The honest answer is a range
Anyone who quotes you a single number without seeing the workflow is guessing. A focused agent that reads one document type and routes it can land in the low five figures. A multi-agent system with evals, guardrails and integrations into your stack runs into the six figures. The spread is the scope, not the "AI".
The three ways to get there
The right choice depends on how many edge cases your workflow has and how much you can afford to learn on the job. Most teams under-scope the verification — the part that decides whether the agent is actually safe to run.
- DIY with no-code tools — cheap to start, breaks on the edge cases that matter
- in-house build — you keep the IP, you also carry the hiring and the maintenance
- agency or specialist build — paid upfront, but you get production discipline from day one
Where the budget actually goes
The model is not the expensive part. The expensive parts are the integrations, the evals, the guardrails and the maintenance — everything that turns a demo into something you can run unsupervised. That is the line between a weekend prototype and a production agent, and it is where the quote is made or broken.
Build vs buy
If an off-the-shelf tool already does the job well, buy it. Build only when the workflow is specific to your business — that specificity is the moat, and it is the only thing an agent adds that a SaaS cannot.
How to not overspend
Scope the smallest workflow that proves the value, ship it, measure it, then expand. Every dollar you spend before you have a working, measured pilot is a bet, not an investment. Run your numbers through the calculator before you sign anything.