Workflow breakdown

How much should a small-business AI or automation project cost?

“AI project” is too broad to have one useful price. Cost comes from the workflow underneath it: systems, data, security, exceptions, permissions, reliability and how much custom interface or change management is required.

Short answer: Publish fixed prices when the boundaries are genuinely repeatable. Scope custom work before implementation when the workflow or integration complexity is uncertain.

Illustrative workflow guide. It does not describe a specific client result.

The cost drivers

What actually makes a project bigger.

  • Systems: each integration adds authentication, field mapping and testing.
  • Data quality: clean structured data is easier than inconsistent free text, scans or old spreadsheets.
  • Security: sensitive customer, financial or employee data raises the implementation bar.
  • Edge cases: ten legitimate exceptions create more design work than one happy path.
  • Reliability: a twice-weekly workflow has different needs from a high-volume operational system.
  • Approvals: human review reduces risk but adds queue and state logic.
  • Interface: a background automation is simpler than a custom internal app.
  • Adoption: even a small build can be hard if several teams must change behavior.
Scoping workflow

How to reach an honest price.

Define the business outcome

State what should be different when the project works: fewer touches, faster handoff, consistent follow-up, less owner intervention.

Map the current workflow

Document triggers, people, systems, delays and exceptions before choosing tools.

Separate the normal path from exceptions

Identify what is predictable and what truly requires judgment.

Choose the minimum useful scope

Build the smallest version that removes meaningful friction without solving every adjacent problem.

Confirm access and security

Know what systems can be connected, what data can move and who can approve actions.

Price the defined scope

Once assumptions and responsibilities are clear, price before implementation begins.

Why productization still helps

Repeatable work should feel buyable.

A productized offer is useful when the core workflow, boundaries and implementation pattern are understood. That is why common fixes can have fixed or starting prices while custom work is scoped separately. Productization should reduce uncertainty, not pretend every company's systems are identical.

Questions that reveal scope

What a useful first conversation should answer.

  • What triggers the workflow?
  • Which systems are involved?
  • Where is the source of truth?
  • What are the common exceptions?
  • What can happen automatically and what needs approval?
  • What sensitive data is involved?
  • What happens if the workflow fails?
  • Who owns it after launch?
  • How will we know it worked?
Failure modes

Why AI projects overrun.

Pricing before understanding the workflow

Hidden integrations and exceptions surface later.

Starting with the model

The team chooses an AI tool before defining the operational problem.

Ignoring operating cost

Hosting, model, monitoring or support costs appear after launch.

No definition of done

If success is simply “use AI,” the project expands indefinitely.

Automating bad process

Technology makes a poor workflow faster without making it better.

When fixed pricing is reasonable

Use it when the boundaries are known.

Fixed pricing works best when inputs, outputs, integrations, support window and exceptions are repeatable enough to define in advance. When those assumptions vary materially, scoped custom pricing is more honest for both sides.

Where FrictionList fits

Start with the problem, not the product.

Related guide: Duplicate data entry.