What an AI Readiness Assessment Costs — and When to Skip It

A paid AI readiness assessment in 2026 runs roughly $1,500–$3,000 for a short discovery review, $5,000–$15,000 for a comprehensive diagnostic, and $15,000–$50,000 or more once a strategic roadmap is attached. Free ones are everywhere too, and they cost you something different. We don’t sell any version of this — which is most of the reason we can be plain about it. Here’s what the market actually charges, what you should get at each level, and why the same money usually does more as one working report than as a document about your readiness to have one.

How much does an AI readiness assessment cost?

Two independently published 2026 pricing guides land on nearly the same bands. ConsultKit’s guide puts discovery-level assessments at $1,500–$3,000 for one to two weeks of work, and comprehensive assessments — a full diagnostic across your data, infrastructure, people, and governance — at $5,000–$15,000 over two to four weeks. Assessment-plus-roadmap engagements run $15,000–$50,000+ over four to eight weeks, with large-enterprise versions passing $200,000.

Aries Consulting Group’s guide corroborates: $2,000–$8,000 for a focused small-business audit, $5,000–$15,000 for mid-market companies (roughly $2M–$50M in revenue), and $15,000–$50,000+ for enterprise-grade work. In an unregulated market with no standard product, two unrelated publishers converging on the same numbers is about as solid as pricing data gets. If a quote lands far outside those bands, ask what’s different about your situation. Sometimes there’s a good answer. Often there isn’t.

Why is there a free version of the same thing?

Because free works — for the firm giving it away. A free assessment is a lead-generation tool, and the deliverable is built to surface problems the vendor’s paid product happens to solve. Aries’ own pricing guide describes the free version as “a sales discovery call wearing audit branding,” which matches what we’ve seen. Nobody staffs a consultant for a week at zero dollars out of curiosity.

That doesn’t make free assessments useless. It makes them sales meetings, and you should read the output the way you’d read any pitch. The real cost isn’t the price, it’s the steering. MIT Project NANDA’s 2025 report found more than half of generative AI budgets went to sales and marketing tools while the largest measured returns came from back-office automation — cutting outsourced processing, agency spend, and manual operations work. Vendor-led advice is part of how that mismatch happens. You get pointed at what’s easy to sell, not at what pays back.

What moves the quote?

Four things move an assessment price more than anything else, and all four are things you can estimate yourself before you talk to anybody:

• Number of systems. A business running on an accounting package, a CRM, and spreadsheets is days of work to audit. One with an ERP, a warehouse, five SaaS tools, and a legacy database is weeks.
• State of the data. If nobody can say where customer data lives or who owns it, the assessor has to reconstruct that map first — and you pay for the archaeology.
• Interview depth. A couple of leadership calls are cheap. Structured interviews across every department are where the hours pile up in the bigger engagements.
• Deliverable format. A written summary with a prioritized list costs less than a scored readiness model, an ROI workbook, and a board-ready roadmap deck. Decide which of those you will actually reread in six months.

What you should get at each level

Under $3,000: a scoped discovery

A systems and data inventory, interviews with one or two leaders, and a short written report naming your two or three best automation candidates with rough effort estimates. That’s it, and that’s fine. At this price you are buying prioritization, not a diagnostic.

$5,000–$15,000: the full diagnostic

Everything above, plus a data-quality and access audit, a security and governance review, integration notes on each core system, and per-use-case ROI estimates. If you are paying this much, insist on the governance piece: Deloitte’s 2026 State of AI in the Enterprise survey of 3,235 leaders across 24 countries found only about one in five organizations has a mature governance model for autonomous AI agents, and only 42% believe their strategy is highly prepared for AI at all.

$15,000–$50,000+: assessment plus roadmap

A multi-quarter implementation plan with sequencing, budget modeling, vendor shortlists, and change-management planning. In our judgment this tier earns its price once you are past roughly 100 employees, carrying real compliance obligations, or refereeing several departments that want the same budget. Below that it is usually overbuying — a map of terrain you could cross on foot in the time it takes to draw it.

Why we don’t sell one

We looked hard at putting a paid assessment on the front of this business and decided against it, for a reason that is more useful to you than it is to us: the readiness score does not predict what people think it predicts.

The companies that score well — clean exports, named data owners, someone in-house who likes this kind of work — are usually the companies that already built the report. A high score is often a sign you don’t need to hire anyone. The companies that score badly already know the data is a mess; paying $8,000 to be told so in a nicer font doesn’t move them any closer to a report that lands on Monday. Either way, the document is not the bottleneck.

The failure data says the same thing from the other direction. S&P Global Market Intelligence found the share of companies abandoning most of their AI initiatives jumped to 42% in 2025 from 17% the year before, with the average organization scrapping 46% of proof-of-concepts before they reached production. MIT’s figure is blunter: about 5% of enterprise generative AI pilots produced real revenue acceleration. Those companies did not fail for lack of assessments. They failed in the gap between the plan and something that runs every week without anyone babysitting it.

