October 30, 2026

Agency Beats Agents: What the Charter AI Pulse Means for Redesigning Work Around People

Charter's AI Pulse finds only 10.5% of AI users are thriving, and agency is the top predictor. What that means for redesigning work around people.
Agency beats agents: in the Charter AI Pulse, feeling involved in how AI fits your job ranks first of 26 conditions, while running agents in the background ranks 25th
Daan van Rossum
By
Daan van Rossum
Founder & CEO, Lead with AI

Presented by

The Charter AI Pulse surveyed 5,002 US desk workers, managers, and executives who use AI, and only 10.5% say AI is making both their work and their life much better. Charter calls them the thriving group.

The strongest predictor of who gets there is agency, or feeling involved in how AI fits into your job. Running agents in the background ranks 25th on Charter's list of conditions.

Charter AI Pulse chart: feeling involved in how AI fits their job ranks first (+45.9 points) among conditions linked to benefiting from AI, while running agents in the background ranks 25th (+14.6 points)
Chart: Charter AI Pulse, Fall 2026

This is the core of what we mean by people-centric AI transformation: transformation works when it starts with real work, real workflows, and real leadership behavior. Charter's data now puts numbers behind all three.

What Does Having a Say in AI Actually Look Like?

Agency is unevenly spread. Managers are 1.8 times more likely than workers to say they have it (67% vs. 37%), and 36% of workers say they are mostly reacting to AI being imposed on their jobs. One systems analyst in Charter's interviews summed it up: "It's a tool. It's not a mandate."

Frequency matters too. People who use AI multiple times a day are twice as likely to benefit as people who use it a few times a week. Agency still ranks above it.

So how do you hand people a real say? You let them redesign their own recurring work, at a level of detail where their judgment counts.

That is what our OVER framework is built for. Nobody can redesign a job title, so OVER starts with the Outputs a person actually produces, breaks each one into the Verbs behind it, Evaluates each verb as AI-first, human-first, or hybrid, and Routes it to the right kind of tool. The person doing the work is the best-placed person to run every one of those steps.

It also builds the daily habit Charter keeps pointing to. When AI is part of a workflow that runs every week, the AI step happens every time the work does.

My colleague Phil Kirschner, a Transformation Partner at Lead with AI, added a useful layer in his recent piece, Get Out of Your Swimlane. Most process maps start from the org chart, with one horizontal lane per team, and AI now moves work between lanes faster than anyone can redraw them.

Phil's answer is to map the work first and attach the "who" last, as a label you fully expect to change when an agent or automation takes over a step. As he puts it, "Swimlanes draw lines around people and organizations, but better maps draw lines between steps."

That shift matters for agency. A redesign that starts from the work invites the people who do it into the room. A redesign that starts from the org chart tends to open with the question everyone is quietly worried about, which is whose lane is getting smaller.

Flagship AI Newsletter
The AI Newsletter That Makes You Smarter, Not Busier
Join over 30,000 leaders and receive our insights on AI platforms, implementations, and organizational change management.
FlexOS Course - AI Content Accelerator - Testimonial Badge

Why Do Teams Get Policies When They Ask for Guidance?

Charter found that 11 of the top 14 factors tied to benefiting from AI are conditions managers control, from time to learn to recognition for using AI well. It also found a demonstration gap of 23 to 30 percentage points between what managers say they do and what their teams see.

One of the widest gaps sits on the behavior teams ask for most. 54% of managers say they give specific guidance on when and how to use AI, but only 24.6% of workers see it.

Charter AI Pulse chart: managers report six AI leadership behaviors 23 to 30 points more often than workers see them, with the widest gaps on sharing what they learn and giving specific guidance
Chart: Charter AI Pulse, Fall 2026

Organizational support shows the same mismatch. Guidance on AI policy is the support most often provided (45.4%) and the least likely to be named most useful (13.4%).

Hands-on training on specific tools is named most useful (32.5%), yet only 39.6% of AI users receive it. Dedicated time to experiment does not fill the gap either, with just 13.9% naming it most useful.

Charter AI Pulse chart: guidance on AI policy is the most provided support (45.4%) but least often named most useful (13.4%), while hands-on training on specific tools is named most useful (32.5%)
Chart: Charter AI Pulse, Fall 2026

Why is specific guidance so rare? It is very hard to give specific guidance on a job description. A manager who tells the team to use AI more has offered encouragement.

