Yesterday, 07:58 AM
Sharing what worked for me moving from generic dev freelancing to mostly AI/automation work, because pricing was the hardest part to figure out.
Mistakes I made at first:
- Charging hourly. Clients don't buy hours; they buy outcomes, and AI work is weirdly non-linear (sometimes a $500-value thing takes 45 minutes, sometimes a $200 thing takes 8 hours).
- Bidding based on effort, not value. If an automation saves a client 10 hours/week at $100/hr, it's worth a lot more than my 6 hours of build time.
- Including "maintenance" for free. Model updates break prompts. Things drift. I now charge a monthly retainer explicitly for monitoring and prompt maintenance.
- Agreeing to "can you just add..." scope creep without re-pricing.
What works for me now:
- Fixed project fee for v1, sized to value delivered rather than hours. Roughly 2-5x what I'd charge as an hourly equivalent for a "normal" dev project of similar size, because AI work has a "magic premium" but also a maintenance risk premium.
- 30-50% upfront, rest on delivery of v1 (not "when they're happy" — define done clearly).
- Separate monthly retainer (15-25% of project fee) for monitoring, goldenset checks, and prompt tweaks.
- A written scope doc listing what v1 does *and does not* do. AI clients' imaginations run wild; boundaries help everyone.
- I say no to "build me an AI that replaces my whole team" gigs. Those never end well.
I'm still learning. What pricing lessons have other AI freelancers learned the hard way? What's a clause you always include now?
Mistakes I made at first:
- Charging hourly. Clients don't buy hours; they buy outcomes, and AI work is weirdly non-linear (sometimes a $500-value thing takes 45 minutes, sometimes a $200 thing takes 8 hours).
- Bidding based on effort, not value. If an automation saves a client 10 hours/week at $100/hr, it's worth a lot more than my 6 hours of build time.
- Including "maintenance" for free. Model updates break prompts. Things drift. I now charge a monthly retainer explicitly for monitoring and prompt maintenance.
- Agreeing to "can you just add..." scope creep without re-pricing.
What works for me now:
- Fixed project fee for v1, sized to value delivered rather than hours. Roughly 2-5x what I'd charge as an hourly equivalent for a "normal" dev project of similar size, because AI work has a "magic premium" but also a maintenance risk premium.
- 30-50% upfront, rest on delivery of v1 (not "when they're happy" — define done clearly).
- Separate monthly retainer (15-25% of project fee) for monitoring, goldenset checks, and prompt tweaks.
- A written scope doc listing what v1 does *and does not* do. AI clients' imaginations run wild; boundaries help everyone.
- I say no to "build me an AI that replaces my whole team" gigs. Those never end well.
I'm still learning. What pricing lessons have other AI freelancers learned the hard way? What's a clause you always include now?