Most ideas in a small business get shared too early or not at all. Too early, and the first meeting is spent on holes you could have spotted yourself. Not at all, and a decent idea drifts because nobody had time to think it through.
AI has changed that for me. It's become my sounding board: not the source of the ideas, but the place where they get into shape before anyone else sees them.
A day instead of weeks
I had a practical problem. How could we check that people understood the questions in the business assessments we run, without manually interviewing everyone who completed one? I was weighing up whether to put the AI Adoption Index online as a two-month beta.
I spent a day working through it with several AI agents, each on a different part of the problem. I set the direction, pushed back on what didn't fit and decided what to keep. We based the design on published research on web probing, an established method for testing questions online by asking people how they understood them.
By the end of the day the idea had changed shape. I'd started with a plan to test one assessment. I finished with a design for a repeatable research quality system we can use on every future assessment, intended to stand up to external scrutiny.
The speed wasn't what surprised me. It was how far the idea moved.
What a good sounding board does
It exposes weak ideas early. When you ask AI to find the weaknesses, some ideas need changing and some aren't worth pursuing. Finding that out in an hour is far cheaper than finding out six months and a few thousand pounds in.
It helps fix weaknesses, not just find them. You spot a problem, ask for three ways round it, pick one and test that. The idea gets stronger in the same sitting rather than waiting for the next meeting.
It means what reaches your team is developed. You walk in with the obvious objections already answered, so the conversation starts further along. Better ideas, earlier, with less wasted time.
Make it argue with you
There is one catch. Left to itself, AI tends to agree with you. Ask "is this a good idea?" and you'll mostly get encouragement. Refine an idea with a sounding board that agrees with everything and you end up with a polished bad idea, which is harder to see through than a rough one.
You have to ask for the argument. The prompts I find most useful are:
- "Argue against this as strongly as you can."
- "You're my most sceptical customer. What would stop you buying?"
- "It's 18 months from now and this has failed. What went wrong?"
An objection from AI needs checking just as much as its encouragement does.
Keep your people in it
AI is good at testing the logic. It can't test what only your people know: the customer with odd habits, the supplier who lets you down, the member of staff who'll quietly resist. Those still need a human conversation.
So take ideas to your team further along, not finished. If they only see the plan once every decision has been made, you're asking for approval, not input.
Learn it hands on
We teach this technique in our Confident & Safe AI course. In the afternoon (or, if virtual, the second session) there is a hands-on round where everyone brings a real idea from their own work, puts it through the sounding board, then tests the result on a colleague. You leave with a better idea and a method you can use on the next one.
Written with Claude, then steered, cut and checked by me. I answer for it.
Want to put a real idea from your own business through the sounding board? That is one of the practice rounds on the Confident & Safe AI day.