How Can I Assess My Organization's AI Readiness?

You can assess your organization's AI readiness yourself in about two weeks, without a consultant or a software trial, by starting from the work rather than the technology. List the work that repeats, pick the three pieces that take the most time, count and time them for two weeks, check whether the information they need lives in a system, and then put a price on each one. That last step matters most, because it tells you the most you should spend before you've spent anything.

Most guides stop one step short of that. We read the 13 readable pages ranking for this question and its close variants in September 2026. Nearly all of them described readiness as a set of areas to rate, such as data, infrastructure, skills, governance, and leadership. One told readers to set a baseline and a target before implementing anything, and one said a score should arrive before the budget is committed. None showed how to turn what you measure into a dollar figure you could hold a purchase against.

This page is the do-it-yourself walkthrough for an owner or operations lead at a business of roughly 10 to 200 people. It's the companion to our guide to AI readiness assessment services, which covers what a full assessment examines and how an outside one runs.

Key Takeaways

• You can assess AI readiness yourself in about two weeks and roughly 12 hours of one person's time, plus a few minutes a day from the people who do the work (our estimate).
• Start with the work, not the tools. In a study of about 200,000 real AI conversations, office and records work scored three to eight times higher for AI applicability than trades and grounds work (Microsoft Research, 2025).
• Measure, don't estimate. Count how often each workflow happens and time five real instances, because the number people remember is rarely the number on the clock.
• Be modest about savings. Workers who used generative AI said it saved them 5.4% of their work hours on average, about 2.2 hours in a 40-hour week (Bick, Blandin, and Deming, NBER, 2025).
• Before investing, set a spending limit for each workflow: its monthly cost times the share you expect AI to remove, plus any losses it would stop, times 12. If a proposal costs more than that, it has to find the difference somewhere you can name.

Table of Contents

• How can I assess my organization's AI readiness?
• Step 1: Name the owner and the decision
• Step 2: List the work that repeats
• Step 3: Shortlist three workflows
• Step 4: Count and time them for two weeks
• Step 5: Check what each workflow would need
• Step 6: Price the work and set a spending limit
• Step 7: Decide, one workflow at a time
• A worked example: a 45-person landscaping company
• Where a self-assessment runs out
• Where Avolis fits
• Frequently Asked Questions
• Continue Learning

How Can I Assess My Organization's AI Readiness?

Assess it one workflow at a time, in seven steps, over about two weeks. The steps move from a list of everything that repeats to a priced shortlist of three workflows, and each one ends with a go, fix-first, or not-now decision. You'll need one person to own it, a way to count, and a stopwatch.

The order matters more than any single step. Most businesses start at the end, with a tool someone saw demonstrated, and work backwards to a reason. Starting from the work means the tool question arrives last, when you know what the answer is worth.

Step 1: Name the Owner and the Decision

Give the assessment to one person, and write down the spending decision it's meant to settle. In most businesses this size the right owner is whoever runs operations, because they know where the work goes and they aren't the one most excited about AI. The owner of the business should set the question, then step back.

A useful question is concrete and has money attached: "Should we spend anything on AI in the office this year, and if so, on which workflow and up to how much?" A vague one ("are we ready for AI?") can't be answered, so it never gets answered. Writing the question down before anyone has measured anything also keeps the result honest, since it's much harder to move the goalposts once they're on paper.

Step 2: List the Work That Repeats

List every piece of work that happens at least weekly and involves typing, copying, looking something up, or writing something for someone else. Don't list tasks by department or by software. List them the way the people doing them would describe their week.

The quickest way to build the list is to ask each person who works in the office, and each field lead who does paperwork, one question: "What did you do more than once last week that involved your keyboard or your phone?" Ask about last week specifically, not a typical week, because people describe a typical week the way they wish it went (why assessors ask about the last time). At around 45 people, expect somewhere between 10 and 20 items. Our pillar on what an assessment should examine explains why the work comes before data and systems, so we won't repeat the argument here.

Step 3: Shortlist Three Workflows

Pick the three items that happen most often, follow roughly the same steps each time, and consist mostly of reading, writing, or moving information. That last filter is the one people skip, and it's the one the evidence supports most clearly.

Researchers at Microsoft classified the work activities in about 200,000 anonymized conversations people had with the company's Bing Copilot assistant in 2024. They found that the most common and most successful AI-assisted activities "involve information work," meaning creating, processing, and communicating information. Then they scored every occupation group on how much of its work AI could help with (Tomlinson, Jaffe, Wang, Counts, and Suri, Working with AI, Microsoft Research, 2025). The spread inside a single operations business is wide.

