Is My Business Ready for AI? A Ten-Minute Test

AI readiness means one specific process passes five checks you can run in ten minutes using your own records and people. If you are asking “is my business ready for AI,” test one process: a company is never ready as a whole.

A typical online readiness quiz asks you to rate your own feelings across about ten questions and returns one company percentage. This test asks you to open files, count rows and name a person who will review the work. Iron Goo set the pass lines, estimated times, score bands and fix order as practical rules for this self-assessment.

According to the US Census Bureau's Business Trends and Outlook Survey national data, 24.6% of US employer businesses said they used AI in any business function during the two weeks from 21 September to 4 October 2026, up from 17.2% in the November 2025 reading. The latest reading was published on 22 October 2026. If that creates pressure to buy, read the case against rushing into AI before booking a demo.

Pick one process in one minute

One process gives the test a defined start and finish. Pick a recurring job you might hand to an AI tool, then write both ends on one line: “Supplier invoice arrives” to “draft bill is ready for approval,” for example. If you cannot name the finish, record the written-instructions check as a fail and define the finish before continuing.

Choose a process whose records and people you can reach now. The ten-minute estimate assumes the relevant person is available to confirm ownership and that someone has already timed an item; if either answer needs a later conversation or measurement, record the check as a fail for now and retest when you have the evidence.

Run the five checks

The five checks test whether you can describe, measure and assign one process using evidence you already have. Set a timer and record a pass or fail for each check on the score sheet.

1. Find written instructions: two minutes

Open the written instructions for the process without asking anyone where to find them. Pass: you find them within two minutes and they include at least one exception rule, such as what to do when a required purchase order number is missing. Fail: the process lives in one person's head, you cannot find the instructions, or the document describes only the easy case. RAND's report on AI project failures, based on interviews with 65 experienced data scientists and engineers, states: “Misunderstandings and miscommunications about the intent and purpose of the project are the most common reasons for AI project failure.” Written instructions give you a way to state the job before someone builds for it.

2. Export last month's records: three minutes

From your own login, export last month's records for the process to CSV or XLSX, then open the file. Pass: one file draws from one system, or two at most, and has a row for each item plus the fields the process uses. Fail: the needed records are scattered through inboxes, paper, chat threads or a personal spreadsheet, or obtaining the export requires IT, a vendor ticket or a higher plan. Gartner predicted that at least 30% of generative AI projects would be abandoned after proof of concept by the end of 2025, naming poor data quality as one of four reasons.

For an export in spreadsheet form, use the AI-ready spreadsheet test to check its structure. An accessible export is a starting point for an assistant that uses records during a task; it does not mean you need years of data to train your own model.

3. Count volume and find a timed duration: one minute

Count the item rows in the export and find out whether anyone has timed the work on one item. Pass: you have a monthly count from a record and a minutes-per-item figure obtained with a timer. Fail: the answer is “a lot,” a guessed duration or a count nobody can trace to a record. Without both figures, you have no measured starting point against which to compare a later change.

In a hypothetical example, 400 items per month at five measured minutes each take 2,000 minutes, or about 33 hours, by our arithmetic. In a Danish study of workers and workplaces, AI adopters reported saving about 3% of work hours; a change of that size can be difficult to spot without a before number. If the export failed, use another existing record for the count if one is available; do not guess to fill the cell.

4. Record the last error and its cost: two minutes

Write down the last mistake in the process, its cost in money or minutes, and the step that catches mistakes today. Pass: all three are written down. Fail: you cannot name the mistake or its cost, cannot identify the checking step, or the only answer is “we never make mistakes.” Ask the person doing the work if you do not know, and mark the check as failed until you have the answer.

Iron Goo's reasoning is that AI output can contain mistakes too, and the cost of an error should determine how much human review remains in the process. A correction that takes two minutes calls for a different review decision from an error that triggers a costly refund. The record gives the future reviewer something concrete to watch for.

5. Name an owner: one minute

Write the name of one person who can judge whether the output is right, has time each week to check it, and has agreed to do so. Pass: that person has agreed and their name is on the sheet. Fail: the answer is a department, “the team,” or your own name when you will not actually review the output. RAND recommends that leaders prepare to commit a team to one specific problem for at least a year; a named reviewer gives this process a person responsible for checking its output.

Copy the score sheet

The score sheet records the evidence behind each pass or fail. Copy it into a document and replace the placeholders as you run the checks.

What does your score mean?

A score of five makes the process a candidate for a fuller readiness check, not a guarantee of savings or project success. A score of three or four means close the failed checks; a score of zero to two means pick a different process.

With five passes, use the full AI readiness standard for the afternoon-long assessment before a vendor demo. You can also check what an AI implementation costs when planning your budget. With three or four passes, use the fix order that follows before buying anything.

With zero to two passes, pick another process using the guide to choosing your first AI use case. Keep the completed sheet: it tells you what was missing if you return to the original process later.

Close failed checks in order

The fix order differs from the check order and starts with the owner; among your failed checks, fix the one highest on this list first. Iron Goo's order moves from a conversation toward a possible system change; the first four fixes can be made without spending money, although they still take staff time.

1. Owner: Ask one person who can judge the output to take responsibility for a weekly check. Record their agreement.
2. Written instructions: Write the start, finish and exception rules with the person who performs the work. Put the instructions where you can open them without asking.
3. Volume: Find a record to count and time one item. Write down both figures and where they came from.
4. Error cost: Ask the person doing the work about the last mistake, its cost and the current catching step. Record the answers.
5. Data export: Try to obtain the needed records in an accessible file. If the obstacle calls for a system migration, picking a different process is often cheaper.

After each fix, rerun the same five checks; allow ten minutes for the retest. Once a process has five passes and has completed the fuller readiness assessment, Iron Goo's operations service can automate the work while keeping the baseline available for comparison.