AI adoption: what to measure and what actually moves it
Licences and logins show access, not change. Here is what 2026 data says about AI adoption, how to measure it, and a 90-day plan to move it.
BY STEFAN MARITZ··8 MIN READ
TL;DR
- Licences issued, logins, and training hours show that people have access to AI. They say little about whether anyone's work has changed.
- Gallup's Q2 2026 survey found that 52% of US workers use AI in their role, and 30% use it a few times a week or more.
- Gallup links frequent use most closely to two conditions: manager support and AI that fits into existing workflows. Only 36% of employees in AI-integrating organisations strongly agree that their manager actively supports their team's use.
- Deloitte reports that 84% of organisations have not redesigned jobs or workflows around AI, which is where most of the value sits.
- Measure the workflows that changed and what they now cost. Treat tool counts as inputs.
- AI champions help when they are picked by workflow, given protected time, and backed by managers who use the tools in front of their teams.
Most AI adoption dashboards count licences, logins, and training sessions. Those numbers are easy to produce, which is why they dominate. They also tell you very little about whether a single workflow in your business runs differently than it did a year ago.
This guide covers what the 2026 data shows, what moves behaviour, how to measure it, and a 90-day plan you can run without outside help.
What we mean by AI adoption
We use a plain definition. Adoption has happened when a person's way of working has changed and stayed changed. Opening a tool does not meet that bar, and neither does attending a workshop.
Deloitte's human capital team makes the same point in a line worth remembering: AI adoption is easy to prove and easy to fake. An employee can open a tool, complete the few interactions needed to count as an active user, and carry on with the old process. The usage figure rises and the work stays the same. Deloitte notes that some organisations now tie AI use to promotions, reviews, and pay, which gives people every reason to produce the number you asked for.
Where AI adoption stands in 2026
Gallup tracks this quarterly with a large US panel. In its Q2 2026 survey, fielded in May with more than 22,000 respondents, 47% of employees said their organisation has integrated AI tools, up from 41% the previous quarter. One in five did not know either way, and that share has stayed flat across recent quarters.
Individual use runs ahead of organisational use. Gallup found that 52% of US workers use AI in their role, 30% use it frequently (a few times a week or more), and 15% use it daily.
These are US figures, and your own numbers will differ. They still give the best large-sample picture available, and they show that roughly half the workforce has tried AI while fewer than a third have made it a habit.
Two gaps behind the headline numbers
Depth. Among AI users, the most common tasks are writing and editing (51%), search and research (49%), and general problem-solving (39%). Coding assistance and automation are each cited by 16%. The productivity ratings run the other way: 77% of people using AI for coding or automation say it improved their productivity, against 68% for writing and 65% for research.
Gallup also found that breadth of use tracks perceived value. Among employees using AI for one or two purposes, 45% report a positive productivity effect. That rises to 66% for three or four purposes, 78% for five or six, and 90% for seven or more. Gallup adds that this does not prove more uses cause more benefit, since people who see value probably go looking for more uses. I read it as a pointer. Someone who uses AI for one thing has adopted a tool. Someone who uses it for seven has probably changed part of their job.
Seniority. In Q2 2026, 51% of leaders used AI frequently, against 36% of managers and 26% of individual contributors. In 2023 those figures were 17%, 15%, and 9%, so everyone has moved, and leaders have moved furthest. The people who decide how work is organised use AI far more than the people whose work would change, which makes it easy for a leadership team to believe adoption is further along than it is.
What moves adoption: managers and workflow fit
Gallup names two conditions most strongly associated with frequent AI use: manager support and the integration of AI into existing workflows.
On managers, the gap is large. Among employees in AI-adopting organisations who strongly agree that their manager actively supports AI use, 78% use it frequently. Among everyone else, 44% do. Those same employees were 9.3 times as likely to strongly agree that AI has transformed how work gets done. Yet only 36% of employees in AI-integrating organisations strongly agree that their manager supports their team's use. Gallup describes effective support plainly: backing AI use at team level, using it visibly, answering questions, and showing how it connects to daily work.
These are associations, and Gallup words them that way, but they match our guide to AI training for employees: people copy what their manager does.
