In the past two weeks, three large studies reported the same pattern. Leaders are using AI agents far more than the people who work for them. In Asana’s survey of 4,502 people in the US and UK, 68% of CIOs and AI leaders use three or more agents every week, compared with 36% of office workers who use AI. Workday’s September survey of nearly 6,000 workers found that 44% of senior leaders are heavy AI users, against 9% of individual contributors. Gallup puts frequent AI use at 51% for leaders and 26% for individual contributors, and that gap has grown since 2023.
CIO Dive called this the agentic gap. I think the name is useful, because this gap decides whether the money companies are putting into agents comes back. Agents create value when the people doing the daily work use them, keep using them and trust them with real tasks. Right now most of that use sits at the top of the organization.
The usual explanations are training, trust and unclear policy. All three play a role. In our work at The Octalysis Group, I keep seeing a more basic cause. Leaders and workers are having two very different experiences with the same agents. Only the leaders’ experience has been designed to motivate. Below I look at the latest research, at both experiences through the Octalysis Framework, and at what companies can change.
What the October 2026 research shows
The Asana study, run by Censuswide in September 2026, is the most detailed of the recent reports. Only 23% of workers say their agents address the tasks they most need help with. A third say agents leave significant gaps, and about one in five say agents mostly handle lower-priority work. Among workers who use two or more agents, 46% say that coordinating those agents takes more than 10% of their working time, and 38% say more than a quarter of the work their agents produce is duplicated. Leaders are also more relaxed about autonomy: 81% are comfortable with agents acting without step-by-step review, compared with 65% of workers.
Workday’s October update adds a knowledge gap. 88% of senior leaders have used agentic AI to complete tasks, compared with 54% of individual contributors. Only 38% of all employees say they understand the term “AI agent” well enough to explain it with confidence. One in three employees say they lose at least half of the time AI saves them to reworking its output, and 41% of AI users say they have deliberately hidden their AI use at work.
The Wharton and GBK Collective 2026 AI Adoption Report, published on 13 October, shows where this leaves companies. Two-thirds of departments say they use AI agents, but only 17% let an agent complete a task without a person approving each step. GBK’s Brian Smith describes this as the “supervised middle”, and the report finds that returns there are the weakest.
Two people, two experiences of the same agent
Yu-kai Chou built the Octalysis Framework around eight Core Drives that explain why people do what they do. When I map the agent experience of a typical leader and a typical worker on those drives, they look like two different products.
The leader’s experience
A leader decides which agents the company buys, what they should do and which work gets handed to them. That is Core Drive 3, Empowerment of Creativity & Feedback: real choices and fast feedback on those choices. The agents and their setup feel like the leader’s own, which is Core Drive 4, Ownership & Possession. Results show up in dashboards and board updates as hours saved and processes automated, which feeds Core Drive 2, Development & Accomplishment. And leading the company’s AI transformation carries a sense of purpose, Core Drive 1, Epic Meaning & Calling.
These are all White Hat drives. They make people feel in control and fulfilled, and people come back to experiences built on them by choice. It is no surprise that, in Gallup’s data, 21% of leaders describe the productivity impact of AI as extremely positive, against 13% of individual contributors.
The worker’s experience
The worker usually gets agents that someone else selected, configured for processes someone else defined. As the Asana numbers show, most of those agents do not touch the work the worker most wants help with. More of the day goes to checking, correcting and coordinating machine output. In the BCG AI at Work 2026 survey of close to 12,000 people, close to half or more of respondents report spending more time reviewing and correcting AI output or managing and directing AI, and 41% report increased mental strain. Research by BetterUp and the Stanford Social Media Lab, published in Harvard Business Review, describes the same burden as “workslop”. In their survey of 1,150 US desk workers, 40% had received it in the previous month, and each case took about two hours to resolve.
On top of this comes uncertainty. Workers often do not know what the agent will do next, what the rules for using it are, or what the agent means for their job. Only 35% of employees in the Workday survey say their employer’s expectations for disclosing AI use are very or extremely clear.
In Octalysis terms, the worker’s experience runs mostly on Black Hat drives. Core Drive 8, Loss & Avoidance, shows up as fear of being blamed for an agent’s mistake and fear for one’s role. Core Drive 7, Unpredictability & Curiosity, shows up in its anxious form: not knowing what the agent will produce or how the company will judge its use. The White Hat drives that the leader enjoys are mostly missing. The choices were made by others, so there is little Empowerment. The agent belongs to IT, so there is little Ownership. And review work is invisible. Nobody sees the error a worker caught before it reached a client.
