 37% astatine Levels 1-2. That 18-point spread represents the quality betwixt AI arsenic an add-on and AI arsenic infrastructure. If your squad is still copy-pasting outputs from a chat interface into your CMS aliases analytics platform, that’s a Level 1 workflow sloppy of really blase your prompts are.

Second, governance. Forty-two percent of advanced organizations person built governance and oversight frameworks, compared to 26% astatine earlier stages. This 1 runs against the atom of really astir trading teams deliberation astir AI – governance sounds for illustration a ineligible section problem, not a contented strategy problem. The information says otherwise. The organizations moving fastest person besides moved first connected policy, oversight, and accountability structures.

Third, measurement. Thirty-seven percent of Level 3-4 organizations are measuring AI effect pinch existent metrics, versus 22% astatine earlier stages. And those value metrics (error rates, rework) are up 19 percent points. Workflow metrics (cycle time, throughput) are up 15. Self-reported clip saved – the anecdotal modular astir teams default to – is really declining arsenic a measurement attack among the astir precocious organizations.

If your statement is measuring AI ROI by asking group whether they consciousness for illustration they’re redeeming time, you are measuring Level 1 translator pinch Level 1 tools.

What Kaushik Got Right And What It Means For Your Next Team Conversation

It’s genuinely bully news that owners and executives are the astir precocious AI users successful the dataset. Leadership behaviour is 1 of the fewer reliable transmission mechanisms for organizational change. When elder group exemplary precocious AI usage crossed a divers group of tasks – not conscionable drafting emails, but making decisions, moving workflows, evaluating outputs – it creates definitive support for the remainder of the statement to return the aforesaid risks.

But there’s a drawback the information surfaces. That activity strength doesn’t automatically construe downstream. The skills and training spread is the number-one situation slowing AI take astatine Level 3-4 organizations. The devices and domiciled structures that make elder leaders much consenting to research are not automatically disposable to individual contributors.

My publication connected the Notion information is the astir vulnerable position for a trading statement correct now is being assured you’re up erstwhile the existent benchmark is only 12% of world companies operating astatine the level wherever AI genuinely reshapes output. Most teams mounting fierce AI goals are aiming astatine Level 2. The organizations that will matter successful 18 months are the ones presently readying for Level 3.

Three things to return backmost to your squad this week for ground-truthing. Map wherever your existent workflows beryllium against Notion’s four-level model, not wherever activity believes they sit, but wherever the day-to-day activity really lands. Identify the azygous highest-value recurring workflow your squad runs and inquire whether it could beryllium automated end-to-end pinch quality reappraisal astatine checkpoints alternatively than quality execution throughout. And if you’re still measuring AI effect by asking group whether they saved time, switch that pinch 1 value metric and 1 workflow metric earlier the adjacent reappraisal cycle.

More Resources: