Sixteen Questions Before You Start an AI Project

A free self-check across eight dimensions: strategy, use cases, data, infrastructure, governance, skills, change and process. Tick what you can evidence, read your band, copy or print the result.

An AI readiness self-check tells you which of eight basics are missing before you pay for an AI build: a stated goal, a shortlist of use cases, reachable data, a test environment, a usage rule, someone to judge the output, a plan for the people affected and a process you can measure. StepTo is a 15-20 engineer nearshore team in Belgrade (CET), and this list was written by StepTo, so read it as a prompt and use it before you talk to us or to anyone else. It is a self-scored prompt, not a validated benchmark and not an audit of your systems.

The sixteen questions

0 of 16

Groundwork first

Most of the basics are open. Close the gaps listed below on paper before you spend on a build. Start with one use case, one data source and one owner.

  • Strategy0 of 2
  • Use cases0 of 2
  • Data0 of 2
  • Infrastructure0 of 2
  • Governance and risk0 of 2
  • Skills0 of 2
  • Change0 of 2
  • Process0 of 2

Tick a row only when you can point to the evidence: a document, a dated decision or a named person. Nothing you tick is stored or sent anywhere. The score is a prompt for what to fix first, not a verdict.

Strategy

  • If no, first step: Write the sentence with the person who owns the budget before you talk to any vendor.

  • If no, first step: Name one sponsor, even for a small pilot, and write down what they decide.

Use cases

  • If no, first step: List the repetitive, high-volume tasks first and pick the ones with a clear owner.

  • If no, first step: Pick a use case where the output is a draft or a suggestion, not an automatic decision.

Data

  • If no, first step: Trace where the data lives today and who can export a sample of it this week.

  • If no, first step: Ask legal or the data owner to classify the data as allowed, restricted or forbidden for external models.

Infrastructure

  • If no, first step: Ask the system owner what integration points exist today and what each one can read or write.

  • If no, first step: Set up a separate environment and a monthly usage limit before the first experiment.

Governance and risk

  • If no, first step: Publish a one-page rule: allowed tools, forbidden data, and who to ask.

  • If no, first step: Ask counsel or a compliance lead for a short written note per shortlisted use case.

Skills

  • If no, first step: Assign a domain expert to build a test set of real examples with the right answers.

  • If no, first step: Decide whether the pilot is built in house, by an outside team or by a mix, and name who operates it.

Change

  • If no, first step: Brief the affected team before the pilot starts and give them a way to report problems.

  • If no, first step: Write the manual path for the process and who switches to it.

Process

  • If no, first step: Map the process on one page with the people who run it, including the exceptions.

  • If no, first step: Measure the current process for two weeks and keep the numbers.

What to do with your score

Dimensions with no tick yet: Strategy, Use cases, Data, Infrastructure, Governance and risk, Skills, Change, Process. Close these first.

  • 0 to 5, Groundwork first. Most of the basics are open. Close the gaps listed below on paper before you spend on a build. Start with one use case, one data source and one owner.
  • 6 to 10, Ready in parts. You have a base. Fix the dimensions that scored zero, then scope one small pilot on the use case with the best data and a human reviewer.
  • 11 to 16, Ready to scope a pilot. Enough is in place to write a pilot brief. Check that each yes has evidence behind it, then define the test set, the budget cap and the stop condition.

How is the score calculated, and what are its limits?

There are two questions for each of the eight dimensions. Every ticked row is one point, so the total runs from 0 to 16. The bands are 0 to 5, 6 to 10 and 11 to 16, and they are simple cut points that StepTo chose. No survey data, norm group or outcome data stands behind them.

Four limits matter. The score is equal-weighted, so a missing data owner counts the same as a missing communication plan, although the first can stop a project alone. It relies on your honesty, which is why the rule is to tick a row only with evidence. It says nothing about how good a particular use case is. And it cannot see the quality of your data or your systems, only whether you have looked.

What do published frameworks say about readiness?

Two published sources on governance and AI literacy, both read on 2026-11-20. The NIST AI Risk Management Framework page says AI RMF 1.0 was released on January 26, 2023, names the four functions Govern, Map, Measure and Manage, and states that AI RMF 1.0 is being revised.