Information and record clerks scored 0.33 and grounds maintenance workers scored 0.04, and the study notes that the lowest scores go to occupations that require physically working with people, operating machinery, and other manual labor. For your shortlist, that means the dispatcher's desk, the bookkeeper's queue, and whoever writes the reports are better places to look than the crews themselves. It doesn't mean field work has nothing to offer, since most field jobs carry paperwork, but the paperwork is where the AI part lives.

A second study adds a twist worth knowing. A nationally representative survey of US workers found that office and administrative support occupations had "low adoption relative to predicted exposure," while managers had high adoption relative to theirs (Bick, Blandin, and Deming, The Rapid Adoption of Generative AI, NBER Working Paper 32966, revised February 2025). In other words, the people whose work suits AI best are often the ones using it least, and the person most enthusiastic about AI in your business may not be the person whose work would benefit.

Step 4: Count and Time Them for Two Weeks

For each of the three workflows, count how many times it happens over two weeks, time five recent real instances with the person who does them, and note every miss, error, or callback along the way. This is the step that separates an assessment from an opinion.

Counting works best from a system when the system records it (invoices sent, tickets closed, reports filed). When it doesn't, a tally sheet on the desk works fine, and it takes the person doing the work a few seconds per instance. For timing, pick five real instances from the last two weeks, not the easiest ones, and have the person walk through each one start to finish while someone keeps time. Five is a small number, but it's usually enough to see whether a task takes six minutes or forty, which is the distinction that matters.

Misses are the number most businesses forget to collect, and they're often worth more than the time. A report that goes out late is a cost. So is an invoice that never gets sent, or a quote that has to be redone because the first one used last year's prices, and every one of those can be counted. Our guide on how businesses assess readiness for AI adoption covers what to measure if you later trial a tool against these numbers. For a full list of the questions an outside assessor would ask each person, see how consultants assess AI readiness in businesses.

Step 5: Check What Each Workflow Would Need

For each shortlisted workflow, answer four questions. Where does the information it needs live? Does it follow the same steps each time? Who owns it? And what happens if the output is wrong? You're not scoring the whole company here, only checking whether these three workflows could take AI as they stand.

Paper records or no clear owner usually mean the first project is a process fix rather than an AI project, and that's a legitimate finding. For the full 36-question version of this check, use our AI readiness checklist. If your gaps are in logins, connections, or where your systems live, our infrastructure readiness guide sorts which of those actually block AI. Why one failed check should stop a workflow, rather than be averaged away by strong answers elsewhere, is covered in our assessment methodology.

Step 6: Price the Work and Set a Spending Limit

Put a monthly dollar figure on each workflow, then turn it into a spending limit: the most a first year of AI on that workflow should cost to pay for itself. This is the "before investing" step, and it's the one that turns a readiness assessment into a purchasing decision.

The arithmetic is simple enough to do on the back of the tally sheet.

1. Monthly hours. Instances per month times minutes per instance, divided by 60.
2. Monthly cost. Monthly hours times the hourly wage of the person doing it. Use their actual wage, or the Bureau of Labor Statistics median for the role if you'd rather not work from payroll. For May 2025, BLS puts bookkeeping clerks at $24.36 an hour and administrative assistants at $22.86 (BLS Occupational Employment and Wage Statistics for bookkeeping clerks and administrative assistants). That's before benefits, so it understates the real cost, which makes the limit conservative.
3. Measured losses. Add anything the misses cost you each month, such as work that was never billed.
4. Expected share. Decide what share of the time you expect AI to remove. We plan with a third for a single, well-matched workflow and then test it, and that figure is our planning assumption, not a research finding.
5. Spending limit. Monthly cost times the expected share, plus any losses you expect it to stop, times 12.

The reason to start modest is what workers themselves report. In the same national survey, people who used generative AI at work estimated it saved them 5.4% of their work hours on average, about 2.2 hours a week for someone working 40. Across all workers, including those who didn't use it, the saving came to 1.4% of hours (Bick, Blandin, and Deming, November 2024 survey wave).

Those are self-reported figures for general use across a whole job, so they aren't a ceiling for one workflow built around AI. A whole-job average also spreads the saving across work AI never touches: two hours a week concentrated on one 23-hour-a-month workflow would already be more than a third of it. If you'd rather plan from a lower floor, rerun the arithmetic at 10%, and the visit-report limit in the example below drops from about $2,130 to about $640. And self-reported savings can run high, since nobody timed them. What the figures offer is a reality check. When a vendor promises to save half your team's time, the national evidence says the typical user reports a saving closer to two hours a week. A spending limit built from your own measured hours is how you tell the difference between those two numbers before you've paid for either.

