AI Strategy

AI Readiness Assessment: Cost, Dimensions & Who Needs One

Almost everyone is using AI now. Few are making money from it. A readiness assessment tells you which side of that line you are on, before you write the check.

Get an Honest Assessment

By Elevated Signal Research Team · May 29, 2026 · 12 min read · Updated October 12, 2026

Key takeaways

1. Nearly nine in ten organizations use AI in at least one function, but only 37% see any profit impact and about 6% are high performers (McKinsey, 2026). Adoption is easy. Readiness is rare.
2. BCG's 2025 survey of 1,250 executives found 60% of companies report minimal value from AI; Gartner found at least half of generative AI projects were abandoned after the proof of concept by the end of 2025. The cause is rarely the model.
3. Only 7% of enterprises say their data is completely ready for AI (Cloudera / Harvard Business Review Analytic Services), and Gartner predicts organizations will abandon 60% of AI projects unsupported by AI-ready data through 2026.
4. Four tiers: free vendor self-assessments, independent engagements, large consulting firms, and assessments folded into a build. Firms quote per engagement. Price tracks data complexity, not brand, and a small independent firm can beat a big name.
5. The single most expensive mistake is not skipping the assessment. It is buying a vanity maturity score that produces a slide deck instead of a costed, owner-assigned roadmap you can actually build from.

An AI readiness assessment is a structured diagnostic that scores whether your organization can actually deploy and sustain artificial intelligence in production, across strategy, data, infrastructure, talent, governance, and culture. It does not ask whether AI is exciting, or whether your competitors are doing it. It asks one question: can you ship it.

Almost everyone uses AI now, and far fewer get much out of it. McKinsey's 2026 survey of 1,719 respondents in 97 countries found nearly nine in ten organizations use AI regularly in at least one business function. Only 37% said AI had added anything to profit, and about 6% qualify as high performers, getting at least 5% of operating profit from AI.

BCG puts the same gap more bluntly. Its 2025 survey of 1,250 senior executives found 60% of companies report minimal revenue and cost gains from AI, and only 5% get substantial value at scale. The distance between "we use AI" and "AI works for us" is the entire reason readiness assessments exist.

We build and run AI systems for clients, so below we say plainly which parts of an assessment are worth paying for and which you can do yourself in an afternoon.

Definition

What is an AI readiness assessment?

An AI readiness assessment scores an organization across six dimensions, strategy, data, infrastructure, talent, governance, and culture, to measure whether it can move AI from idea to production and keep it running. The deliverable is a maturity score plus a prioritized, costed roadmap of the gaps that would otherwise sink the first real project.

It is not the same as AI strategy. Strategy decides what to build and why: the use cases, the sequencing, the expected return. Readiness decides whether you can build any of it without the project collapsing the moment it leaves the sandbox. Readiness sits before strategy and before any build. A company can have a brilliant strategy and a data warehouse nobody trusts.

The most useful reframe comes from BCG's 2024 research: AI leaders put about 70% of their effort into people and processes, 20% into technology and data, and only 10% into the algorithms themselves. So readiness is mostly a business problem, and five of the six dimensions an assessment scores have nothing to do with which model you pick.

A good assessment ends with an order of work. A score of "3.2 out of 5" tells you nothing you can act on. "Your data is fine, your governance is a liability, and your first use case should be the one in customer support because that is where you already have clean labeled data" tells you exactly what to do Monday morning.

Hold any assessment to that standard. A score nobody acts on is the most common way an assessment wastes its fee.

The failure pattern

Why do most AI projects stall after the pilot?

Pilots succeed because they are hand-held: a small team, clean sample data, no compliance review, no integration with real systems. Production is where all of that catches up. Gartner found that by the end of 2025 at least 50% of generative AI projects had been abandoned after the proof of concept, undone by poor data quality, weak risk controls, escalating costs, and unclear business value.

The failure numbers measure different things over different windows, and they all point the same way. S&P Global found the share of companies abandoning most of their AI initiatives jumped from 17% to 42% in a single year, with 46% of proofs of concept scrapped before production (late-2024 survey). RAND's 2024 study, built on interviews with 65 data scientists and engineers, cites estimates that more than 80% of AI projects fail.

A widely cited MIT NANDA report (2025) says 95% of organizations got zero return from generative AI, but it rests on interviews at 52 organizations and a survey of 153 leaders, so treat it as a contested upper bound rather than settled fact. The measured figures put it at roughly half, and organizational readiness, data first, is the consistent cause.

None of Gartner's causes (poor data quality, inadequate governance, runaway cost, fuzzy business value) is about the model. Each is a readiness failure an assessment would have flagged before the first dollar was spent. Gartner separately warned that through 2026, organizations will abandon 60% of AI projects unsupported by AI-ready data, which lines up with Cloudera's finding that only 7% of the data leaders it surveyed call their data completely AI-ready.

