Always misjudging client needs?
Last week I was at a coffee shop discussing requirements with a client. When I got back and organized my notes, I realized I had completely missed the mark. I used to make this kind of misalignment mistake all the time. For us running one-person companies, the biggest fear is a client saying a bunch of stuff, us missing the key points, and then realizing we lost money on the quote. I've gotten stuck here too—when client requirements get long, I easily miss critical risk points and end up pulling all-nighters to fill the holes myself.
Using AI to diagnose like an ER
Recently, Harvard did a study putting OpenAI's o1 reasoning model (the kind that does deep thinking) head-to-head with doctors in an ER. It turns out it had higher diagnostic accuracy than two senior doctors across 76 real cases, even catching a rare flesh-eating bacterial infection 12 hours early. If it can help doctors diagnose patients, we can use it to "diagnose" our client needs. My friend Dada, a freelance designer, was at her WeWork desk last Wednesday night. She tossed a messy 20-minute voice memo transcript from a client into a reasoning model. In 5 minutes, the AI picked out two logical contradictions the client didn't even realize they had, helping her avoid a massive rework trap.
Replicate cost today
Money: About 140 RMB/month (for the premium version with reasoning capabilities; you can also make do with the free version). Time: 15 minutes. Technical barrier: As long as you can type and send a WeChat message. First step: Open ChatGPT, switch to the model with "reasoning" or "deep thinking" in the top left corner (if unavailable, use the strongest one). Then, select all and paste your client communication logs, and ask it: "As a senior project consultant, help me identify the hidden contradictions and risks that could lead to delays or budget overruns in this requirement."
Advice by stage
Just starting: If you're still in the exploration phase, not everyone needs this tool. It's fine to skip it for now; you can throw your own thoughts at a regular AI to help clear your mind first.
1-2 clients: If you're working on a new proposal for an existing client, I'd feed the requirements doc to a reasoning model for a "risk checkup" before sending my next quote to prevent issues.
Scaling: If you're leading a small team on big contracts, I'd suggest turning the risk points flagged by the AI into a standard requirements confirmation checklist. Run through it before internal team meetings to avoid unnecessary detours.