---
title: "AI readiness check: 10 questions in 15 minutes"
description: "A 10-question AI readiness check a company can do in 15 minutes. What each answer means, how to score it, and what to fix before you spend money on AI."
url: "https://madsalt.tech/ai-readiness-check/"
author: "Denys Zaichenko"
author_url: "https://madsalt.tech/about/"
language: "en"
published: 2026-11-12
---

# AI readiness check: 10 questions in 15 minutes

TL;DR

Most companies that thought about AI and did not start gave the same reason: no one who knows how. In the EU, 70.3% of firms that considered AI and did not adopt it named lack of expertise, ahead of legal doubts (53.6%), privacy (52.7%), data quality (43.5%) and cost (38.4%). So a useful readiness check does not ask "are you innovative". It asks whether you have a job worth automating, data to feed it, someone to check the output and someone to own it afterwards. Below are ten questions, a score, and what to fix at each level. The scoring thresholds are my own rule of thumb, not a standard.

I'm Denys, a forward deployed engineer in Porto. I use these questions myself before I quote a first job.

## What is an AI readiness assessment?

A short test of whether your company can use AI on a real job and keep it running. Big consultancies sell long versions with maturity levels. A company with no data team needs a shorter one: the answers tell you whether to build now, fix something first, or do nothing yet. Doing nothing is a fine result.

## Why is an AI readiness assessment important?

Because the stated barriers are about readiness, not about tools. The numbers:

![Why EU firms that considered AI did not adopt it. % of firms that had considered AI and did not adopt it. Several answers allowed. Lack of relevant expertise: 70.30; Unclear legal consequences: 53.60; Data protection and privacy: 52.70; Data availability or quality: 43.50; Incompatible existing systems: 41.60; Costs seem too high: 38.40.](https://madsalt.tech/charts/en/ai-readiness-check.svg)

*Figure 1. Eurostat, EU firms with 10 or more staff excluding finance, 2025 (KS-01-26-009, Table 7), read 12 November 2026.*

The EU figures are from Eurostat's 2025 survey of firms with 10 or more staff. They are shares of the firms that had considered AI and did not adopt it, which is about 14% of firms that do not use AI. 20.0% of all EU firms did use at least one AI technology in 2025. The reasons are multiple choice, so they add up to more than 100%.

The UK government's 2026 survey of 3,503 businesses points the same way. Limited AI skills affected 67% of firms using AI and 78% of those planning to. 65% of non-adopters cited no clear business case. The report notes that cost was more about the time and effort to embed AI than about tool prices.

## What does an AI readiness checklist include?

Answer each with Yes (2 points), Partly (1) or No (0). Do it with the person who runs the job, not alone.

1. **Can you name one repeated job that costs real hours every week?** One job, such as answering the same customer questions or retyping invoices. "Improve efficiency" is a No.
2. **Can you say how big it is in numbers?** Hours a month, items a month, cost of a mistake. Without a number you cannot tell later if it worked. This is the "clear business case" that 65% of UK non-adopters said they lacked.
3. **Does the job have a clear input and a clear output?** For example: an email in, a CRM record out. Fuzzy jobs that depend on judgment are poor first candidates.
4. **Do you have 20 to 30 real examples you can test on?** Real invoices, real emails, real tickets. Tests on invented examples prove little.
5. **Is the information the job needs written down somewhere?** In documents, a CRM, a database. If the answer lives in one person's head, the first task is to write it down.
6. **Is that information current and consistent?** Two price lists with different numbers will give two different answers. Data quality was a barrier for 43.5% of the EU firms.
7. **Can your systems hand data over?** An export, an API, an email inbox the system can read. Systems that cannot talk to each other were a barrier for 41.6% of the EU firms.
8. **Do you know what data may leave the company?** Customer personal data, contracts and health or finance records need a decision about where they may be processed. Legal doubt (53.6%) and privacy (52.7%) were the second and third reasons in the EU.
9. **Will a named person check the output?** Someone who can tell right from wrong and has the time. 81% of the UK businesses in the survey kept some human oversight of AI output.
10. **Will someone inside own it after it launches?** Fix it when it breaks, update the documents, read the alerts. This is the "lack of expertise" problem in a smaller form: you do not need to build it, you need someone to own it.