Readiness also predicts how long projects last. Gartner found that 45% of organizations with high AI maturity keep their AI projects operational for at least three years, versus 20% for low-maturity ones. In other words, readiness predicts whether what you build this year is still running in three.

The six dimensions

What does an AI readiness assessment actually evaluate?

Frameworks differ in how they slice it. Cisco uses six pillars, Microsoft seven (including model management), Gartner seven, but they agree far more than they differ. All converge on the six dimensions below, and for each one the useful question is whether it will hold up in a real deployment, not whether you have it.

Strategy & use-case alignment

Is there a defined business problem AI is meant to solve, with an owner and a number attached? Cisco's 2025 index found 99% of its top-tier "Pacesetters" have a well-defined AI strategy, against 58% of all companies surveyed. The common failure is starting from the technology ("we should use AI") instead of the problem.

Data readiness

Is the data accessible, accurate, governed, and labeled well enough to ground or train a model? Gartner lists data availability and quality among the top barriers to getting AI into production, and only 7% of data leaders in Cloudera's survey call their data completely AI-ready. Gartner also stresses that data is ready only relative to a specific use case, with the patterns, errors, and outliers that use case needs.

Technology infrastructure

Can your environment run, deploy, and monitor models? Cloud and compute, the integration layer, MLOps, and the path from a notebook to a production endpoint. Plenty of pilots work on a laptop and then have nowhere to live. In Gartner's late-2023 survey, only 48% of AI projects made it into production on average, and going from prototype to production took about eight months.

Talent & data literacy

Two separate questions. Do you have the technical people to build and maintain AI, and does the broader organization have the literacy to use it without being fooled by it? A model nobody trusts and nobody knows how to interrogate gets shelved regardless of how good it is.

Governance & risk

Are there policies for acceptable use, model risk, data privacy, and regulatory compliance? "Inadequate risk controls" is one of Gartner's four named reasons projects get killed. In regulated industries this is not optional, and a weak score here is what turns a promising pilot into a legal review that never ends.

Culture & change capacity

Will the organization actually adopt what you build? AI changes workflows, and workflows are political. Cisco found 91% of Pacesetters have a full change-management plan for AI, against 36% of all companies. Culture is the soft dimension that hard ROI numbers depend on.

Five of the six dimensions have nothing to do with which model you pick. The teams that get past the pilot usually stop adding pilots and fix the data first, before a model is even chosen.

Cost

How much does an AI readiness assessment cost?

The market splits into four tiers, and price tracks usefulness only loosely. Free vendor self-assessments score you in under an hour. Independent firms, large consulting firms and build programs quote per engagement; nobody publishes a market rate, so treat any "typical cost" range you see online as a sales anchor.

Tier	Cost	What you get
Vendor self-assessment	Free	Self-scored across 6-7 pillars in under an hour (Cisco and Microsoft offer one; AWS and Google publish frameworks you can score yourself against). Fast and directional, but self-reported and shaped to sell the vendor's product.
Independent / boutique	Quoted per project	A study of a few weeks that inspects your real data and systems, interviews stakeholders, and returns a costed roadmap. The right fit for most mid-market companies.
Big-Four / strategy firm	Quoted, highest	A longer engagement with many stakeholder interviews and a board-grade narrative. Right when the real need is executive consensus or heavy regulation; overkill when you just need to know what to fix.
Strategy + implementation	Priced as part of the build	Assessment folded into a build: proof of concept, production system, MLOps, and a multi-quarter program. Not really an assessment anymore.

Price is driven by data complexity and governance gaps (data clean-up is usually the biggest variable), regulatory overhead, and how much of the work a senior practitioner actually does versus a junior.

At the large-firm tier, the partner who sells the engagement is rarely the one doing the work. Ask for the CV of the person who will actually run your assessment, not the one in the sales meeting.

Signs a "cheap" assessment is really a sales call:

a long paid discovery phase with no concrete deliverable
a "credit-back" structure that rebates the audit fee against a future build (a conflict of interest, not a discount)
a sub-$2,000 price tag, which buys a questionnaire, not an evaluation

Start with a free self-assessment to find the obvious gaps, and pay for the next tier only when you need someone to look at your real environment and tell you how ready you actually are.

Timing

Who actually needs an AI readiness assessment, and when?

You need a real one before a big AI spend, after a stalled pilot, or when leadership cannot agree on where to start. The simplest rule: if you have a specific use case worth shipping in the next 6 to 12 months, get a paid, use-case-specific assessment. If you do not, run a free one and define the use case first. Pay when guessing wrong would be expensive.

The clearest trigger is a proof of concept that worked in the demo and then went nowhere. That is the moment the readiness gaps became real, and the cheapest time to diagnose them, while the memory of what broke is fresh.

The second is the opposite: you have not started, and three executives have three different ideas about which use case comes first. An assessment turns that argument into a sequenced plan. Regulatory exposure and a board mandate without execution clarity round out the list.

