Two failures, usually together. The workflow hasn’t changed, so the AI makes the old pitch faster. And the AI is generic: it knows neither the firm nor how to persuade this client, so it writes in a default voice that any firm with the same tool shares. Fix the workflow first, then the AI.
Why doesn’t faster mean better?
Speed changes the cost of a pitch, not what it says. If the old pitch started from the firm’s credentials while the client was thinking about something else, the AI version does the same in less time. The hours saved are real, and they show up on a timesheet. The win rate moves only if someone spends them on work that wins.
Searchlight Group, a recruitment firm, pitches to government and not-for-profit evaluation committees, which score proposals the way a law or accounting firm’s panel review does. It rebuilt how it pitched first, then built the AI it pitches with. Proposals went from six hours to two, and the win rate from 35% before the work to 85% in 2025. The time saving was the smaller result. Our guide on showing the return on AI in BD sets out how to count the difference.
Hasn’t this happened before?
Yes, with computers. In 1990 Michael Hammer argued in the Harvard Business Review that companies had invested heavily in information technology and got disappointing results, because they used the machines to speed up existing processes instead of redesigning them. He called it “paving the cow paths”.
In BD the old path is the pitch built from the last similar pitch and padded with credentials. Put AI on that path and you get the same pitch sooner. So redesign the path before you pave it: decide what a pitch for this client should start from and what it has to prove, then give that job to the AI.
Why does generic AI make BD writing sound alike?
It writes in its model’s default voice unless someone tells it otherwise. OpenAI’s developer guide to its top model, GPT-6 Astra, warns that it “may use recurring phrases across sessions” and tells developers to “specify the writing style and structure your application needs”. A firm that doesn’t specify gets the model’s house style.
Firms add a habit of their own. Taleist used AI to code every sentence on the homepages of 55 law firms from the Australian Financial Review’s partnership survey, captured on 25 July 2026, then checked the counts. Of the 738 sentences that put either the firm or the client in the frame, 694 were about the firm: 94%. A homepage isn’t a pitch, but it is the easiest public sample of how a firm writes about itself. Load copy like that into an AI as the firm’s knowledge and it has 694 examples of the firm talking about itself to work from, and 44 of the firm talking to the client.
The fix is to give the AI two things a default model doesn’t have. One is the firm’s knowledge. The other is a method: what this client values most, drawn from the tender, the client’s own words and the partners who know them, and how a pitch for them should run.
Which failure is costing your firm?
Look at your last lost pitch. If it was built from an earlier pitch and opened on the firm, the workflow is the problem. If it started from the client but read like any firm’s, the AI is. If both, fix the workflow first, because a better-written pitch that starts from the firm still loses to one that starts from the client. The service page sets out all four combinations, and our guide to whether the next AI model will fix your pitches explains why a newer model fixes neither.
What turns AI into BD results?
Fixing both, in order. Rebuild the workflow so the pitch starts from the client. Build the firm’s knowledge and a method for persuasion into the AI. Then measure results against a baseline set before the change: pitches won, fees, replies to client emails.
Taleist does this for professional services firms, and calls it consulting-grade AI: an advisor built on the firm’s knowledge and the craft of persuasion, running inside the AI the firm has already approved. For what the change looks like on the page, see an executive summary, before and after. For the other kinds of provider, see who builds AI for business development.
Sources: Michael Hammer, “Reengineering Work: Don’t Automate, Obliterate”, Harvard Business Review, July–August 1990, pp. 104–112. OpenAI, “Using GPT-6”, OpenAI API documentation, developers.openai.com/api/docs/guides/latest-model, read 10 October 2026. Taleist homepage study: 55 firms from the AFR’s partnership survey, homepages captured 25 July 2026. Searchlight Group results as published on taleist.agency.