## How long does an AI readiness check take, and what does your score mean?

Add up the points, 0 to 20.

| Score    | What it means                                                                           | What to do                                                                                                                     |
| -------- | --------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------ |
| 0 to 7   | Not ready for a build. The gaps are in the job and the information, not the technology. | Do questions 1, 2 and 5 first. Pick one job and write down its numbers and its rules. Spend nothing on AI yet.                 |
| 8 to 14  | Ready for a small test. A few things will break, and you can find them cheaply.         | Fix any question you scored 0. Then run a test on 20 to 30 examples, with a written pass mark, before building anything large. |
| 15 to 20 | Ready for a first build.                                                                | Choose the job with the highest score on questions 3, 4 and 9, and scope it as one deliverable.                                |

One rule beats the total: a No on question 8 or 9 stops you whatever the score. A system that sends private data somewhere it should not go, or that nobody checks, can cause more harm than the job was costing you.

## What happens after the check?

By the question that scored 0:

- **Questions 1 to 3 (no job, no numbers):** spend a week writing down one process. Who does it, how often, how long it takes, how you know it was done right. This costs nothing and half the time it shows the job needs a simpler fix than AI.
- **Questions 4 to 6 (no examples, scattered or conflicting information):** collect the examples, merge the duplicate documents and name one source of truth. Do it before the build, not during it.
- **Question 7 (systems cannot hand over data):** often a small job, such as turning on an export or an API key. Ask before assuming it is a big one.
- **Question 8 (data rules):** write down which data may go to an outside AI service and which must stay inside. For EU companies the AI Act dates moved in 2026: the high-risk rules for the Annex III list now apply from 2 December 2027, while transparency rules (Article 50) apply and national enforcement starts on 2 August 2026. Ask a lawyer about your case, because I am not one.
- **Questions 9 and 10 (nobody to check or own it):** name the person before you start, with the hours they will have. If no one has those hours, a build will go unwatched.

Then test before you commit. The method, with a test of four models on 30 invoices, is in [invoice data extraction with AI](https://madsalt.tech/blog/invoice-data-extraction-with-ai/).

## Which tool or template is best for an AI readiness assessment?

Large vendors and consultancies publish their own assessments. I have not compared them, and most are built for big organisations. For a company with no data team, the ten questions above work as the template: they fit on one page, and the result is one job and a list of fixes. What decides whether a tool is useful is whether it ends with a named job, not a maturity score.

If the check says you need someone to build it, the next question is who. Use [hire an AI automation specialist](https://madsalt.tech/hire-ai-automation-specialist/) for the options and prices. Put your score and your one chosen job into the brief; the free [AI job description generator](https://madsalt.tech/tools/ai-job-description-generator/) turns them into a post or message you can send. For what running the result will cost each month, see [how much an AI agent costs per month](https://madsalt.tech/blog/how-much-does-an-ai-agent-cost/).

## My similar case

A client was paying about USD 500 a month for a client and time management subscription, and every member could still open every client. I built the client a system of its own: clients, engagements, weekly timesheets with approval, reminders, and confidential clients visible only to named members. It now runs on Supabase for USD 25 a month, and everything else in the stack is on free tiers. The build is in the [time tracking case](https://madsalt.tech/work/audit-firm-time-tracking/).

Want a second pair of eyes on your answers? Send me your score and the one job you picked, and we'll work out the best first step for you.

Send me your score

Tell me the job, your score and the weakest question. I'll reply with whether I would build now or fix something first, even if that means not hiring me yet.

[Write to me](https://madsalt.tech/#contact)

## Sources

- Eurostat, [The use of artificial intelligence technologies in the EU in 2025 (KS-01-26-009)](https://ec.europa.eu/eurostat/documents/7870049/23260410/KS-01-26-009-EN-N.pdf/37d063cb-28cf-3b4e-91f3-c3784c970842), Table 7, 2026: barrier shares of firms that had considered AI, all enterprises. Adoption 20.0% and 14.2% considered, same document. Firms with 10 or more staff, excluding finance.
- UK Department for Science, Innovation and Technology, [AI adoption survey report 2026: executive summary](https://www.gov.uk/government/publications/ai-adoption-survey-2026/ai-adoption-survey-report-2026-executive-summary), 29 October 2026 (3,503 businesses, fieldwork January to April 2026).
- European Commission AI Act Service Desk, [timeline for the implementation of the AI Act](https://ai-act-service-desk.ec.europa.eu/en/ai-act/timeline/timeline-implementation-eu-ai-act), opened 12 November 2026.
- The scoring and thresholds are my own rule of thumb. They are not from any of the sources above.

## More on starting with AI

- [Do you still need invoice OCR? I tested 4 AI models](https://madsalt.tech/blog/invoice-data-extraction-with-ai/)
- [How much does an AI agent cost per month? 4 real builds](https://madsalt.tech/blog/how-much-does-an-ai-agent-cost/)