Black Hat drives can push people to act for a while. They also make people feel they are not in control, and people leave such an experience as soon as the pressure drops.
The supervised middle is the least motivating seat
The Wharton finding deserves extra attention. When a person has to approve every step an agent takes, they are reviewing work they did not shape, at a pace they do not control, with no visible result for their effort. From a motivation design point of view, this is close to the worst job one can design.
The report also points to a way out. Departments that monitor their agents automatically have transformed their workflows at a rate of 24%, compared with 9% for departments that do not. Gartner expects more than 40% of agentic AI projects to be canceled by the end of 2027. In my view, many of those projects will fail because the people around the agents were given a role nobody wants to stay in.
We described the same pattern in agent pilots in why AI agent adoption fails after the pilot. Every month that workers spend in a poorly designed agent experience adds to what I call motivational debt, and that debt makes the next rollout harder.
Give workers the leader’s experience
The practical answer is to give the people doing the work the same motivational experience that leaders already have. Here is how we approach that.
1. Let people choose and shape their agents
Give teams a say in which agents they use and how those agents are set up. This activates Core Drive 3 and Core Drive 4. As we described in AI adoption is a motivation design challenge, BBVA gave its first licenses to motivated employees rather than senior ones, let low users lose their license to eager colleagues, and made it clear that employees own the output. Managers matter here too. Gallup found that employees who strongly agree their manager supports AI use are far more likely to use AI frequently, 78% compared with 44%.
2. Start with the work people want help with
If only 23% of workers say agents address their most important needs, the first step is to ask them. Each team knows which tasks drain its time. An agent built around one of those tasks gives immediate and visible relief.
3. Turn review into progression
Design the human role so that it grows. When a person’s review record shows that an agent is reliable for a type of task, let them move that task out of step-by-step approval and into monitoring. Show people the errors they caught and the hours they recovered. This is Core Drive 2, and it turns the supervised middle into a path toward more responsibility.
4. Make good agent work visible to peers
When someone builds an agent setup that works, let others copy it and credit the person who built it. Core Drive 5, Social Influence & Relatedness, spreads good practice much faster than a training program does. It also gives workers a reason to talk openly about their AI use instead of hiding it.
5. Lower the Black Hat pressure
Be clear about the rules for using and disclosing AI, about who is accountable when an agent makes a mistake, and about what happens to the time agents free up. People who believe the saved time will be used to cut their role will protect themselves. We went into this further in meaningful work when AI can do the task.
What this looks like in practice
The closest parallel in our own work is Navo Orbico, which operates across 19 countries. Its sales teams were tired of cash incentives, and routine work in the CRM was being neglected. We did not replace their systems. We built an engagement experience on top of the CRM they already had. Sales reps became captains who chose their own growth strategy, which is Core Drive 3, and a shared virtual “Port City” tied each person’s effort to the team, which is Core Drive 5. The rollout was non-mandatory and reached a 99.5% adoption rate, with a 28.6% increase in sales revenue. The full details are in the Navo Orbico case study.
The same principle shows up across our case studies. At CAIXA, our work led to a 49% increase in net income in six months, propelling CAIXA to become Brazil’s #1 public bank. At Astana Hub, the result was a 504% increase in active users.
How to measure your own agentic gap
Most companies measure agent licenses and log-ins. These numbers say little about whether work gets done. A more useful set of measures:
- Agent usage split by level, from executives to individual contributors, so the gap becomes visible.
- Continued use after 30 and 90 days, which shows whether people come back by choice.
- The share of each team that says agents help with the work they most need help with.
- Time spent reviewing, reworking and coordinating agent output.
Asana’s Arnab Bose put it well in CIO Dive: companies should “move away from measuring agent activity and start measuring whether work actually gets finished.”
Conclusion
Leaders have a motivating experience with AI agents because they have choice, ownership and visible results. Most workers have the opposite: agents chosen for them, extra review work and a lot of uncertainty.
Companies that want a return on agents need to design the worker’s experience with the same care they put into choosing the technology. The Octalysis Framework gives a practical way to do that, by building White Hat drives into the daily work and reducing the Black Hat pressure that makes people stop using the tools.
If your agent usage data shows the same split between leaders and teams, bring the numbers. contact us and we will look together at where the worker experience breaks down.
