Hey there,
Recently I sat down with an architecture firm's CFO and its AI lead to scope the next phase of their AI work. Business development was their priority, and proposals were the obvious place to start: slow, painful, and exactly the kind of work AI is supposed to be good at.
Before we got into the how, I asked a different question. Of the last few projects you lost, why did you lose them?
"That's a great question," the CFO said. "I wish I had the answer to that." The firm gets debriefs, but he admitted they tend to justify a loss rather than learn from it. (This is incredibly common in every industry I've worked in.)
One of his colleagues in the meeting suggested we start by summarizing their pain points instead. I pushed back. Pain points matter, because they make the work slower and, well, more painful. But if fixing one isn't going to move the needle on winning work, you've just built a better process for writing losing proposals.
Then they articulated the real problem: the firm is often late to the game. On one recent pursuit, the strategic partners the firm called had started talking about the project with other companies 12 months earlier and teamed up nine months ago. Where were you? By the time the firm heard about it, it was too late: the teams were set.
Faster proposals wouldn't change any of that. They would just help the firm lose more efficiently.
This is how most AI programs pick their first projects. Someone asks where the pain is, and the team lists the work people complain about. It's a reasonable place to start. Painful work is visible, easy to measure, and easy to celebrate when it gets faster.
Nobody writes a proposal for its own sake. Its job is to win the work. A colleague in the meeting said it plainly: their top proposal writer is excellent, but that's "not the full picture of winning the award."
You've just built a much better process for writing losing proposals.
It's a classic jobs-to-be-done mistake: optimizing the task and ignoring the job the task exists to do. And it isn't unique to proposals. A faster month-end close doesn't help if nobody acts on the numbers. A smoother onboarding checklist doesn't help if new hires still take a year to get productive.
AI makes the mistake more tempting, because it's very good at making painful tasks less painful. The speedup shows up right away. Whether you solved the right problem shows up much later, if anyone checks at all.
The fix is to change the order of the questions. Before you ask what's slowing you down, ask why you lost the last five. Before you ask where AI can help, ask what outcome you're trying to change.
If you can't answer the first question, that's your first finding. And most organizations can't. Win/loss analysis is one of the most useful habits a business can build, and one of the rarest. When it happens at all, it's usually a quick debrief that explains the loss away and gets filed.
When it's done well, it tells you things you'd never guess. Years ago at Universal Electronics, we lost some of our early deals for a high-end custom remote control. When we dug into why, the answer surprised us: our price was too low. (Yes, too low.) In custom home theater installation, dealer and installer margins matter a lot, and our initial price left them too little. So, we raised the price, they made more on every sale, and our sales picked up. No amount of proposal polish would have found that.
Without that kind of analysis, this firm has no way to know whether any change to its proposal process will work. No AI tool fixes that. A habit of honest debriefs does.
By the end of the meeting, the goal for the next phase of my work with them had become clear. It wasn't faster proposals. It was visibility: what's out there, what they're working on, and what they're chasing. Hearing about the next pursuit nine months sooner will do more for their win rate than any proposal tool. Now, is that something AI can help with? Maybe...maybe not.
Most firms pick their first AI project by asking what hurts. Ask instead what result you're not getting, and why.
Break a Pencil,
Michael
P.S. What's the painful process your team most wants AI to fix? Reply to this email, or if you're reading on the web, write me at [email protected], and tell me whether it actually decides the outcome you care about. I'm curious how often the answer is no.
P.P.S. If someone you know is scoping their first AI project right now, share this with them before they pick it.