Step 7: Decide, One Workflow at a Time

Give each of the three workflows one of three answers. Go small means the work is measured, the information is in a system, someone owns it, and the spending limit comfortably covers a modest first step. Fix first means the work is worth money but something in Step 5 blocks it, usually paper records or no owner. Not now means the volume or the savings are too small to justify the effort, or the work depends on judgment nobody has written down. These map onto the go, wait, and no calls in our guide to assessment services, split so that a process fix gets an answer of its own.

"Not now" isn't a failure. It's the answer that protects you from the projects that go wrong for predictable reasons, and our analysis of why AI projects fail shows how often those reasons trace back to skipping this kind of check.

A Worked Example: A 45-Person Landscaping Company

Here's the full walkthrough on one business. It's a composite we've built for illustration, not a client: a 45-person commercial landscaping company with 36 people in the field and 9 in the office, maintaining about 40 commercial properties. The wages are BLS May 2025 medians, and every other figure below is illustrative.

The question. The owner had been pitched two AI products in a month and asked the operations manager to answer one question: "Should we spend anything on AI in the office this year, and on what?" The operations manager owned the assessment.

The list and the shortlist. Asking each office person and crew lead about last week produced 16 recurring items. Three passed the Step 3 filter: the monthly visit reports each commercial client gets, turning crew leads' extra-work tickets into invoices, and rebuilding the crew schedule after rain days.

The count. Over two weeks, the operations manager got these numbers.

The check and the price. The visit reports passed every Step 5 question, and at a third of the time, their spending limit came to $533 times one-third times 12, or about $2,130 for the first year. That's enough to justify a subscription tool tested properly, and nowhere near enough to justify a custom build. The extra-work tickets failed the first question, since crew leads wrote them on paper, and that failure turned out to be the most valuable finding of the two weeks. At an illustrative average of $380 per extra-work invoice, last month's five unbilled tickets were worth about $1,900 in work that had been done and never charged, more than four times the entire monthly cost of processing the tickets. The rain-day rebuilds depended on knowledge that lived only in the supervisor's head.

The decision. The visit reports became a go small, with a trial capped well under the spending limit. The tickets became a fix first: crew leads now enter extra work on their phones in the scheduling app the company already pays for, which is a process change, not an AI project, and the bookkeeper matches tickets to invoices weekly. Once the tickets are digital, they'll be worth assessing again. The rain-day schedule became a not now, with a note to get the supervisor's rules written down over the winter.

The owner turned down both AI pitches. The biggest dollar figure in the assessment came from a process fix, and the only workflow ready for AI was worth about $2,000 a year, which is a useful thing to know before a vendor quotes you ten times that.

Citation-ready summary: You can assess your organization's AI readiness yourself in about two weeks by working one workflow at a time. List the work that repeats, shortlist the three most frequent information-heavy workflows, count and time them for two weeks, and check whether each one's information lives in a system and has an owner. Then set a spending limit for each: monthly cost times the share of time you expect AI to remove, plus any losses it would stop, times 12. Workers who use generative AI report saving 5.4% of their hours on average, so plan modestly.

Where a Self-Assessment Runs Out

A self-assessment works well when the work sits with one or two people and the owner can find a dozen hours to run it. It runs out in three situations. The first is when the expensive work crosses several people or systems, because nobody inside sees the whole handoff. The second is when nobody has time to count, since an assessment with guessed numbers is worse than none; it gives a guess the authority of a measurement. The third is when the answer will justify real money, and an outside view is worth paying for before a five-figure build.

If you're at that point, the next question is who should do it. Our sibling pages cover how consultants assess AI readiness in organizations, including leadership and team readiness, and how consultants assess AI readiness in businesses, including how to tell an assessment from a survey. For what outside assessments cost and the free and subsidized routes, see AI readiness assessment consulting firms. If you're comparing firms, start with which AI consulting company to choose.

Where Avolis Fits

Our diagnostic follows the same logic as this page, at more depth and across more of the business. We work through each workflow in scope with the people who run it, count how often it happens, time real instances, map where the work stalls between people and systems, and then rank what's worth building. Every engagement starts with that diagnostic, and the ranked result is yours whatever you decide next.

You don't need us to do the seven steps above, and we'd suggest doing them first. If the count turns up workflows that cross several people and systems, or a spending limit that justifies a real build, that's when a second set of eyes is worth it.

Frequently Asked Questions

How can I assess my organization's AI readiness?

