The first time I met the sales director, he spent exactly one minute explaining to me why this project was a waste of time. "We sell. We don't have time for AI workshops. My people are out with clients — they're not sitting around playing with chatbots." And he didn't say it unkindly. He meant it sincerely.
And he wasn't wrong. His people were busy. Sales cycles were long. Competition was fierce. Time for a course was time not spent selling. That's a rational concern, and it deserves a rational response.
The problem with most AI implementations in sales teams is that they start with the technology and try to convince people it's relevant. We started from the opposite direction: with what salespeople actually hated about their work.
We asked the sales director for one thing: access to three of his salespeople for 30 minutes, one at a time. No presentation. No demo. Just conversation. He agreed with some scepticism.
We asked the three salespeople the same question: What is the part of your work you spend most time on and hate most? The answer was consistent across all three. Meeting preparation and proposal writing. Meeting preparation because it required research on the client's company, industry and competitive landscape — and because it took two to three hours per client meeting. Proposal writing because each proposal was a template-based exercise that still ended up requiring two to three hours of customisation.
That was what we worked with. Not AI in general. Not "the future of work." Precisely these two processes, for precisely these three salespeople.
We ran a two-hour session — we deliberately didn't call it a course. We called it a demonstration. We took a real client situation from one of the salespeople — an upcoming meeting, as it happened — and showed, step by step, how AI could halve the preparation time. We searched for public information about the client's company and used AI to summarise it into a briefing. We identified the likely pain points based on industry and company size. We generated five opening questions designed to surface the real challenges.
Then we took an existing proposal and showed how AI could generate a structured draft in six minutes, which the salesperson then customised in a further 20. Total time: 26 minutes instead of two to three hours.
We were very conscious about what we didn't say. We didn't say AI writes their proposals for them. We didn't say AI replaces their knowledge and experience. We said the opposite: AI gives you back the hours you currently spend on templates and research — so you can use them for what you're best at. What clients choose you for is you, not the documents.
The three salespeople tried it the next day. Not because we asked them to — but because they were curious. One called me two days later to say he had had his best-prepared client meeting in three years. He knew the client's situation better than the client themselves, and the client noticed.
He shared it with his colleagues. Not in a meeting, not in a management email — but over lunch. "This is actually brilliant." Colleagues asked if they could try it. We ran an informal session for four of them. Then ten. Then all 22.
Six weeks after the first session, the sales director himself asked for structured onboarding for the whole team. The person who wouldn't allocate time for a course now allocated an entire afternoon. "I can see it in the numbers," he told me. "Preparation time is halved. We're sending more proposals. And the proposals are better — clients say so."
Tripling in AI adoption. Meeting preparation time halved. Proposals written 60% faster. But the number that matters most to him is none of those. It's that two salespeople who hadn't hit their quarterly targets in the previous two quarters are both above target. He doesn't attribute that directly to AI. But he says they have more time to actually sell — and it shows.
The most important lesson from this project is one that transfers to any kind of change initiative: don't start with the technology. Start with the problem people actually have. And show that you're solving it. The rest happens by itself.