The text of Article 4 of the EU AI Act, AI literacy, as published on artificialintelligenceact.eu, says providers and deployers of AI systems must take measures to ensure "a sufficient level of AI literacy" among their staff and others operating AI on their behalf, and that, according to that page, it comes into force on 2 February 2025. This is not legal advice. Check the current text with counsel for your use case.

How do other vendors offer an AI assessment?

Other vendors publish assessments too. We link them as examples of the format, not as endorsements.

  • Vega IT reports an interactive 8-step AI readiness self-assessment on its site as of 2026-11-20.

    Source: vegaitglobal.com/digital-product-development/ai-readiness-assessment

  • Andela reports an agentic engineering assessment made of a 15-minute knowledge check, a 5-minute behaviors survey and a 1.5-hour real-world simulation, human-graded, ending in a diagnostic report, as of 2026-11-20. It measures how an engineering team uses AI coding tools, not whether a company's data and processes are ready.

    Source: andela.com/assess/agentic-engineering

Where a graded, human-reviewed assessment is what you need, such as Andela's, a self-scored list cannot match it. These vendors have larger teams and published client references. What this page offers instead is a free list you can finish in one sitting, scored in your browser, with every question and every limit visible.

What should you do after you have a score?

Fix the dimensions that scored zero before you commission a build. The StepTo AI strategy consulting page describes an AI readiness assessment that audits data quality and accessibility, existing systems and team skills against candidate use cases. When a use case is a knowledge assistant, read the page on RAG development for enterprise knowledge search. When it is a system that acts on your behalf, read the page on AI agent development services.

Before you hand any AI work to a vendor, run it through the AI vendor governance checklist and the twelve-check vendor scorecard. StepTo's own answers to common procurement questions are on the due diligence answers page. To turn your answers into a brief, use the software RFP template. For terms, see the pricing page, the staff augmentation page and the page on hiring AI developers.

Frequently asked questions

Is the StepTo AI readiness self-check a validated or certified assessment?

No. It is a self-scored list of sixteen questions written by StepTo. Each question counts one point, nothing is weighted and no benchmark data sits behind the score bands. StepTo makes no certification claim for it. Use it to decide what to fix first, and use an independent assessment if you need a measurement.

Does StepTo store my answers?

No. The score is computed in your browser and the ticks are not sent to StepTo or stored. The page only records that the check was used, never which rows you ticked.

Does StepTo run an AI readiness assessment for clients?

The StepTo AI strategy consulting page describes an AI readiness assessment that audits data quality and accessibility, existing systems and team skills, scored against candidate use cases. This self-check is separate from it and free to use. To talk about the assessment, start with a call. It starts with a 15-minute intro call. If there is a fit, a 60-minute discovery call follows.

What does StepTo publish if my score says I need engineers?

StepTo is a 15-20 engineer nearshore team in Belgrade (CET). Its published rate card is $25-85/hr by seniority. Staff augmentation starts from $4,500 per developer per month, with a 3-month minimum and 30 days' notice. The pricing page has the full terms.

Get in touch

Or reach us directly

Office hours

Monday – Friday09:00 – 18:00 CET
WeekendClosed

Sent on a weekend: we reply on the next business day.

Contracting party: StepTo d.o.o. Company details are on the Imprint page.

Send us a message

StepTo's Belgrade team replies within one business day.

It starts with a 15-minute intro call. If there is a fit, a 60-minute discovery call follows. Engineers can typically start within one to two weeks of first contact; a full team usually takes two to four weeks.

* required

Your details are handled as described in our Privacy Policy.

Performance-led engineering

Want senior engineers who move work forward, not just tickets?

Work with accountable, English-fluent professionals who communicate clearly, protect quality, and deliver with a steady operating rhythm. Cost efficiency matters, but performance is why clients stay with us.

Delivery signals · senior engineering team
Senior ownership
Lead-level
Delivery rhythm
Weekly
Timezone overlap
CET
1 teamaccountable for outcomes, communication, and execution