How Do You Run an AI Readiness Check? A Practical Guide
Last Updated: September 2026
An AI readiness check is a short test of whether one named process could run with AI help next month. You look at the steps people really follow, the data they use, the person who answers for the result, and how you judge good work. It tests one workflow, not the whole business, and it ends in a decision.
AI Smart Ventures has guided growing businesses through AI adoption for over a decade, and the pattern rarely changes. Owners arrive holding a tool they already bought, with no process to point it at. Work that starts from one workflow you know well moves faster, because the team sees the before and the after.
Skip this step and the cost is rarely the software. It is the quarter spent defending a tool no one opens, the data cleanup you find halfway through, and the team that decides AI does not work here. A one-process check turns that risk into a short test you can repeat.
Key Takeaways
- Check one process, not the company: pick a single workflow that runs every week and test that, rather than scoring the whole business on a scorecard.
- Ready means data, owner, and output: the process needs inputs you can find, one owner who can act, and an agreed way to tell good work from bad.
- Write the standard before you buy: decide what a good result looks like while the tool choice is open, so the first pilot has something real to prove.
- Name the owner and what they may change: a person who cannot change a step or approve an exception is a contact, not an owner.
- Re-run the same measures after 30 days: readiness is a measure you repeat, and the second run tells you whether anything really moved.
Each point holds up because it covers one process. A score for the whole business averages your cleanest workflow with your messiest, and that average describes nothing you can fix on Monday.
What Does an AI Readiness Check Actually Measure?
An AI readiness check measures four things about one process: the steps, the data they use, the person who answers for the result, and the standard the output must meet. Everything else is context. You are asking a narrow question. Could this piece of work be done with AI help next month, without anyone losing sight of quality? That has a yes-or-no answer, and a scorecard does not.
The check works because it is bounded. You watch one real run from start to finish, note where each input comes from, and mark every point where a person makes a judgment call. Judgment points are where AI either helps or quietly does damage, so they matter more than the tools do. You finish with a short page that your team, a vendor, and an outside partner can all read the same way.
Why Check One Process Instead of the Whole Business?
Check one process because company-level scores hide the thing you need to fix. Publicis Sapient’s 2026 global AI report surveyed 1,550 AI decision-makers and was published on 17 June 2026. It found 73% say AI is used regularly or across most business processes, while only 10% say AI is core to how their business runs. A further 42% say AI is capable, but their own business is not set up to capture the value.
A score out of ten averages your cleanest workflow with your messiest, then hands you a number no one can act on. One process gives you control. You know the cycle time today, you know who complains about it, and you can take the same measure four weeks later. It also limits the damage. If the answer is no, you have spent a week on workflow optimization, not a budget cycle and a team’s goodwill.
How Do You Evaluate AI Readiness for One Process?
Evaluate AI readiness for one process in five passes: name it, time it, trace its inputs, name its owner, then define good output. Do all five on the same workflow in a week, using a real run rather than the version in the handbook. The written version is always tidier than the lived one, and the gap between them is where AI projects fail. Finish with a verdict: yes, no, or not yet.
Run the passes in that order, because each one narrows the next. Timing tells you whether the process is worth changing. Tracing inputs tells you whether the data is within reach. Naming the owner tells you whether a change can be approved this quarter. Teams that jump straight to tool demos end up with a shortlist and no rules for choosing. AI Smart Ventures observes that written rules settle most tool arguments before they start.
- Name the process: write one sentence covering what goes in, what comes out, and who receives it.
- Time a real run: record how long the work takes today and how often it happens.
- Trace every input: list where each piece of data lives and who can reach it.
- Name one owner: the person who can change a step without asking three teams first.
- Define good output: agree on the test a result must pass before anyone sends it onward.
Bring in help if the first pass stalls. AI Advisory is built for this work: choosing which process to test first, then reading the answer honestly.
Is Your Data Clean Enough for This One Process?
Your data is clean enough when the inputs this process uses are easy to find, current, and held in one format. That is a lower bar than a company data program, and the right bar here. dbt Labs’ 2026 report, published on 14 April 2026, drew on 363 data practitioners. Trust in data rose from 66% to 83% as a priority in a year, while unclear data ownership was named by 41% as a live blocker.
Test it directly. Take five recent runs and check whether every input can be found in under a minute, in one place, in the same format. If three of the five need a phone call, this process is not ready, and the repair is small and specific rather than a data program. Fixing five fields for one workflow is a week of work. Fixing everything is a year of it, and the AI project waits.
Who Owns the Process and Can They Change It?
The owner is the person who can change the steps, approve exceptions, and answer for the output. The check should name them on paper. Claimed owners are common; real power to act is rarer. Ivanti’s 2026 AI Maturity Report surveyed 3,900 people across six countries in February and March 2026. It found 85% of IT professionals name an accountable owner for every AI agent and workflow, while only 42% say accountability is actually clear.
The owner matters most on a bad day. When an AI-assisted step gets something wrong, a person has to notice, stop the work, and decide what the process does instead. If that person needs three sign-offs to change a template, the workflow is not ready, however clean the data is. Write down the owner’s name, the change they may make alone, and who they call next. Their reaction tells you more than the page does.
