AI made the draft cheap. Everyone read that as “proposals get faster.” The real change is stranger: the work moved to the two ends. The plan that becomes the prompt, and the review that catches what a cheap draft misses. The middle compressed. The judgment didn’t.
For decades the long pole of a proposal was the draft. Getting compliant, persuasive words onto the page was the slow, expensive part, so the process front-loaded toward it: assign sections, write, write, write, and review with whatever hours were left before the deadline. Planning was a meeting; review was a scramble. The draft was the mountain.
When a compliant-looking draft takes minutes, the mountain in the middle flattens. The leverage moves to the two ends.
Compliance matrix, bid/no-bid, win themes, an outline mapped to the score. Heavy, and where the bid is decided.
AI writes the volume against the plan; a human writes the load-bearing argument. Fast, and not precious.
Compliance, strategy, score-it, omissions, legal, does-it-read-human. Heavy, and where the bid is won or lost now.
The craft didn’t disappear when the draft got cheap. It moved out of the middle and into the two ends, where it was always the scarcest.
Here’s the front bookend, and it’s bigger than an outline. The kickoff plan and the prompt are the same artifact now, and it’s everything that decides the draft: the positioning against the field, the data and proof points that survive scrutiny, the context(the customer’s mission, the hot buttons, how it’ll be scored), the content you’re building from, the win themes, the complianceeach section answers, and the scoring hook it has to hit. Assemble that, section by section, and the draft is a formality. Skip it and AI doesn’t save you; it produces the wrong thing faster and you iterate on garbage. The hour you spend here is the highest-leverage hour of the pursuit.
And this is where the inversion meets the asset. Most of that plan isn’t invented per bid — it’s drawn from your pursuit memory: the positioning, the proof, the win themes, the language that survived the last review. A company on its fifth bid assembles a sharper plan in an afternoon than a cold start manages in a week. So the plan, not the draft, is the real product— and the reason the work compounds.
The back bookend is where the era really shows. When anyone can generate a draft that lookscompliant in minutes, the only thing of value is someone who can tell compliant from compliant-looking — and strategic from generic, and true from plausible, and written-by-a-person from written-by-the-same-model-everyone-else-used. That last one is new. In a field filling with look-alike submissions (see proposal Kessler syndrome), sounding like a human is a discriminator, not a nicety. Review isn’t one pass anymore; it’s a stack of lenses, each asking a different question the cheap draft can’t answer for itself.
Press-button-get-proposal is the compressed middle, sold as the whole thing. It’s the part that got cheap, which is exactly why it’s the part that matters least. The tools optimize the mountain that already flattened and leave you alone at both bookends — where the bid is actually decided. That’s the same squeeze that’s folding the specialized platforms: they automated the easy 30% and assumed you had people for the 70% that's hard.
On your next bid, move your senior hour. Spend it on the kickoff plan — section by section, detailed enough to prompt on — and on a real review stack at the end. Let the draft be fast and unprecious in between. If your best person is spending their time in the middle, they’re spending it on the part that got cheap.
A senior operator owns the plan and the review; AI compresses the middle under their hand. The two ends where the bid is decided, run by someone who has sat on the evaluation side.
See how a pursuit runs →