An executive summary, before and after

Open on the problem the client named in the tender, make every heading a result the client gets, and put proof under each one. AI can draft that if you give it the tender and those rules. Give it only your past summaries and it tends to copy what they do, including opening on the firm.

We wrote the “before” ourselves, but not from nothing. Every habit in it comes from years of reading and rewriting firm pitches and tender responses. You be the judge: does it look like the summaries you’ve read across your career? The “after” is our rewrite for the same tender, and both are scored below.

What did the tender ask for?

Panel firms that would add capacity to the council’s own legal team as the district grew: local, accessible and quick. The tender is real and public. In January 2024 Queenstown Lakes District Council in New Zealand went to market for a legal services panel, and its published procurement plan says why: “in response to the high growth, Council must augment their internal legal services capability and capacity”. It wanted firms that would “demonstrate a local perspective and act as an extension to the internal team”, be “accessible” and deliver “quick turnaround time frames”. The work covered environmental and RMA law including the District Plan review, infrastructure delivery, public and local government law, commercial, disputes and property, and each firm had to show capability in at least four of those.

In both summaries, “[Firm]” stands for an invented firm, and square brackets mark proof the firm would supply.

How did the two score?

The rewrite answered all seven of the council’s stated needs; the original answered none. A separate AI marker scored both summaries without being told which was meant to be better. We chose the tests, and most of them measure the rules the rewrite was written to, so read the table as a check that the rules were applied, not as independent judging.

The seven needs: added capacity as the district grows, working as an extension of the in-house team, a local perspective, accessibility, quick turnaround, the District Plan review, and infrastructure delivery.

The last row is the surprise. Most of the rewrite’s sentences still have the firm as their subject. What changed is what each sentence does: it acts on a need the council named. A summary doesn’t have to stop saying “we”. It has to stop saying “we” about nothing the client asked for.

The marker found two gaps in the rewrite as well. Neither summary says anything specific about value for money, which the plan weights at 20%. And the rewrite names only two or three practice areas when the council asked for capability in at least four. The fix is a sentence on how fees will be set and a line covering the other areas. The proof slots matter too: leave them empty and the three promises become claims with nothing behind them.

What made the difference?

Three changes: the opening, the headings and the proof. The rewrite opens on the problem the council itself named, in nearly its own words, so the reader sees in the first line that the firm read the tender. Each heading is a result the council gets, so a panel member skimming the headings reads the case. And each promise carries its proof, or a visible gap where the proof goes, so nothing rests on an adjective.

Cutting the courtesies did the rest. The greeting, the thank-you, the delight and the closing hope came to 44 of the first summary’s 190 words, and none of them told the council anything.

Is the “before” typical?

You’ll know from your own reading. The data we have points the same way, on firms’ own websites at least. 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%. None of the 55 opened with a line about the client. We haven’t counted tenders. But a homepage is how a firm describes itself when it controls every word, and a firm’s own copy is what an AI imitates when it’s given nothing else.

Can AI write the “after”?

Yes, if you give it the right inputs. Give it the tender’s stated objectives and evaluation criteria, and three rules: open on the client’s problem, make each heading a result, and put proof under each heading. Then tell it to mark missing proof in brackets rather than supply it.

That last rule matters. Asked to fill a gap, AI can write a plausible project or figure that never existed. Brackets keep the gap visible until someone who knows the work fills it with something true.

What AI won’t do on its own is decide which results matter most to this client. Give it last year’s summaries and nothing else, and it has nothing to imitate but last year’s habits. Our guide to whether the next AI model will fix your pitches explains why a newer model doesn’t change that.

Taleist builds AI that drafts this way 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. It isn’t the only route. Our guide compares the others: who builds AI for business development.