Why we sell AI solutions at fixed prices

The pricing models that don't work

Three pricing models dominate AI services. Each is misaligned in a specific way.

Hourly consulting rewards taking longer. The customer's incentive is "ship fast and right." The vendor's incentive is "stay billable." We have watched 6-week projects stretch to 14 weeks under hourly billing on the exact same scope. The customer ends up paying more for a worse outcome because the vendor optimizes for the wrong thing.

Per-seat SaaS punishes successful adoption. If your AI assistant is good, more people use it. The vendor charges more. The customer's reward for getting their team to adopt the tool is a higher bill. The customer eventually rolls back adoption to control cost, then renews at lower seats, then doesn't renew at all.

Per-token / usage-based sounds aligned but isn't. The vendor's incentive is to make the agent verbose. Token efficiency improves the customer's bill but reduces the vendor's revenue. The vendor will not optimize against their own revenue, so they don't.

We picked productized fixed-tier pricing because none of the three default models align the vendor with the outcome the customer is buying.

The three tiers, and what each one buys

Starter ($25-$250 one-time or first month) — productized AI tooling for individuals and small teams. AI workflow audit, prompt library, single-channel assistant setup, one workflow automation with eval harness. Ships in days. The customer gets a working thing without committing to a relationship.

Pro ($250-$400/month) — multi-step agent workflows wired into the customer's stack. Includes hosting, eval harness, dashboards, weekly iteration cadence with a senior engineer, Slack support with a 24-hour SLA. The most popular tier. Production-grade scaffolding by default.

Custom (priced after a 30-min discovery call) — embedded engineering pod for ambitious roadmaps. Series-A+ companies, multi-tenant platforms, regulated workflows. Outcome-linked fees on the second engagement.

The math: most projects we've ever shipped fit cleanly in one of these three buckets. The exceptions either grew naturally from one tier to the next (Starter → Pro after 4-6 weeks of value) or were too small to be worth either side's time. The bucket sizes were calibrated against the actual distribution of what people need, not against what we wanted to sell.

Why fixed pricing is honest

Fixed pricing forces honesty in three places.

Scope is the contract. "Multi-step agent workflow with eval harness and dashboards" either ships or doesn't. There's no "we ran out of hours" conversation. The customer knows what's coming. We know what we're delivering. If the scope was wrong, that's our problem to solve, not theirs to fund.

Speed is rewarded, not punished. Shipping a Pro build in 8 days instead of 30 is good for both sides — the customer gets the value sooner, and we free engineering capacity for the next customer. Hourly billing actively works against this.

Adoption is decoupled from price. Once a Pro tier ships, the price is the price whether 5 people or 500 use the agent. Customers actually push adoption because the marginal user costs them nothing extra. Per-seat would make them ration access.

The economics underneath

Fixed pricing only works if the vendor has invested in productization — i.e., the same scaffolding ships on every engagement so each new build is additive effort, not from-scratch. Our Pro tier ships the same eval harness, the same observability layer, the same cost guardrails, the same dashboard structure on every build. The customization is the agent itself: the chain, the corpus, the integrations.

This is the move from consulting (every engagement is from-scratch, billed by hour) to productized services (most of the engagement is reused, billed by tier). The unit economics get better with each build because the scaffolding gets battle-tested without any specific customer paying for it. By Pro build #20, the scaffolding is stronger than what most one-off custom projects ever ship.

The Custom tier has different economics — embedded pods are closer to traditional consulting in cost structure, which is why they're priced separately. But we still ship them with the productized scaffolding underneath, so the engineering team starts at a higher altitude than greenfield.

What customers actually buy at each tier

Starter customers buy certainty — they want a working thing this week and they're done if it doesn't work. The Starter price is low enough that committing is easy and exiting is cheap. Most Starter customers never become Pro customers, and that's fine. The tier exists to be a clean entry point.

Pro customers buy the scaffolding plus iteration. They could in theory build the agent themselves. They are paying us to skip the 6 weeks of figuring out evals, gating, observability, and cost guardrails — and to keep iterating against a moving target. The retainer pays for being the team that owns the system's improvement curve.

Custom customers buy engineering capacity at senior altitude with productized infrastructure underneath. They have ambitious roadmaps and limited senior bandwidth. We become a pod that ships against their priorities while bringing the scaffolding patterns from every Pro build before them.

The pricing question we get most

"What's the discount for an annual commit?" We don't offer one. The price is the price. Annual commits don't change our cost structure on Pro tier — the work is monthly iteration, not seat licensing. What we do offer is a 30-day exit on Pro: try the tier for a month; if it's not landing, leave at the end of the month. The exit clause is more valuable than a discount because it builds trust that the price is the actual price.

Fixed pricing is the simplest expression of "we're confident this is worth what we're charging." If we needed hourly billing to make the unit economics work, we wouldn't yet have productized enough to be charging this way.

[See the tier breakdown](/pricing) or [browse what each tier ships](/services/playbook).