How Do You Measure AI Readiness After the Check?
Measure AI readiness by running the same test again after 30 days. Compare four numbers: cycle time, rework, the share of runs that finish without a handover, and how many people chose to use the new step. Write those numbers down before the pilot starts, so the comparison is fair. They are workflow measures, not adoption dashboards, and they move within a month. If nothing moved, this process was never the constraint.
The second run is where change management shows up. Kyndryl’s 2026 People Readiness Report asked 1,100 business and technology leaders across eight countries. It found 77% have already scaled generative AI across several functions, while fewer than a third have fully established change management programs. Capability building closes that gap. The training has to name the workflow: short AI upskilling tied to this one process gets used, and a general tour of tools does not.
What Makes an AI Readiness Check Fail?
An AI readiness check fails when it scores the business instead of testing a process, when it runs after the software is bought, and when no one signs the verdict. Scope is the other common failure. A review covering eleven workflows gives you a report, and a report is not a decision. Tool-first AI agencies often start from the platform and work backwards, which is quick to sell and slow to use.
Two smaller mistakes cost more than they look. The first is testing a process that runs twice a year, where you learn nothing before the next budget round. The second is treating AI literacy as a training problem for later. If the people running the work cannot say what the tool is doing, they will not catch the day it is wrong. You can avoid both in the first hour.
Frequently Asked Questions
How do you check AI-generated content?
Check AI-generated content against the sources it claims to use, not against how well it reads. Open the record, page, or link behind each fact and confirm it says the same thing. Then check tone and any number a reader might act on. Give one named person sign-off, and keep a short written standard so two reviewers reach the same verdict.
What are the five pillars of AI readiness?
The five pillars most frameworks name are strategy, data, technology, people, and governance. They make a useful checklist and a poor decision tool: a business can score well on all five and still have no process ready to change. Use the pillars to spot a talk no one has had, then test one workflow against them. A pillar matters here only if it blocks the process you picked.
What is the 30% rule in AI?
The 30% rule is a rule of thumb rather than a law. If an AI-assisted step does not save roughly 30% of the time a task takes today, the checking and upkeep often cost more than the savings. No standards body sets that number. Treat it as a bar you agree on before the pilot, then measure the real before and after on one process.
Is there a free AI readiness check you can run?
Yes. The version described here costs nothing but time: one workflow, one week, five passes, and a written yes, no, or not yet at the end. Free online scoring tools return a number in ten minutes, which is fine for starting a talk and weak as evidence. If you want a template, put the five questions on one page and reuse it.
What questions should an AI readiness check include?
Ask what the process produces and who receives it, and how long a real run takes. Then ask where each input lives, who may change the steps, and how you would know the output was wrong. Also ask what happens today when it is wrong. Six questions answered about one workflow, with evidence, beat a fifty-item form filled in from memory by one leader.
How long does an AI readiness check take?
A single-process check takes about a week of part-time work. That covers one session watching a real run, a day tracing the inputs, a short talk with the owner, and an hour to write the verdict. Broader reviews take longer because they wait on other people’s calendars. If a check runs past three weeks, the scope has usually grown from one process into several.
Can a business be unready for AI?
Yes, and it is usually short-lived and narrow. A business is unready when the process is not written down, its inputs sit in five systems no one agrees on, or no single person may change the steps. None of that is permanent. Repair the inputs for one workflow, name the owner, and the same business is ready for that process. It stays unready for the others.
How do you start an AI readiness check?
Start by picking a process that runs at least weekly, matters to someone outside your team, and irritates the people who run it. Watch one real run this week and write the five answers down. Scope drives what comes next: some processes need a data repair, others need AI training first. Schedule a consultation for a second pair of eyes on the first one.
Executive Summary
An AI readiness check answers one question: could this named process run with AI help next month? You test the steps, the data they use, what the owner may change, and the standard the output must meet. A score for the whole business leaves nothing to act on. Run the five passes in a week, write a yes, no, or not yet, then repeat the same measures after 30 days. Readiness is a decision you make one workflow at a time.
What Should You Do Next?
Pick the process this week: one that runs weekly, has a named owner, and produces something a colleague or client receives. Watch a real run, note where every input comes from, and record how long it takes. Then decide whether the gap is data, owners, or skills, before you look at a single tool.
AI Smart Ventures offers AI Advisory for growing businesses deciding where AI belongs first. Schedule a consultation to turn your first readiness check into a sequenced plan.
People Also Read
About the Author
Nicole A. Donnelly is the Founder of AI Smart Ventures and an AI Adoption Specialist with 20 years of experience as a founder and CEO and over a decade leading AI adoption initiatives. She helps businesses integrate artificial intelligence with clarity and confidence, driving innovation and sustainable growth. Nicole has trained over 20,217 professionals in Applied AI, delivered 624 workshops, and worked with close to 1,000 organizations across diverse industries.
Expertise: AI Transformation, AI Strategy, AI Implementation, AI Adoption, Applied AI, Marketing, Business Operations
Disclaimer: This content is for informational purposes only and does not constitute professional business or technology advice. Results vary based on industry, existing systems, and implementation commitment. Contact AI Smart Ventures for a consultation regarding your specific situation.
