A tool for the wrong blocker went unused. The ones built for real problems got used.

A proposal generator built in 2025 for my own team worked. Nobody needed it.

I've since built tools that I use every day. The difference between them wasn't the technology. It was whether I had the real blocker right before I started.

The tool nobody needed

We were taking on consulting and vendor projects, and we weren't sending out many proposals. I assumed the reason was the time each one took to write. So I built a tool to produce them faster, to help my business partner, who was the one sending them.

Once it existed, I saw that writing speed had never been what held us back. There were three other things:

• Bandwidth. He didn't have the hours to pursue more deals, however fast the document was.
• Not enough contacts. We weren't doing the outreach that fills a pipeline, so there were few people to send a proposal to.
• The thinking before the writing. The slow part of a proposal is understanding what the customer wants, packaging it into something they can buy, and working out whether it suits our business. A generator doesn't do that.

I had heard "we can't send more proposals" and decided the cause myself. I picked the cause I knew how to fix.

This has happened to me more than once. It's easy to believe "the blocker is X" when X is something I can build.

The tools that got used

This year I built two things for my own work.

Scheduled lead-finding jobs. Four recurring jobs: two that look for leads daily and two weekly research runs. Each reads a ledger of what it found before, skips anyone already in my customer list, and files a short report. The first test runs cost between $0.16 and $0.36 in data each. I wrote up the setup, and what broke, in I moved my AI jobs off a US$99-a-month service.

A message digest. I get messages across four apps, and people were going unanswered. Every two hours my laptop reads the conversations where someone else spoke last and writes a short list: who to reply to now, who can wait, and who has been waiting more than a day. It can read and it can't send.

Why these were different

With the digest, I wasn't guessing at the blocker. I was the one with unanswered messages. I knew the problem wasn't typing replies faster. It was not knowing who was waiting.

With the proposal tool, I was guessing about someone else's work, and I never asked the question that would have settled it: if this took zero time, what would you do differently? The honest answer would have been "nothing", and I'd have spent that week on outreach.

I made the same kind of mistake with buyers, which I described in A demo is not a deal. Building is what I'm good at, so a missing tool is the explanation I reach for first.

The check I use now

Before building anything for someone else, I ask them three questions:

1. What stops you doing more of this today? Let them answer without suggestions.
2. If that were fixed tomorrow, what would stop you next?
3. If this task took zero time, what would change?

If the answers point at time, contacts or judgment, a faster tool won't help. That's also the first step of Audit, Optimize, Automate, Verify: look at the process before you automate any of it.