Summary: AI hasn’t cut anyone’s workload, it’s just made people produce more, faster. The problem is nobody’s changed what happens after: work still gets stuck in the same approvals, the same meetings, the same manager’s inbox. Unless organisations rethink those processes, all that saved time is lost.
The conversation around AI at work is still largely focused on tools:
- Which platform should we buy?
- How quickly can we roll it out?
- How do we encourage more people to use it?
- And how do we build enough AI capability across the organisation?
These are important questions but, following a number of leadership activation events I’ve been running with Occupational Psychologist, Dexter Winters, I’ve been doing a lot of thinking about the challenge coming next.
We have witnessed time after time that once people start using AI successfully, something rather uncomfortable happens: they produce more work.
More: reports, presentations, analysis, ideas, content, recommendations.
All created at a speed that would have been impossible only a few years ago. On the surface, this looks like productivity but producing more does not necessarily mean becoming more productive.
In fact, many people I speak to are telling me that they feel busier, more overwhelmed and under greater pressure than before.
AI has increased the flow of work, but most organisations have not changed the system around it.
The fire hydrant and the O-ring
This is where the plumbing analogy comes in (my thanks goes to Dexter for the inspiration on this one!).
AI has turned on a fire hydrant of work, but we are still trying to push everything through the same old pipework.
Somewhere in that system is a tiny O-ring which is usually a manager, senior leader or leadership team who is expected to review, challenge, approve or sign off everything now arriving at twice the speed.
The employee might have used AI to reduce three hours of work to 45 minutes, but the output still enters the same approval process. It waits for the same manager, gets discussed in the same meeting and eventually joins the same decision-making queue.
The creation of the work has accelerated. The organisation has not. Eventually, something must give.
The manager becomes overwhelmed, standards begin to slip or people stop reviewing work as carefully as they should. Alternatively, the organisation creates an enormous volume of perfectly acceptable work that nobody particularly needs.
This is why we need to stop treating AI adoption as purely a technology or skills challenge. It is rapidly becoming an organisational design challenge.
Many people I speak to are telling me that they feel busier, more overwhelmed and under greater pressure than before
More output is not the same as more value
Individual productivity and organisational productivity are not the same thing.
If I save an hour writing a report, that is an individual productivity gain. But if the report still spends four days waiting to be approved, requires three meetings and is then rewritten because nobody agreed its purpose in the first place, the organisation has gained very little.
This is where a great deal of anticipated AI return on investment could quietly disappear.
Organisations may be measuring how many employees are using an AI tool; how many licences have been activated or how much time people say they have saved. Those measures can be useful, but they do not tell us whether work is moving through the organisation more effectively or creating better outcomes.
The more valuable questions are:
- What happened to the time saved?
- Did decisions improve and, if so, why and how? If not, why not?
- Did customers receive a better service? How do we know?
- Did teams reduce low value activity? How do we know?
- Were people able to focus on work requiring judgement, creativity or human connection? If not, what got in the way?
If we cannot answer those questions, we may simply be using AI to produce more widgets.
We need new norms for AI-enabled work
There is another issue that many organisations have not yet addressed: people are being encouraged to use AI without a shared understanding of what good AI-enabled work looks like.
More questions to ask which can help this journey include;
- When should AI be used, and when should it not?
- What level of human review is expected?
- Who owns the final decision?
- Does every AI assisted document need approval?
- When can employees act autonomously?
- What evidence should accompany a recommendation?
- How do people declare where AI has contributed?
Note: These questions cannot be answered by an AI policy alone. Policies set boundaries, but people also need practical working norms that reflect what really happens inside their roles and teams.
Without those norms, managers become the safety net for everything. Employees are told to experiment but still feel the need to seek reassurance before acting. Leaders want innovation while maintaining the same controls, hierarchies and approval processes that existed before AI. That is not transformation. It is old work moving faster.
Individual productivity and organisational productivity are not the same thing
L&D has a much bigger role to play
This is where learning and development needs to broaden its contribution.
AI training cannot stop at teaching people how to use a tool or write a better prompt.
Employees need to understand how to judge AI outputs, where human expertise adds value and how to collaborate effectively with AI. Managers need support to redesign work, set clearer expectations and lead teams whose capacity may change significantly.
This also means helping teams examine their work honestly with questions like:
- Which activities still need to exist?
- Which approvals genuinely reduce risk?
- Which meetings are now redundant?
- Where are decisions getting stuck?
- What should people do with the time AI releases?
If we build capability without addressing these questions, we risk creating faster employees inside slow organisations.
The next phase of AI adoption
The next phase of AI adoption will be less about access to technology and more about work cadence, decision making and organisational design.
It will require leaders to explain why AI is part of the organisation’s future, what that means for employees and, equally importantly, what it does not mean.
It will require managers who feel confident enough to lead differently rather than simply coping with more. And it will require organisations to decide where the AI Dividend
(the time, energy and capacity released by AI) should go.
The fire hydrant has already been switched on. The question now is whether organisations will redesign the pipework or wait for the pressure to expose its weakest points.
The next phase of AI adoption will be less about access to technology and more about work cadence, decision making and organisational design
Actionable insights
- Ask what happened to the hours AI saved last month. If nobody can answer, they probably didn’t go anywhere useful.
- A three-hour task done in 45 minutes still hits the same four-day sign-off. Fix the sign-off, not just the task.
- Get specific about AI norms: when it’s fine to use, when it isn’t, who has final say. Vague policies won’t hold up in practice.
- Some approvals exist purely because things used to move slower. Check which ones still earn their place.
- Push AI training past prompting: The real skill gap now is knowing when to trust what AI gives you and when to push back, not how to write a better prompt.
Note: This article was created by Erica Farmer in collaboration with ChatGPT 5.6 on 22/07/2026, which supported the development and refinement of her original ideas.
Fire Hydrant and the O ring metaphor: credit to Dexter Winters and used with permission.
Erica Farmer’s new book AI for People Professionals: Understand How to Use Artificial Intelligence in Your HR Role is available now.
The post Has AI turned work into a fire hydrant?: More output, not more value appeared first on HRZone.
