Why we're not the cheapest, and when cheaper is the right call
The honest answer to the three questions behind most pricing conversations: why quotes differ so much, whether you could just build it yourself, and how to tell a good quote from a cheap one.
By Ahsan Ahmad, Chief Executive Officer · Last reviewed
The short answer
AI agent quotes range from about $5,000 to $100,000 because suppliers are quoting different things: a demo at one end, a system with integrations, approval paths, an evaluation set and monitoring at the other. Ours sit in the middle - fixed prices from $4,500, most builds between $6,000 and $26,000, model costs at cost. A freelancer, a SaaS agent or doing it yourself is often cheaper and sometimes the right answer. This page shows when.
Why do AI agent quotes range from $5,000 to $100,000?
Two quotes for "an AI agent that answers customer questions" can differ tenfold and both be honest. The difference is almost never the model or the hourly rate. It is what the quote includes, and who pays when the estimate is wrong.
| What is included | Around $5,000 | Our fixed price | $60,000+ |
|---|---|---|---|
| Integrations | One, often read-only | As scoped: typically 1-5 | Many, plus SSO and enterprise systems |
| Evaluation set from your real cases | Rarely | Always, agreed before build | Always |
| What it must never do, in writing | No | Yes, with approval before irreversible actions | Yes |
| Monitoring and cost alerts | No | Yes, from day one | Yes, often with a dashboard |
| Handover and runbooks | Thin | Recorded handover, runbooks, your repos | Extensive |
| Who absorbs an overrun | You, usually | Us | Usually you, via change requests |
| Who you talk to | The builder | The engineer building it | An account manager, then the team |
The four things that move a quote most, in order: the number of systems the agent touches, how rigorously it is tested before launch, whether it acts or only answers, and where the engineers are paid. None of them is the model you choose. Our AI agent cost guide has the market-wide numbers with the arithmetic shown.
Can't I just build it myself with Claude Code or Cursor?
Sometimes, yes. AI coding tools have made the first working version of an agent cheap. A technical founder can get a prototype answering questions over their own documents in a weekend, and for an internal tool where a wrong answer costs little, that may be all you need.
What the tools do not do is the part that costs money in a professional build. They will not tell you what the agent must never do, build a test set from your real tickets, notice that it answers confidently from a document that was replaced last month, or keep it working after the model provider ships an update. They write the code you ask for, and they cannot see the failures you did not think to ask about.
The full answer, with a checklist for deciding, is in can I build it myself with Claude Code or Cursor?
Freelancer, DIY, SaaS agent or a firm: which is cheapest?
Each of the cheaper options is the right answer for some projects. The mistake is choosing on the first invoice rather than on what happens in month nine.
| Option | Typical cost | Right answer when | Where it goes wrong |
|---|---|---|---|
| Build it yourself | Your time plus model costs | Internal tool, prototype, low stakes, you can maintain it | Customer-facing, or it acts on money or data |
| Freelancer | $60-150/hr | Scoped work under about $5,000 | They move on, and nobody else knows the system |
| SaaS agent (e.g. per-resolution) | Around $1 per resolution, plus seats | A product covers 80% of the need at your volume | Volume grows and you are paying rent on every answer |
| AI engineering firm | From $4,500 fixed, or $2,800/engineer/month | The system has to keep working, or needs several skills | The scope is small enough that overhead dominates |
For the freelancer question in detail, see AI agency vs freelancer. For hiring, see white-label AI vs in-house.
Is it cheaper to rent a SaaS agent or own a custom one?
At low volume, renting is cheaper and it is not close. A SaaS support agent priced per resolution costs a few hundred dollars a month at a few hundred resolutions, with no build and no maintenance. Owning only wins once the monthly rent passes what a custom agent costs to run and maintain, plus its build cost spread over the years you use it.
In our arithmetic that crossover sits at roughly 1,500 to 2,000 resolutions a month. Below it, rent. Above it, the gap widens every month. The worked three-year comparison is in custom AI agent vs SaaS AI agent: the three-year cost.
How do you check whether any AI quote is good, including ours?
Six questions separate a quote for a working system from a quote for a demo. Ask them of everyone, us included, and treat a vague answer as the answer.
- Show me the evaluation set. How many real cases will the system be tested on, who labels them, and what score counts as done?
- What will it never do? Ask for the written list of actions that need a human first. No list means nobody has thought about it.
- Whose accounts is it built in? Repositories, cloud and model API keys should be yours from the first commit.
- What happens when the model changes? Providers retire models on a schedule. Ask who notices and who pays to re-test.
- Who absorbs an overrun? Fixed price, capped hours or open-ended: get it in writing.
- Can I meet the engineer who builds it? Not the person who sells it - the one writing the code.
For the longer list, see our checklist for hiring an AI development partner.
How do we price an AI build?
Fixed prices for defined work, a monthly rate for ongoing work, and nothing hidden in between.
| What you buy | Scope | Price | Timeline |
|---|---|---|---|
| Single-purpose agent | One job over your own content, one or two integrations | $4,500-8,000 | 2-4 weeks |
| Production agent | Multi-step, a few integrations, human handoff and approval paths | $8,000-18,000 | 4-7 weeks |
| Agent platform | Several agents or systems, shared tooling, permissions and evaluation | $14,000-26,000 | 10-16 weeks |
| Embedded engineering pod | Scope still moving, or an ongoing roadmap | $2,800 per engineer a month | Rolling monthly |
What is always included
A written capability boundary, an evaluation set agreed before build, monitoring from launch, a recorded handover and runbooks. Everything lives in your repositories and cloud accounts.
What is never added
No markup on model API or hosting spend - it is billed to your accounts at cost. No annual contract. No percentage of anything.
Where we are the wrong choice
Under about $5,000 of scoped work, a freelancer is better value. If a SaaS product covers most of what you need at your volume, buy it. We would rather tell you that on a call than take work we are wrong for. Full rates are on pricing.
Frequently asked questions
Because they are quoting different things. The cheap quote is usually a working demo: one prompt, one integration, no evaluation set and no plan for what happens when it is wrong. The expensive one includes the integrations, the approval paths, a labelled test set, monitoring and a handover. Ask each supplier for a line-item list of what is included and the gap usually explains itself.
No. Freelancers, no-code tools and off-the-shelf SaaS agents are all cheaper to start. Our fixed-price agents run $4,500-8,000 for a single-purpose agent, $8,000-18,000 for a production agent and $14,000-26,000 for an agent platform. We are cheaper than most US firms, and we say plainly when one of the cheaper options is the right answer.
Often, yes - for an internal tool, a prototype or anything where a wrong answer costs little. AI coding tools write working code quickly. What they do not do on their own is decide what the agent must never do, build a test set from your real cases, or keep the system working after a model update. Those parts are where most of a professional build's cost goes.
So that the risk of an overrun sits with us, not you. A fixed price only works with a written scope, so the capability boundary, the systems involved and the evaluation threshold are agreed before work starts. Changes to that scope are priced as change requests, in writing, before they are built.
No. Model API spend and cloud hosting are billed to your own accounts at cost. We never resell tokens with a margin, and the keys are in your name from the first day, so you can see exactly what the system costs to run.
Under about $5,000 of well-scoped work, hire a freelancer. If an off-the-shelf product does 80% of what you need at your volume, buy it. If you have engineers and the stakes are low, build it yourself. We will tell you the same thing on a call.
Want a straight answer on your project?
Tell us what you are building on a free call. If a cheaper option is right, we will say so.