Work one workflow at a time. List the work that repeats weekly, shortlist the three most frequent information-heavy workflows, count and time them for two weeks, and check whether their information lives in a system and has an owner. Then price each one and decide whether to go small, fix a process first, or wait.

How do businesses assess AI readiness before investing?

They measure the work and set a spending limit before buying anything. Multiply a workflow's monthly hours by the wage of the person doing it, take the share of time you expect AI to remove, add any losses it would stop, and multiply by 12. A proposal that costs more than that limit needs a clear reason.

How long does a DIY AI readiness assessment take?

Plan on about two weeks from start to decision, and roughly 12 hours of the owner's time by our estimate. Most of it goes on counting and timing three workflows, which needs two weeks of real volume to mean anything. Everyone else spends a few minutes a day on a tally sheet.

What should I measure in an AI readiness self-assessment?

Measure three things for each shortlisted workflow: how often it happens, how many minutes a real instance takes, and how often it goes wrong, such as work that is never billed or reports that go out late. Count from your systems where you can, and use a simple tally sheet where you can't.

When should I bring in outside help for an AI readiness assessment?

Bring in outside help when the expensive work crosses several people or systems, when nobody inside has time to count and time it properly, or when the result will justify a significant build. For smaller, single-person workflows, a self-assessment usually gives you enough to make a sound decision on your own.

Continue Learning

Assessing AI readiness yourself comes down to measuring the work before anyone measures a tool. The spending limit at the end is the part most businesses never calculate, and it's the part that makes every later conversation with a vendor easier.

Before you take the next AI pitch, time the work it claims to change.

How assessments work:

• AI readiness assessment services
• AI readiness assessment methodology
• AI readiness assessment checklist and PDF
• How businesses assess readiness for AI adoption

Assess it with help:

• How consultants assess AI readiness in organizations
• How consultants assess AI readiness in businesses
• AI readiness assessment for SMBs

Choosing who does it:

• AI readiness assessment consulting firms
• Which AI consulting company should I choose?
• Why do AI projects fail?

---

Sources

All sources retrieved 2026-09-24.

• Kiran Tomlinson, Sonia Jaffe, Will Wang, Scott Counts, and Siddharth Suri, Working with AI: Measuring the Applicability of Generative AI to Occupations, Microsoft Research, arXiv:2507.07935, submitted July 2025, retrieved 2026-09-24 — https://arxiv.org/abs/2507.07935
• Alexander Bick, Adam Blandin, and David J. Deming, The Rapid Adoption of Generative AI, NBER Working Paper 32966, September 2024, revised February 2025, retrieved 2026-09-24 — https://www.nber.org/papers/w32966
• US Bureau of Labor Statistics, Occupational Employment and Wage Statistics, May 2025 national estimates (median hourly wage, SOC 43-3031, 43-6014, and 37-1012), via the BLS Public Data API, retrieved 2026-09-24 — https://www.bls.gov/oes/current/oes433031.htm, https://www.bls.gov/oes/current/oes436014.htm, https://www.bls.gov/oes/current/oes371012.htm

On the page review. "The 13 readable pages" are those returned for "how can i assess my organization's ai readiness" and "how to assess business readiness for ai before investing" on 2026-09-24: HBS Online, Prosci, Knack, Leapsome, Microsoft (two pages), OvalEdge, Fullstack, Mimacom, Authentic, Symmetria Partners, 305 Spin, and Agility at Scale. A third Microsoft result is a self-assessment tool rather than an article and isn't counted. 305 Spin recommends setting a baseline and a target before implementing, and OvalEdge says a score should arrive before the budget is committed. Udemy Business and SotaTek were blocked to automated retrieval and aren't counted. It's a snapshot of one day's results, not a market survey.

On the studies. The Microsoft study measures which work activities AI helps with in real conversations with one free assistant. It isn't a measure of how much time AI saves. The Bick, Blandin, and Deming time savings are self-reported by workers across their whole job, not measured on a single workflow. We cite both for direction, and neither is a study of small operations businesses.

On the time estimates and planning share. The hours in the steps table and the "plan with a third" share are Avolis's own planning figures, not research findings.

On first-party claims. Statements about "our diagnostic" describe Avolis's own engagements. They are not independent research. The landscaping company is an illustrative composite, not a client. Its wages are BLS medians and all other figures are illustrative.

---

About Avolis Research Group

Avolis Research Group is Avolis's in-house research practice, focused on how operations-heavy small and mid-sized businesses actually adopt AI. It synthesizes primary economic research, government survey data, and results from real implementations into practical, vendor-neutral guidance.

More about Avolis and how we work · Get in touch