Short answerThere is no price list for AI automation. An AI pilot gets a written quote based on scope, after a first conversation that is free. What moves the quote is how tidy the data is, how many exceptions there are and how much human review is needed, not how clever the task sounds. This guide contains no amounts.

Why is there no price list for AI automation?

Because the name of a task says little about the effort. “Processing documents” can mean one tidy PDF or many different photos and scans. So every pilot starts with a first conversation, which is free, and continues with a written quote based on scope.

The quote separates the build, the upkeep and the cost of outside tools. This guide gives no amount and no range.

What changes the quote?

The table lists ten things, why they affect the scope and what you can prepare before the conversation.

What changes the quote for AI automation, why, and what to prepare
What changes the quoteWhyWhat to prepare
Process boundariesThe more steps and branches, the more there is to build and testA description of one job from first step to last, and what lies outside it
Format of the inputsA form with fields is simpler than free text, a scan or a photoA few real examples, both tidy and messy
Access to toolsNot every tool supports every connection, and the limits are checked before buildingA list of the tools you use and who has access to which data
ExceptionsEvery unusual case needs a rule or a person to take itThe cases that do not follow the usual path
Human reviewThe more review and approval, the more building around itWho reviews, what they decide and what must never be automated
Number of tools and connectionsEach connection between tools is a separate piece of workWhich tools have to exchange data
LanguagesBosnian, English and a mix are each tested separatelyThe languages in which data arrives and messages go out
Outside servicesSubscriptions and usage are paid under the provider’s terms and listed separately from the buildWhich services you already have and under whose account
Where the solution runsIt runs in your tool or in an outside service, and someone has to keep it running and pay for itWhat you already have and who maintains it
Support and maintenanceInputs and tools change over timeHow much support you expect after launch

Why do exceptions and review often come as a surprise?

The tidy case is the easy part. The effort goes into what does not follow the usual path: incomplete data, duplicates, an unreadable scan, a message in another language. For each, someone decides whether a rule or a person handles it.

So ask how many cases are messy and who takes them. We cannot estimate the share of exceptions in advance, because your data decides it. A pilot shows it on a sample.

What is one-off and what is ongoing?

Ask to see these two groups separately.

  • one-off: process mapping, building, testing, instructions and training
  • ongoing: subscriptions and usage of outside services, the place where the solution runs, support and maintenance, regular review of results

A quote that does not separate the groups does not tell you what you will pay after launch.

Two tasks that sound alike

Sample, fictional data

A fictional company says it wants to “automate document intake”. Task A: invoices arrive only as PDFs from one program, always in the same layout. The system saves them, checks for duplicates and creates a task for a person.

Task B: documents arrive by email as photos, scans and PDFs, from clients with different habits, in Bosnian and English. The system has to read amounts, match them and flag what is unclear, and a person reviews every suggestion.

Task B asks for more on every row of the table: more test cases, more exceptions, more review and more languages to check. Its quote would be larger, although both are “document intake”. This is not a real project.

How is a pilot accepted, and what is handed over?

Before building, agree what success means: the criterion, the test cases, the person who judges and what happens if it is not met. At the end we review the result, time, quality and remaining manual work with your team, and you decide whether the solution goes into use.

At launch we prepare instructions, training and agreed support. What is handed over, who has access to what and who owns the solution belongs in the quote.

What should you bring to the first conversation?

  • a description of one job and who does it today
  • how the data comes in, in which formats and roughly how many items a month
  • a few real examples, including bad ones, preferably without personal data
  • the tools you use and who has access to which data
  • the languages in which data arrives and messages go out
  • who reviews the result
  • what must never be automated

What should a written quote contain?

From us or from anyone else, a quote should answer these points:

  • which process is included and what is outside it
  • the inputs and formats
  • the tools and the connections to be built
  • who reviews and what they decide
  • the success criterion and the test cases
  • what happens if the pilot does not meet the criterion
  • the build separated from the ongoing costs
  • who pays for outside services and under whose account
  • what support and maintenance include
  • what is handed over at the end and who has access to what
  • ownership of the finished solution
  • how VAT is shown, if it applies

Frequently asked questions

Is there a price list for AI automation?

No. Every pilot is quoted by scope, because two tasks that sound the same can need very different effort. You get the quote in writing, and the first conversation is free.

Is there a monthly fee?

It depends on the solution. Support after launch and outside subscriptions are listed separately from the build. Whether there is an ongoing monthly cost is agreed in writing before the start.

Who pays for the outside tools?

We list outside services and their costs separately in the quote, and the provider sets the terms. Whose account the subscriptions run under is agreed in writing before the start.

What if the pilot fails?

The success criterion is agreed before building, and at the end we review the result with your team. We do not promise an outcome. What happens if the criterion is not met is agreed in writing before the start.

Who owns what is built?

Your business data stays yours. Who owns the finished solution and the access it needs is agreed in writing before the start.

What does maintenance mean?

In general it is work after launch: fixes when inputs or tools change, and a review of mistakes. What the agreed support includes, and whether it is charged separately, is stated in the quote.

How long does a build take?

We cannot name a duration in advance, because it depends on scope. The deadline is agreed in writing once the scope is known.

APPLY IT TO YOUR BUSINESS

Describe one job in five sentences: where the data comes from, who sees it first, what is checked, what happens to exceptions and who decides. That is enough for a first conversation.

See also what a business can automate with AI, AI automation for accounting offices, where to start with AI automation, AI and automation or contact.

How Midaz can help ↗