Jonathon's AI Wiki

Pricing and Selling AI Automations on Value (Nate Herk)

8 min read

Source: raw/Build_Sell_with_Codex_5+_Hour_Course.md — the selling modules of Nate Herk’s “Build & Sell with Codex 5+ Hour Course” (youtube.com/watch?v=X-pbJWKmwi0): his “Still Relevant Next Year” talk at the Workless AI event in Montenegro and a follow-up pricing lesson. Transcript fetched 2026-09-29. Creator: Nate Herk. Platform: YouTube. Skill level: operator (agency or freelance builder).

Nate Herk’s method for selling AI automation work without guessing: find the business’s real constraint, agree on one number the build will move, then price at a fraction of the value the client has stated. It answers the pricing question left open in The 5 AI Automations Businesses Actually Pay For with worked numbers, payment milestones, a maintenance definition and scripts for budget pushback. All figures are one practitioner’s self-reported experience; his credentials (an agency scaled past 20,000/month minimum retainer, then sold) are unverified.

Key Takeaways

  • Three mistakes behind an unhappy client. An early client told him “We are not getting the value that we paid for.” Nate’s diagnosis: he built something that was not the constraint (a personal assistant, when the real pain was hand-written proposals), chose no KPI, and guessed the price. Your job is to “diagnose and prove”.
  • Find the constraint first. A business is either supply-constrained (more demand than it can serve) or demand-constrained (not enough coming in). His two questions:
    • Supply: “if you had 10x the business tomorrow, what would break first?” Then be quiet: “whatever breaks first is the first clog.” Fixing one clog reveals the next, and that sequence is how a project becomes a retainer.
    • Demand: “What would you do to 10x the business coming in tomorrow?” — then build systems that support that initiative.
  • Example: a med spa. The owner asked for a lead-generation agent. On the call she explained that leads came in steadily but did not show up to appointments, and there was no follow-up after visits. The better build was reactivation plus appointment reminders.
  • Use silence. After a diagnostic question, stop talking; clients fill the pause with the real answer. On discovery calls, “your job is to sit there and listen.”
  • Sometimes the answer is not AI. A supply-constrained service business may just need to hire; knowing “when not to use AI” is part of the value.
  • Agree one KPI with the stakeholder before any payment, with a baseline and a target — e.g. an appointment-setting agent taking a business from 5 to 10–12 appointments a week. An objective number is how you later prove the build worked.
  • Price on value, not hours. “Cost is the floor… value is the ceiling”, and the price is anything between them. Hourly billing rewards slowness: the slower developer earns three times as much for the same result. “Cost doesn’t justify price. Price justifies cost” (Nate credits Jonathan Stark, who spoke on pricing at Nate’s AIS Live event, with the landscaper example: a landscaper who buys a new truck cannot double the price of the same lawn).
  • Rule of thumb: about 10% (flexing to 10–20%) of first-year annualized value, so the client can see a 10x return. Assume the next question is “how did you get that number?” and walk them through the math.
  • Hold the price; cut the scope. When the number exceeds the budget, “reduce the scope a little bit and start with a smaller project” rather than discounting, which teaches the client “your prices move down every time that they frown.”
  • Maintenance is not new features. His standard $400/month keeps the build doing what was agreed (breakages, API changes, new models, edge cases). New functionality is a separate conversation.
  • Measure the after-state. His biggest mistake on the appointment-setter project: he never measured results after launch, which “cost me on the next conversation.” Check the KPI at one, two and three months and show it to the client.

Worked examples

  • Appointment-setting agent (real build). 20 leads/week × 1 hour each × 800/week × 52 = 5,500, roughly a 7.5x first-year return. He counted on growth in the lead baseline to make up the other 2.5x, and warns not to guarantee that. Maintenance: $400/month.
  • **Customer-support rep, 1 hour/day at ~12,000/year → a $1,200 build at 10%.
  • Construction crew phone orders. Converting daily phone orders into the crew’s text format saved only 45 minutes a day but avoided about $12,000/month in scheduling errors. Once you have experience, price from bottom-line impact rather than hours saved (e.g. what each booked appointment is worth in closed sales).
  • Budget mismatch. Value comes to 7,000 budget: scope a $7,000 V1 and schedule the rest for when the budget is there or the V1 has earned money back.

The sales process, step by step

  1. Discovery. Let them “dump everything”, then run LRP: listen, repeat, poke. Ask what changes for the business once the project succeeds.
  2. Why this, why now, why me.
    • Why this checks the request gets the outcome — the real fix may be “a really simple deterministic script and a Slack notification”.
    • Why now surfaces urgency.
    • Why me (“couldn’t you vibe code this? …hand it to an intern?”) brings objections out early. If they say a nephew could build it, agree, then ask who will watch it “at 2 in the morning when the model updates”, and wait for their answer.
  3. Size the value when they won’t share salaries. “How long does this take you today? How many people touch it? And what happens when it goes wrong?” Add proxy questions (locations, volume on the busiest days). On a job marketplace, state your assumption (“priced assuming about 200 of these a month”) and offer a 15-minute call to adjust. If you still can’t land on a number, you are not ready to write the proposal.
  4. Proposal. Open with three short paragraphs about the client — where they are now, in their words and numbers; where they want to be; why you can get them there. Then offer three tiers at about 10%, 25% and 50% of first-year value, each with more functionality or results. The middle one is “designed to win”; three options change the question from “should we work with this person?” to “how should we work with this person?”
  5. Payments tied to objective milestones about 30 days apart, e.g. a $9,000 project in three payments (start, midpoint, production). A milestone must be provable (“the agent pulls from the database and responds within a minute”), not “the inbox agent is working as expected”.
  6. Scope creep goes to a V2 backlog. Treat new requests as a good sign — they are already imagining working with you longer.
  7. Retainer. Start with one value-priced project; once the KPI moves, show the numbers and propose a retainer to keep removing constraints without re-scoping each time.

Accounts, costs and early projects

  • Client’s account, client’s card for API, cloud and token costs: “tokens are a utility bill”. Put an expected monthly run cost and its volume assumption in the proposal.
  • Testing costs can run from 1,000–3,000 to the price, pays for testing on his own keys, and switches to the client’s keys at production.
  • Hourly is fine for your first two or three projects (about $100/hour) while you have no proof; stop after that.
  • Red flag: a prospect who only wants a quote to compare vendors may just want the cheapest labor; he suggests walking away.
  • Before quoting, get the client to say out loud what the problem costs them. “Your price is just a fraction of a number that they have already stated.”

Try It

  1. On your next discovery call, ask “if you had 10x the business tomorrow, what would break first?” and wait at least a few seconds before speaking.
  2. Write the KPI as baseline → target in the proposal before any price appears.
  3. Compute first-year annualized value from their numbers, take 10–20%, and prepare the “how did you get that number?” walkthrough.
  4. Build three tiers (about 10/25/50%) with different functionality in each.
  5. When they say it’s over budget, cut scope, not rate.
  6. Book the one-, two- and three-month KPI check-ins at kickoff.

Open Questions

  • Garbled example. In the Montenegro talk, a support rep at 10 hours/week × 2,100” a year and “20,800/year and about $2,080; the transcript figures are not reliable.
  • No outcome data. There is no data on close rates or client retention for value-priced versus hourly proposals; everything here is his own experience.
  • Retainer pricing beyond the 20,000/month engagements were structured) is not covered.
  • Tier content. How the 10/25/50% tiers differ in deliverables is left to the builder.
  • Nate mentions a free 27–30-page pricing guide; it was not available as a source for this article.

Graph View

Backlinks