The most common pricing mistake we see isn't picking the wrong number. It's anchoring to the wrong thing entirely. Builders look at what the AI costs to run and add a margin. Buyers don't care what your agent costs to run — they care what the job is worth. Price the job.
Start with the job, not the tokens
Every useful agent replaces or upgrades a specific piece of work: qualifying leads, drafting responses, running weekly check-ins, analysing documents. That work already has a price in your buyer's world — an hourly rate, a freelancer invoice, a salary slice, or the cost of it simply not getting done. Your pricing conversation starts there.
A useful exercise: write one sentence in the form “This agent does [job] that currently costs [what] and delivers it [how much faster/better].”If you can't fill that in, the problem isn't pricing — the agent's job isn't sharp enough yet. Fix the workflow first.
The three models, and when each wins
Almost every agent price is one of three shapes — and the right one follows from how the buyer experiences the value.
- Subscription (monthly) — the default for agents that work continuously: check-ins, inbox handling, content pipelines, monitoring. The buyer forms a habit; recurring value earns recurring billing. Most niche agents land between the price of a software tool and a fraction of the human alternative — which still leaves generous room.
- Pay-per-use — best when value arrives in discrete, high-stakes units: a document analysed, an audit run, a report produced. Feels fair at low volume, scales automatically with heavy users, and lowers the barrier for a first try.
- One-off — right for setup-shaped deliverables: a configured agent handed over, a brand strategy produced, an initial build. Often strongest paired with a subscription (setup fee + monthly).
Anchor high, because the alternative is a human
The comparison your buyer is silently making is not “this agent vs a cheaper agent.” It's “this agent vs paying a person, or vs doing it myself at 11pm.” A consultant who charges by the hour for audits doesn't undercut themselves by shipping an agent that runs the same rubric — they extend themselves. The agent works for people who could never afford the consultant's calendar, at a price that would be absurd for software and a bargain for expertise.
You're not selling compute. You're selling your judgment, running while you sleep.
The four mistakes that cap your revenue
- Cost-plus pricing. Metering your price to model costs broadcasts 'this is a commodity.' Your knowledge base, your rubric, your voice — that's the product, and it doesn't get cheaper per token.
- One tier. A single price forces every buyer into the same box. Two or three tiers (self-serve / standard / premium-with-you-attached) lets the market tell you who they are.
- Free without a wall. A free tier is a funnel only if something real sits behind the wall — a usage cap, a premium capability, or your personal review. Free forever with no wall is a donation.
- Never raising prices. Your first price is a hypothesis. If nobody flinches, it was too low. Early customers can keep founding terms — that's loyalty, not lost revenue — while new customers pay what the job is worth.
Where the mechanics stop being your problem
Everything above is strategy, and it's yours to own — nobody can price your expertise for you. What shouldn't be yours to own is the machinery: billing, payment collection, payouts, taxes, refunds, usage metering, and the awkward dunning email when a card fails.
That's the half we built Squidgy to handle. On the platform you pick the shape — one-off, subscription, or pay-per-use — set the number, and the rails do the rest, with AI and messaging costs metered from a usage balance so a busy month never quietly eats your margin. You keep the majority of what you charge; the platform takes one simple fee on revenue collected. Change the price whenever the market teaches you something.
And if you're still one step earlier — no agent yet, just deep knowledge of a niche and a workflow worth automating — that's exactly the gap our founding-builder cohort exists for: twenty-five builders, hands-on with us, from described workflow to first revenue. Pricing included in the hand-holding.