47% Direct AI Instead of Doing the Work: The Tool-Stacking Trap
Published on 8/24/2026 · André Hellmann
Many job descriptions now include a task nobody advertised: supervising AI. 47 percent of workers say they spend more time managing and directing AI than doing the work itself (Source: BCG, AI at Work 2026). Skipping AI workflow redesign relocates work — it does not reduce it.
Discuss your next step in a free diagnosis call. Book a slot →
Contents
- What the BCG survey shows
- Strategy beats tooling five to one
- Workflow redesign instead of tool stacking
- Where does the recovered time go?
- Conclusion: the bottleneck is the workflow
- Frequently asked questions
- Sources
What the BCG survey shows
Boston Consulting Group published the fourth edition of its “AI at Work” study on 3 June 2026. It draws on 11,749 workers surveyed across 14 markets (Source: BCG, AI at Work 2026).
Three findings belong together. 74 percent of frontline employees now use AI regularly — up more than 20 percentage points in two years. 72 percent say AI has already considerably changed the skills expected in their role. And 47 percent spend more time directing AI than doing the work itself (Source: BCG, AI at Work 2026).
This is not an adoption gap. The tools are there and people use them. What is missing is the redesign of the work around them. That is exactly the People-Process Gap: people change their behavior faster than organizations change their processes.
The free Self-Check shows where a company stands in a few minutes.
Strategy beats tooling five to one
The most interesting number in the study is a ratio. Companies with a clear AI strategy see 25 percentage points more measurable business impact. Companies that only buy better tools land at roughly 5 percentage points (Source: BCG, AI at Work 2026).
A factor of five is not a rounding error. It describes two different approaches with very different returns.
The tool route is attractive because it is fast. Buy licenses, hand out access, schedule training. After that, every person decides individually where AI helps. That produces activity — and precisely the 47 percent who manage more than they work.
The strategy route starts with an uncomfortable question: which workflow should look different after the rollout than it did before? Skip that question and you stack tools on unchanged processes.
A tool on an unchanged workflow does not create progress. It creates one more task.
Workflow redesign instead of tool stacking
The study quantifies the difference from the other direction, too. Companies that redesign their workflows are 24 percentage points more likely to see measurable improvement. They are 22 percentage points more likely to save at least one workday per week and 20 percentage points more likely to report higher job satisfaction (Source: BCG, AI at Work 2026).
Redesign sounds large but is a craft task. Four steps are enough to begin:
- Map the workflow. Note every step from trigger to result — with duration and people involved. One page is enough.
- Delete steps. Which intermediate steps exist only because they used to be necessary? Deleting comes before automating.
- Define the handovers. What does the system do, what does the person do, and where exactly is the line? Blurred lines create duplicate work.
- Define the result. How does the team know a run went well? Without that definition, everyone applies their own standard.
Why the order matters — workflow first, tool second — is covered in the article on workflow-first.
One side effect matters: redesign does not work on a drawing board. The people who run the workflow every day know the exceptions. That is why we build with the team, not for the team — see Built with the Team.
Where does the recovered time go?
Time does get saved. 42 percent of regular frontline users save at least a full workday per week (Source: BCG, AI at Work 2026).
Then very little happens. 66 percent receive limited or no guidance on what to do with that time. More than half do not redirect it into strategic work (Source: BCG, AI at Work 2026). The hours spread across the day and disappear.
That is the most expensive part of the finding. Unused time gains show up in no metric at all. Which removes the evidence every AI budget needs in the next round — the connection is covered in the article on measuring AI ROI.
The fix is unglamorous: decide the use up front instead of hoping for it afterwards. More quotes per week. Shorter response times in service. One additional customer conversation per day. One decision per workflow is enough.
Which workflow to start with is something we map out in the free diagnosis call.
Conclusion: the bottleneck is the workflow
The BCG numbers disprove a comfortable assumption: that productivity arrives with the tool. It arrives with the redesign of the work the tool was meant to support.
The pressure is rising. 61 percent of respondents believe AI agents could take over at least half of their tasks within three years. At the same time, 52 percent still have only a limited understanding of what agents are (Source: BCG, AI at Work 2026). Anyone stacking tools on unchanged workflows today will stack agents on them tomorrow.
The way out is the same as before: workflow first, tool second, result measurable. This is not a method for large corporations. It is the minimum order for an AI operation that returns something.
Frequently asked questions
Why do 47 percent manage AI more than they work?
Because AI gets introduced on top of existing workflows instead of changing them (Source: BCG, AI at Work 2026). Prompting, checking and correcting are added to the existing work rather than replacing it.
What does workflow redesign mean in practice?
Map the workflow, delete redundant steps, define the handovers between people and system clearly, and define the result. Only then decide on tools.
Are better AI tools not enough?
No. Better tools alone lift measurable impact by roughly 5 percentage points, a clear strategy by 25 percentage points (Source: BCG, AI at Work 2026).
What happens to the recovered time?
In most companies, nothing defined. 66 percent of workers receive little guidance on it (Source: BCG, AI at Work 2026). Without a defined use, the gain evaporates.
How long does a first redesign take?
For a clearly scoped workflow, plan four to six weeks until the first measurement. Which workflow fits is something we clarify in the free diagnosis call.