The return on AI in business development is your whole team working closer to the level of its best people. It shows in results, not hours: proposals won and the fees behind them, pitches you were right to skip, replies to client emails and newsletters, and RSVPs that turn into people in the room. Measure each against a baseline set before the AI went in.
Why hours saved doesn’t answer the partners
Hours saved is the figure AI tools lead with, because a tool can see it. The tool knows how long a draft used to take and how long it takes now. It doesn’t know whether the client said yes.
A partner hears the figure and asks what the team did with the time. If the answer is “wrote the same proposals faster”, the firm has spent less on work that still loses. That’s a saving. It isn’t revenue.
Hours turn into a return only when someone spends them winning work: more bids at the same win rate, more client meetings, a newsletter good enough that clients answer it. So count what the hours bought, not the hours.
Where saved hours go
The return is the whole team working near the level of its best
In a BD team, some people write proposals that win and emails that get answered. Others write drafts that need rewriting before they go out. The result to look for is that gap closing: the weaker drafts start reading like the strong ones, and the strong people stop spending their week fixing them.
The best outside evidence points the same way. In a randomised field experiment at Procter & Gamble, 776 professionals worked up ideas for new products and campaigns, and evaluators scored them blind (Dell’Acqua and colleagues, Harvard Business School working paper, 2025). Some were working outside their usual job. Alone with AI, they scored about as well as specialist teams working without it. The study didn’t check whether AI narrows the gap inside a team.
That’s what to measure in your own team. Take a result you already track, such as win rate by bid lead or reply rate by sender, and compare your strongest people with the rest, before the AI and after. With a small team the numbers per person are thin, so read the trend over months, not a single quarter.
The gap closing inside a team
What to measure, by kind of work
Pitches and proposals. Win rate, the fees won, and the shortlist rate on panel work. Count wins and fees as totals as well as rates. A firm that drops the bids it was least likely to win will see its win rate rise even if it wins no more work.
How a win rate rises without more wins
Pitches you didn’t enter. AI can read a tender against your record of wins and losses and flag the kind you rarely win: a panel tender in a sector where you keep losing, say, flagged before anyone opens a template. The return here is the days not spent on a likely loss, moved to a bid you can win. Where the award is published, you can check who won the bids you skipped.
Client emails and newsletters. Count replies, not opens. Apple Mail now loads messages in the background, which registers as an open whether anyone read them or not. A reply is a client starting a conversation, and nobody has to argue about whether it happened.
Events. RSVPs, then the share who turn up, then the follow-up meetings that come out of the room.
Hours. Keep the figure, and put it last. It shows what the AI freed. The numbers above show what the team did with it.
How to set the baseline before the AI goes in
The return is the difference between before and after, so “before” has to exist. Take the twelve months ahead of the change from your bid register, CRM and email platform. Where hours were never logged, ask the people who do the work for an estimate, and label it one.
Then compare like with like afterwards: panel tenders with panel tenders, the monthly newsletter with the monthly newsletter.
Why comparing AI users with everyone else misleads
The obvious test is to compare work done with the AI against work done without it. That test flatters the AI. The people who pick up a new tool first tend to be the keen ones, and the keen ones can be the people already winning. You end up measuring who chose the tool, not what the tool did.
A misleading comparison and a fair one
Compare the same people before and after instead. Or put the AI into one practice group’s whole workflow and compare that group with its own history.
Why some AI moves the hours and not the results
An AI that drafts from your last proposal will cut the hours. It gives the results no reason to move, because the client reads the same argument as last time, written faster.
You can check your own tool. Put a proposal it drafted beside the document it drew from. If the argument is the same, the hours fell and the win rate had no reason to follow.
Results move when the work changes: each proposal starts from what this client is considering, and each email from what this reader cares about. The work changes first, then the AI is built to do the new work.
Searchlight Group shows both effects. Searchlight is a recruitment firm that pitches for government and not-for-profit work. Evaluation committees score its proposals the way a law firm’s panel review works: a committee, a scoring sheet and a stack of responses that sound alike. Its proposals took six hours each, and it won 35% of them. Taleist rebuilt how the firm pitched first, then built the AI it pitches with. Its win rate reached 85% in 2025, and proposals now take two hours.
Searchlight is one firm, and your bids, clients and volume differ from its. That is why your own baseline matters more than anyone’s case study, ours included.
Who can help you measure it
Your own BD or finance team, if the bid register and CRM are in order. It’s the cheapest route when the AI is in place and the question is only whether it’s working.
Your AI tool’s vendor, for adoption. Its dashboard shows use, not whether the client said yes.
A consultancy that rebuilds the work before building the AI. Taleist does this for professional services firms: we rebuild the workflow to start from the client, then build AI that knows the firm and knows how to persuade, for pitches, proposals, client emails and events. Taleist calls this 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. It fits a firm whose AI is saving hours without lifting results. It doesn’t fit a firm that only wants a usage report.
Automation agencies, pitch software and AI strategy consultancies sell into BD too, and they change different things. Our guide compares them: who builds AI for business development.
Sources: Dell’Acqua and colleagues, “The Cybernetic Teammate: A Field Experiment on Generative AI Reshaping Teamwork and Expertise”, Harvard Business School working paper 25-043, 2025. Searchlight Group results as published on taleist.agency.