So what we sell instead is the thing on the far side of that gap. Standing Reports is $1,500 a month, month to month: one recurring report or workflow taken end to end — source connections, the transformation logic, scheduled delivery, monitoring with proactive fixes when something upstream changes, format changes included, and a monthly note on what it saved you. No setup fee. If the first correct report doesn’t land within 14 days of kickoff, month one is free.

Run the arithmetic against the bands above. The cheapest comprehensive diagnostic is about three months of that. The expensive end of the same band is ten. At the end of the diagnostic you have a document; at the end of ten months you have a report that has been landing correct every week, plus a very concrete map of where your data actually breaks — which is the finding the assessment was trying to sell you, learned the only way it’s ever really learned.

What you actually need before you automate one workflow

You can do the part of a readiness assessment that changes a decision yourself, for one workflow, in an afternoon. Five questions, in the order they matter:

1. Can the numbers leave the system? A tool with a CSV export or an API is a normal build. A system that will only show you a number on a screen — no export, no API, a PDF at best — is a different and much larger project. This single question moves a quote more than the other four combined.
2. Can you name a human per source? One person per system who can grant access and answer a question about what a field means. If you can’t name them, that is the finding, and it costs nothing to discover.
3. Are the rules written down anywhere? Every hand-built report carries logic that lives only in the head of the person who builds it: which jobs get excluded, which month a number belongs to, the one client billed differently since 2019. Writing that down is the cheapest hour in this whole process and the most expensive thing to reconstruct later.
4. Is the shape stable? A report with a steady format is cheap to keep running. One that grows a new column every time somebody asks a question in a Monday meeting isn’t a build, it’s a standing claim on someone’s attention — price it that way.
5. Who would notice if it were wrong? The person who assembles the report by hand is also, without anyone writing it down, the quality check. Automate the assembly without replacing the noticing and you have made the process faster and more confident at once. That is the failure mode we see most, and we broke it down in what breaks a weekly report automation.

That list is free, it takes an afternoon, and it produces the two or three findings you would actually have acted on. If you work through it and the answers are ugly, that is not a reason to buy a bigger assessment — it is the normal starting condition. We spent six years cleaning and connecting small-company data before we touched AI, and messy inputs are the job, not a disqualifier.

When a paid assessment is genuinely worth it

There are real cases. If you are past roughly 100 employees, if governance is itself the deliverable because a regulator or a customer’s security review will ask for it, if three departments are competing for one budget and somebody neutral has to arbitrate, or if you are mid-consolidation after an acquisition and nobody can draw the systems map — pay for the assessment. In those situations the document is the product, and it is worth the $5,000–$15,000.

Below that, the adoption data argues for starting small. The US Census Bureau’s Business Trends and Outlook Survey measured AI use among US businesses at 17–20% from December 2025 through May 2026 — 37% among firms with 250 or more employees, under 20% among firms with four or fewer. A Federal Reserve FEDS note from April 2026 puts firm-level adoption around 18% at the end of 2025, while noting that because large firms adopt first, roughly 78% of the labor force works somewhere that has. If you run a small company, you are not behind a pack of AI-native competitors your own size. You have time to do one thing properly.

And some workflows shouldn’t be automated at any price — the report nobody reads, the genuinely one-off analysis, the process whose rules haven’t actually been decided yet. Turning that work down is cheaper for everyone than a build that gets switched off in month three, which is why we keep a public list of what we won’t automate.

Frequently asked questions

Is a free AI readiness assessment worth taking?

As a free hour with someone who has seen a lot of businesses, sure — you may learn something. Treat the recommendations as a pitch, expect them to lean toward whatever the vendor sells, and don’t grant system or data access for it. Anything that requires real access to your systems should be a paid engagement with a contract behind it.

How long does an AI readiness assessment take?

Per ConsultKit’s 2026 market data: one to two weeks at the discovery tier, two to four weeks for a comprehensive assessment, and four to eight weeks when a strategic roadmap is included. For a small business, anything quoted in months is scope inflation rather than thoroughness.

Can I do the readiness work myself?

A good chunk of it. List your systems, note which ones share data, and ask each department lead for the three most repetitive tasks in their week — that is most of the discovery tier, for free. The parts that are genuinely hard to self-serve are technical: auditing data quality and access, and reviewing security and governance. Pay for those if you need them. Don’t pay someone to interview you about your own business.

Do you run an assessment before you build?

Not as a product you buy. The fit check is free and comes back in writing. Once you start, the first week of onboarding includes a systems-and-data inventory, because we can’t automate what we can’t reach — you just get it as a byproduct of a working report instead of as the whole deliverable. If it turns out your data genuinely can’t be automated, we say so then and refund the $500 deposit that held your slot — it’s credited against your first month otherwise, not an extra fee.

What do we own if we stop?

Everything — the connections, the logic, the outputs, the documentation. We build inside systems you already own, so there is no proprietary box you have to keep renting to read your own numbers. Month to month means month to month. We cap onboarding at four new clients a month, which is a real constraint rather than a sales line: there are two of us, and one has a day job.