Small businesses and single teams with one obvious use case can skip the expensive version. Run the free self-assessment, fix the one or two things it flags, and spend the rest of the budget actually building.

If you can already answer three questions, you are probably past needing a paid diagnostic: do you have a written AI strategy tied to business goals, do you have a pilot in production, and can you name your single weakest dimension? Three yeses means build, not assess.

Where we fit

What separates a useful assessment from a checkbox exercise?

Evidence. A checkbox assessment asks you to rate your own data quality on a scale of one to five. A useful one opens the data warehouse and looks. The difference shows up immediately, because the dimension everyone overrates is the one that breaks the most projects: data readiness.

The skeptics have a point. Vendor self-assessments are sales funnels by design. A CIO.com columnist calls the pattern "AI strategy theater": a board slide listing fifteen pilots, none tied to a measurable business outcome. An assessment that only produces that slide is part of the problem. Maturity scores become stalling tactics, a way to study a problem instead of fixing it.

And there is a structural conflict of interest whenever the firm doing the assessment also sells the implementation: one white-label provider sells its readiness assessment to agencies on the promise that every assessment generates three to five follow-on implementation projects. A readiness assessment is only worth paying for if it is evidence-based, specific to your use case, independent of the downstream build, and ends in a decision.

We build AI systems for a living, including our Company Brain platform that runs retrieval, knowledge graphs, and document ingestion on a client's own infrastructure, so an assessment we run is anchored to what it actually takes to ship, not to a maturity-model abstraction.

Our AI readiness work ties every finding to a specific use case and a specific next action, ranked by impact and effort, and ends with what a real assessment should produce:

a use-case-specific gap analysis
a feasibility-ranked list of what can ship in 90 days
a costed roadmap
a clear decision: build now, build with prep, or fix the foundations first

It has the same shape as a good intelligence report: what is true, what it means for you, and what to do about it.

A readiness assessment is a snapshot, and AI moves fast enough that it ages within a quarter. Its job is to keep you out of the half or so of projects that get abandoned after the pilot.

The readiness-to-value link is also largely correlational, high performers invest more and execute better, so spend alone is not the cause. Re-check the dimensions that were weak once you have fixed them, ignore the ones that were already strong, and never let the assessment become the reason you are not building yet.

Common questions

AI readiness assessment FAQ

What is an AI readiness assessment?

An AI readiness assessment is a structured diagnostic that scores whether your organization can actually deploy and sustain AI in production, across six dimensions: strategy, data, infrastructure, talent, governance, and culture. It does not ask whether AI is exciting. It asks whether you can ship it. A good one ends with a prioritized, costed roadmap, not a maturity badge.

How much does an AI readiness assessment cost?

There is no published market rate; firms quote per engagement. The tiers are clear, though. Free online self-assessments (Cisco's and Microsoft's; Microsoft's takes about 45 minutes) give a directional score. An independent firm that inspects your real data and systems costs a project fee. Large consulting firms charge the most, and assessments folded into an implementation program are priced as part of the build. What drives the price is the state of your data and your regulatory scope, not the brand on the report.

What is the difference between AI readiness and AI strategy?

AI strategy decides what to build and why. AI readiness decides whether you can build it without the project dying after the pilot. Many stalled AI programs had a strategy. What they lacked was usable data, governance, and a path from the demo into production. Readiness sits before strategy and before any build.

How long does an AI readiness assessment take?

A free online self-assessment takes under an hour (Microsoft's takes about 45 minutes). A serious evidence-based engagement that inspects your actual data and systems takes weeks, depending on scope and how many business units are involved. The slow part is never the scoring. It is getting honest access to the data warehouse, the integration map, and the people who know where the problems are buried.

What are the dimensions of an AI readiness assessment?

Most frameworks converge on six: strategy, data, infrastructure, talent, governance, and culture. Cisco uses six, Microsoft seven (including model management), Gartner seven. The substance is the same either way. Cisco's 2025 index shows readiness gaps across all six pillars, and Gartner finds data availability and quality among the top barriers to getting AI into production.

Do small businesses need an AI readiness assessment?

Not an expensive one. A small business with one obvious use case is better served by a free vendor self-assessment to surface the obvious gaps, then targeted help on the one or two dimensions that actually block that use case. The expensive enterprise-grade assessment is built for organizations with a large data estate, multiple business units, and real regulatory exposure.

Related Insights

Want to keep reading?

These guides go deeper on the strategy and economics behind getting real value out of AI and intelligence work.

AI Business Intelligence

Where AI actually helps with analysis, and where it fakes it

BI Consulting: Costs & ROI

What business intelligence consulting costs and what you get

Outsource Market Research?

When to build the capability vs. buy the deliverable

Before you spend

Half of AI projects die after the demo. Find out why yours won't.

We assess readiness against your real data and systems, not a questionnaire, then hand you a prioritized roadmap tied to specific use cases. US-based, human expertise plus AI power, no offshore subcontractors.

Get an Honest Assessment See AI Readiness