Pricing
Most agencies in this category won’t publish a number. We think that’s a tell.
The short version: most agencies in this category won’t publish a number, and we think that’s a tell. The bands below are real; your exact quote comes from the audit, because integration depth — not agent count — is what drives cost.
Last updated:
Here are real bands. Your actual quote comes from the audit, because integration depth — not agent count — is what drives cost.
The audit
Credited in full against a subsequent build.
Builds
| Scope | Band | Typical timeline |
|---|---|---|
| Single-workflow agent, supported platform | [BAND A] | 2–4 weeks |
| Single-workflow agent, custom integration | [BAND B] | 4–8 weeks |
| Multi-workflow deployment | [BAND C] | 8–16 weeks |
| Multi-agent system, enterprise integration | [BAND D] | 16 weeks+ |
For reference, published market ranges for custom AI agent development in 2026 sit around $5,000–$30,000 for a simple single-function agent, $30,000–$120,000 for mid-market autonomous workflow agents, and $150,000–$450,000+ for enterprise multi-agent systems (Tenfold, 2026; ProductCrafters, 2026; SoftTeco, 2026). We sit where our scope sits and we’ll tell you which band you’re in before you commit.
Running costs — the part usually omitted
A build quote without these is incomplete.
| Cost | Typical |
|---|---|
| Voice usage | $0.05–$0.15 per minute (mainstream); up to $0.40 premium |
| Model inference | Varies by task; modelled per workflow in the audit |
| Vector storage and retrieval | Usually modest; material at high document volume |
| Cloud infrastructure | Billed direct to your accounts, not marked up by us |
We do not resell infrastructure. Your cloud bill goes to you at cost.
AgentOps retainer
Scaled to agent count and workflow criticality. Covers monitoring, evaluation runs, model upgrades and deprecation migrations, integration repair, drift detection, and a defined allocation of workflow expansion each month.
What we don’t charge for
- Discovery calls before the audit
- The ROI calculator
- Re-quoting when we got the scope wrong
- Infrastructure markup
Frequently asked questions
Why is there a range rather than a price?
Because integration depth varies enormously. The same agent against a modern API and against a twenty-year-old on-premise system are different projects. The audit determines which you have.
Is it cheaper to build in-house?
Sometimes, genuinely. If you have engineers with agent experience and capacity, in-house is often better — you keep the knowledge. What people underestimate is evaluation infrastructure and ongoing maintenance, not the initial build.
Can we start smaller?
The audit is the small start. It’s designed so you can test how we work for a fixed, modest amount before committing to anything.
Do you do revenue-share or performance pricing?
No. It sounds attractive and it aligns incentives badly — it pushes toward workflows that are easy to measure rather than valuable to fix.
Are there ongoing licence fees?
Not to us. You’ll have model and infrastructure costs, billed to your accounts. We don’t licence you software we wrote for you.