AI tools get sold on the promise of time saved. For a lot of people, the first few weeks feel like the opposite. They type a request, get back something generic or slightly wrong, try to fix it with a follow-up, get something different-but-still-wrong, and eventually give up and just write it themselves — twenty minutes gone, no output to show for it.
This isn't a sign the tools don't work. It's what happens when a genuinely powerful tool is used with a vague instruction. Ask a new employee to "sort out the report" with no further detail and you'd expect some back-and-forth too. AI tools aren't different in this respect — they just don't push back and ask clarifying questions the way a person would, so the gap between what you meant and what you asked for shows up in the output instead, silently, without anyone flagging that something got lost along the way.
Why the output sounds generic
Ask an AI tool for a client email, a policy draft, or a marketing blurb with no other context, and it will reach for the most statistically average version of that thing — because that's all it has to go on. It doesn't know your tone, your standard caveats, or the three things your clients always ask about. Generic input, generic output. This is one of the reasons a business's own knowledge and standards matter so much once AI is in the mix; without them, every answer starts from zero.
It's not really about the tool you use
This applies whether your team is on Claude, Copilot, ChatGPT or something else entirely. The specific interface changes; the underlying skill doesn't. Someone who prompts well gets good results from more or less any capable AI tool. Someone prompting vaguely will get mediocre, generic results from all of them, and is liable to conclude the tool itself isn't very good — when the actual gap is in the instruction, not the model behind it.
Share what already works
Individually, people tend to figure out a handful of requests that work well for their own role and quietly keep using them. The time saving multiplies when those get shared rather than rediscovered independently by everyone on the team — the request that reliably produces a good first draft of a client update, the one that summarises a meeting the way your business actually likes meetings summarised. Writing them down where colleagues can see them turns one person's trial and error into everyone's starting point.
The cost, multiplied
None of this looks expensive in a single instance — a few extra minutes here, a rewrite there. It adds up differently across a team. Ten people losing twenty minutes a day to back-and-forth prompting isn't a rounding error by the end of a month; it's a meaningful chunk of paid time spent getting AI to do what a clearer instruction would have gotten in one pass.
What actually fixes it
Not a single clever trick, despite what a lot of "10 prompts that changed my life" content implies. It's a habit, built on a few consistent things:
- Say what you actually want, including the format, the audience and the length
- Give it the context a new employee would need, not just the task
- Iterate on the same thread instead of starting a fresh, blank request each time
- Reuse prompts that already worked, rather than reinventing them from scratch
The second point is where a lot of the real time saving comes from, and it gets much easier once your standards, processes and tone are written down somewhere an AI tool can read them. Then the context doesn't have to be retyped from memory every time, and the starting point is already three-quarters of the way to right.
A quick before and after
"Write me a client email about the delay" is the kind of prompt that produces something you'll spend ten minutes rewriting. "Write a short, apologetic email to a trades client explaining a two-week delay on their kitchen install due to a supplier issue, offering a firm new date and a small goodwill gesture, in a warm but professional tone" produces something you might send with one tweak. Same tool, same amount of typing effort either way — the difference is entirely in what you gave it to work with.
Where to build the skill properly
Our AI Discovery workshop covers exactly this ground — what AI is and isn't, using it responsibly, and writing a prompt that actually works — for anyone who's new to working with these tools or has been getting by on trial and error. It's the foundation everything else builds on.