On workflows, Deloitte's figures are stark. Worker access to AI tools grew by 50% in a single year, according to its 2026 State of AI in the Enterprise report. Fewer than 60% of workers with access use AI in their daily workflow, and 84% of organisations have not redesigned jobs or workflows around AI. BCG's fourth annual AI at Work survey, published in June 2026, frames the same issue as AI changing jobs faster than companies redesign their operations.
Adding a tool to a process built for a world without it produces occasional use. Changing the process produces habits. The same cause sits behind pilots that stall: nobody owns the workflow around the tool.
How to measure AI adoption in a way that means something
Keep licences and logins on the dashboard as inputs. For the headline, use measures that track behaviour:
- Workflows with an AI step people follow. List the recurring workflows that matter most in each function, then count how many now include a documented AI step the team uses.
- Weekly use by team. Track the share of each team using AI at least weekly, reported by team and manager rather than by licence. Gallup's threshold of a few times a week is a sensible benchmark.
- Breadth of use. Count how many distinct tasks each person uses AI for. Gallup's data links this to perceived value.
- Outcome against a baseline. A baseline is the before measurement you compare against. For each changed workflow, record time, cost, and error rate before the change and again 90 days later. Without it you are relying on impressions.
- Review behaviour. Deloitte names judgment, knowing when to trust, challenge, or reject an output, as one of three behaviours that separate people who adapt from people who merely use the tool. Sample outputs and check whether people edit and question them. Usage logs cannot show this.
Where AI champions fit
AI champions are employees who help colleagues turn AI into routine practice. OpenAI's Academy guidance describes two kinds: leaders, who handle strategy, sponsorship, governance, and measurement, and activators, who sit inside teams, build workflows, share use cases, and report back on what works. It adds that strong champions start from where work gets stuck rather than from a feature, focus on workflows rather than prompts, and package what works so someone else can reuse it. I agree with all three.
One caution on sources. Most published guidance on champion programmes, including target ratios and time commitments, comes from vendors who sell programmes or seats, so treat any specific ratio as a starting guess.
Our recommendations are simple. Choose champions by the workflow they own, because someone who understands the work can see where AI fits and where it does not. Put their champion time in the calendar. Judge them by what they produce, which is a workflow someone else adopts. Pair each one with their manager, since manager support is among the strongest predictors of use in Gallup's data, and a champion working against an indifferent manager has an uphill job.
A 90-day starting plan
- Weeks one and two: choose three recurring workflows in one function, ideally where the work is repetitive and the owner is willing. Record the baseline for each.
- Name a manager sponsor for each workflow. The sponsor uses the tool in their own work and talks about it in team meetings.
- Name a champion for each workflow, with time set aside every week.
- Weeks three to 12: change the workflow itself. Remove the old step, update the templates and checklists, and change how output is reviewed, so the new way is the easiest way.
- Day 90: compare against the baseline. Keep what worked, fix what nearly worked, and stop what did not. Stopping is a legitimate outcome, and it leaves you with evidence rather than a licence nobody opens.
If you want a scored view of strategy, data, people, process, and tooling before you start, the AI Readiness Diagnostic is a free place to begin, and our readiness assessment guide covers the internal version.
Frequently asked questions
What is AI adoption?
AI adoption is the point at which people's way of working has changed and stayed changed because of AI. Licences issued, logins, and training attendance are inputs. They do not show adoption on their own.
What percentage of employees use AI at work?
In Gallup's May 2026 survey of US workers, 52% used AI in their role, 30% used it frequently, and 15% used it daily. 47% said their organisation has integrated AI tools. These are US figures.
How do you measure AI adoption?
Track workflows that now include an AI step, weekly use by team, breadth of use per person, outcomes against a before-and-after baseline, and whether people review and challenge outputs. Treat licences and logins as supporting inputs.
Why do employees not use the AI tools their company provides?
Gallup finds that manager support and workflow integration are the two conditions most associated with frequent use, and Deloitte reports that 84% of organisations have not redesigned work around AI. Other reasons will be specific to your teams, so ask the people involved.
How many AI champions does an organisation need?
No reliable benchmark exists, and most published ratios come from vendors. Start with one champion for each team that owns a target workflow, protect their time, and adjust after the first 90 days.
WORK WITH PRAXES
Ready to move from reading about AI to getting your organisation ready for it? Start with the diagnostic.
Take the AI Readiness Diagnostic