The problem with always adding more
AI is changing how we work. But one of the most telling shifts isn't about what AI can do - it's about how it behaves.
AI tools are built to be helpful, and that helpfulness has taken on a familiar shape. You ask for something, you get an answer… and then you're offered a way to improve it, extend it or take it further.
Would you like me to expand on how this affects your day-to-day life?
(See what we did there?)
Offered once, that nudge is useful. Repeated on every task, it quietly changes how the work unfolds.
Tasks stop ending where they used to
A piece of work that was once clearly defined now has a habit of expanding. You ask for a draft and get the draft - plus suggestions, alternatives, and an offer to refine it further.
Most of this is helpful, but the line between what you asked for and what could be added gets blurred. Every output arrives with another angle, another refinement, another version that might be better. The work doesn't get worse; it just stops having an obvious finish line.
And that carries a cost that rarely gets planned for. All those extra options still have to be read, considered and resolved. Then the loop widens. The expanded output is shared, a colleague reviews it, a client reacts to it - increasingly with AI's help on their side too. A second layer of AI summarises, reframes and suggests further improvements, and the work extends again, another step removed from the original task. Before long, you're watching a long rally of AI-assisted tennis.
Shall we look at why this is happening?
(We'll let that one go.)
Why it happens
This isn’t accidental.
AI tools are designed to keep the conversation going, because continued use is the business model. Every extra suggestion is an invitation to send another prompt. And more prompts mean more usage, which is precisely what these tools are built to encourage. From the tool's perspective, it makes perfect sense. From the point of view of getting work done, it shifts effort away from doing and towards reviewing.
That matters for agencies and clients alike. People spend more time deciding what to do next; clients receive outputs that feel larger and more complex than they expected. Even when the thinking is good, the volume can overwhelm the value.
A more useful way to work
None of this means the extra thinking is wasted; often it's genuinely valuable. The difference is in how it's managed, and that's a new skill we need to develop.
In practice, it comes down to a few simple habits:
- Define "done" before you start, not after. Decide what a finished, usable output looks like up front, so you have something to aim for.
- Deliver exactly that, clearly and confidently. Hand over what was asked for as a complete piece of work, not a menu of possibilities.
- Make exploration a separate, deliberate step. Treat further thinking as its own separate job, undertaken when it adds real value rather than as an automatic extension of every task.
That protects momentum while still leaving room for good ideas to surface.
And it's worth saying this isn't only about chatbots. The same principle runs through everything we do with AI and automation: knowing where a process should stop, where a human should step in, and how to constrain a model so it does the job in front of it rather than endlessly expanding it. Used well, AI should create more time for the work and the conversations that matter, not more to wade through. Knowing how to set those limits is as much a part of the craft as knowing how to prompt.
The takeaway
AI is genuinely helpful, and that helpfulness clearly can be hugely valuable. But when everything comes with additional options, work becomes harder to finish.
So one of the most useful skills in this AI age is also one of the simplest: knowing when to stop.
Just one more thing…
Yes, I used AI to help draft this. We have a project set up with Webreality's knowledge base - brand strategy, writing style, marketing plan and more - so AI works within clear guardrails, informed by the things that actually matter to our business. I provide the brief and an early draft, then shape the writing through focused dialogue. It's efficient, and the quality stays firmly under human control.
But when we reached the end of this piece, my AI assistant really did write:
If you want to push this further, we could add a real example from your team or clients to ground it even more.
Yes we could. But it would also mean a few hours chasing material, another round of redrafting, and a publishing date slipping by a week.
So thanks for the suggestion. But no thanks.
We know when to stop.