Specific guidance sounds more like this: AI drafts the variance commentary on the weekly forecast, and you own the recommendation. That kind of guidance is only possible once you understand the work at the level of outputs and steps.

We call that skill workflow literacy, and it is why AI training cannot be a one-time event. The work and the tools keep changing underneath everyone.

Charter's best examples came from coworkers meeting weekly to show each other real work, with leaders demonstrating their own workflows. A team that walks through one redesigned workflow together each week gets hands-on training, specific guidance, and visible manager use in the same hour.

Where Should the Human Check Happen?

Charter's most uncomfortable finding sits at the top of the org chart. Executives are twice as likely as workers to trust AI more than their own judgment (32.8% vs. 16.3%), and about one in five executives regularly uses AI outputs they have doubts about, compared with 6% of workers.

Charter AI Pulse chart: managers and executives are about twice as likely as workers to trust AI outputs, and to trust AI more than their own judgment or a knowledgeable colleague
Chart: Charter AI Pulse, Fall 2026

That is despite executives reporting the most checking, with 81% saying they always or usually fact-check AI output. The people whose decisions carry the most weight are also the most exposed to letting AI do the thinking.

The reasons are practical. Time pressure leads (43.7%), followed by not knowing how to verify (38.7%). Executives are the most likely to decide the stakes are simply low (42%).

Charter recommends teaching people how to verify and defining which decisions count as high stakes. That is the job of the DRAFT check we published earlier this month: Doubt the confidence, Reasoning, Alternatives, Facts and sources, and Take ownership.

It is scaled to the stakes, so a quick internal note gets a fast pass while board material gets all five steps.

Verification also needs a fixed place in the workflow. In OVER's Evaluate step, anything that relies on tacit knowledge, political judgment, or real accountability is human-first. Phil's version is to flag every decision point, handoff, and approval on the map, because those are the places a person should stay in the loop whatever the agents end up doing.

This is a burnout question as much as a quality one. Checking and fixing AI output is the most-cited source of AI burnout in Charter's data (20.7%). Designing the checks in and budgeting time for them keeps accountability with people without turning everyone into a full-time proofreader.

Why Would Anyone Hide Something That Helps Them?

35.9% of AI users have hidden their AI use from a manager, colleague, or client, and managers hide it most (43.7%). In Charter's analysis, hiding is the top factor separating people whose burnout rose with AI from those whose burnout fell.

Charter AI Pulse chart: having hidden AI use (+28.5 points) is the factor most common among AI users whose burnout increased, followed by often hitting usage or cost limits
Chart: Charter AI Pulse, Fall 2026

The leading reason is fear of looking less capable (49.3%). Underneath it sits real job anxiety: 76% of workers say it is not at all or only slightly clear how their jobs will evolve because of AI, compared with 34% of executives.

Ethan Mollick calls these people Secret Cyborgs, and as I wrote in our people-centric guide, hiding is a rational response when people suspect their efficiency will be used against them. Charter's data on saved time suggests the suspicion is reasonable.

Only 7.9% say the time AI saves goes to rest or time outside work, while 17.6% say it goes to doing more of the same.

That is why we ask leaders to decide what reclaimed time is for before it arrives. It is also another argument for redesigning workflows in the open, because a mapped workflow makes AI use visible by default. The AI steps are written on the map where everyone can see them.

The Bottom Line

  • Give people a say by having them redesign their own recurring work. Agency is the top predictor of benefit in Charter's data, ahead of frequency and far ahead of advanced use.
  • Map the work before you assign the who. Steps will keep moving between people, agents, and automations, so treat ownership as a label you expect to change.
  • Turn manager support into workflow-level guidance. Specific guidance is what teams ask for most and see least.
  • Build verification into the workflow. Mark the human checkpoints, scale checks to the stakes, and budget time for them.
  • Make AI use safe to show. Hidden use is the strongest burnout signal in Charter's data.

Charter's appendix includes ten questions managers can run with their own teams. Answer them yourself first, then ask your team, and the distance between the two sets of answers is your demonstration gap.