Cost to build an AI agent: scope the setup work

By Jeremy Kenerson, Founder at BreadHead.

Last updated October 7, 2026

$3,866 for setup and $716 a month to run are our estimates for the example Zapier workload below. They cover setup time, the tool plan, model access and upkeep for 10 workflows with 20 steps each, run 50 times each per month. Treat this as a workload example rather than a quote for your agent, and ask for setup and ongoing work to be priced separately.

How much does it cost to build an AI agent?

For setup of this Zapier workload, our estimate is $3,866, based on 60 owner hours valued at $64.44 an hour. Ask for a separate quote naming your agent's job, inputs and review needs.

Example

This is our example workload, not measured use: 10 workflows with 20 steps each, run 50 times each per month, or 500 runs and 10,000 steps. We assume each step is one successful billable task, with no extra retries or billable actions. We assume Zapier's hosted service handles these steps, with no separate server bill.

For every hour, including owner hours, we use the BLS 2025 hourly median for Software Developers, Quality Assurance Analysts, and Testers as a wage yardstick. A contractor or agency may charge more. Payroll taxes, benefits and hiring costs aren't in that wage.

Token use and models aren't specified. A chat plan price doesn't establish workflow API rights or enough usage. Ask the person quoting the work to confirm the access and usage your job needs.

These costs vary and aren't in the totals:

• Connected tools' subscriptions and API use, accessed through credentials such as logins and API keys.
• Taxes and extra usage.
• Payroll taxes, benefits, hiring costs and contractor or agency markup.

Choose the task before asking anyone to price it. Use the one-page task inventory to narrow the request to work you can name.

Use this check for your setup request:

• What starts the agent's job, and what should it produce?
• Which records may it read or change?
• Which accounts and paid tools must be connected?
• Who supplies the instructions and sample inputs?
• What testing and changes are included in setup?
• What work begins only after setup is accepted?

Name the actions you want to approve yourself. Use the agent permission boundary to set where the agent must stop.

Include how you'll check any facts in a draft. Use the draft claim check to prepare that part of the request.

Read the full basis before using the estimate to plan a budget. See What it costs for the full math, running costs and assumptions.

When should you hire an AI consultant?

Ask a consultant for help when you can't yet define the agent's job or judge what's included in a quote. Ask for help with those open questions before agreeing to a build.

Prepare the request in these steps:

1. Write the job in plain words. Attach sample inputs and describe the output you want to check.
2. List each tool involved. Ask which subscriptions and account access you'll need to supply.
3. Separate setup from upkeep. Ask who handles later changes, faults and usage bills.
4. Name what you can't yet decide. Ask for each unknown to be answered or clearly left out of the quote.
5. Agree on how you'll accept the work. Ask what happens if the build doesn't meet those checks.

Copy this table into your quote request. Fill in what you know, then ask the person quoting to answer the open questions.

Replace the task and its limits with your own. Ask the person quoting to restate the job in their answer. Mark any difference before agreeing to proceed.

Set the checks before accepting a build. Use the pilot acceptance sheet to record the work you'll test and who will approve it.

Compare software against the job and limits in your quote request. When you're ready to compare platforms, see the best AI workflow automation tools.

With BreadHead, you review and approve the work before it goes public or into use.

What changed

• October 7, 2026: First version. Facts checked.