The AI readiness checklist: 12 questions to answer before you buy another AI tool
Twelve yes/no questions across data, systems, process and people that predict whether an AI initiative reaches production. Under an hour, no vendor required, and “mostly” counts as a no.
11 min read
01
WHY A CHECKLIST
02
THE 12 QUESTIONS
03
SCORING AND NEXT STEP
TL;DR
Precisely’s 2026 State of Data Integrity and AI Readiness study found 88% of leaders say their data is ready for AI, while 43% of the same group call data readiness their biggest barrier. Cloudera and HBR Analytic Services put the share of enterprises whose data is completely ready at 7%.
Cisco has measured the same 13% of organizations as fully AI-ready two years in a row. Only 19% have fully centralized data, against 76% of that ready minority; 35% have a change-management plan, against 91%.
Twelve yes/no questions in three layers (data, systems, process and people) surface the gaps that stall pilots. They take under an hour, need no vendor, and are written for a revenue operation running Salesforce or HubSpot rather than for a data-platform team.
Ten or more yes answers: take the clearest use case to production. Six to nine: pilot-ready, not production-ready. Five or fewer: stop evaluating vendors and fix ownership and definitions first, because those cost decisions rather than software.
The readiness gap is a confidence gap
Ask an executive team whether its data is ready for AI and the answer is almost always yes. Ask the same team what is holding its AI program back and the answer is almost always the data. Precisely’s 2026 State of Data Integrity and AI Readiness study, published in January 2026 from a survey of more than 500 data and analytics leaders, captured both answers from the same respondents: 88% claimed their data was ready for AI, and 43% named data readiness as their single biggest barrier. Only 31% had AI metrics tied to business KPIs at all.
Independent surveys land in the same place. Cloudera and Harvard Business Review Analytic Services reported in March 2026 that only 7% of enterprises describe their data as completely ready for AI, 27% say it is not very or not at all ready, and siloed data is the most-cited obstacle, at 56%. Cisco’s AI Readiness Index scored 7,985 leaders across 49 metrics in 2024 and found 13% fully ready; the December 2025 edition, with roughly 8,000 leaders, found 13% again, even as 98% reported rising urgency. Microsoft’s AI Readiness Advisor, a May 2026 study of 1,000 organizations across 15 countries, split readiness into technology and organizational dimensions and found about 30% strong on each. Only the companies strong on both were delivering impact.
The pattern behind those numbers is simple. Readiness is being assessed by feel, in a meeting, by the people who bought the tools. Nobody has written the questions down and answered them in the presence of someone who would know the real answer, such as the Salesforce admin, or the analyst who rebuilds the pipeline report every Monday because the dashboard version is wrong.
Why a checklist beats a strategy deck
There is no shortage of frameworks. Cisco scores six pillars: strategy, infrastructure, data, talent, governance and culture. AWS’s Cloud Adoption Framework for AI has six perspectives of its own, from business through security and operations. Microsoft’s advisor uses ten domains. Salesforce published a five-part test in September 2025 for whether data is ready for an agent, covering unification, identity resolution, governance, real-time activation and feedback loops. All of them are sound. All of them were written for a large enterprise with a data-platform team, and none of them tells a 600-person software company running Salesforce or HubSpot what to look at on Tuesday.
A checklist does something a framework cannot: it forces a yes or a no. Gartner’s data-management survey found 63% of organizations either lack the practices AI-ready data requires or are not sure whether they have them, and predicted that through 2026, 60% of AI projects unsupported by AI-ready data would be abandoned. “Not sure” is the operative phrase. The questions below exist to replace it.
We use these twelve in the first conversation with every client. They can be answered honestly in under an hour. They require no vendor, no tooling and no data-science team, and the pattern of answers predicts the next six months more reliably than any demo. Three rules: answer in writing, answer with the people who actually run the systems in the room, and treat “mostly” as a no. The checklist below groups the twelve questions into three layers, from the data up to the people who are supposed to be accountable for it.
The twelve questions
Each question comes with the reason it is on the list. The order is deliberate. The data questions are the ones most companies fail without knowing it; the people questions are the ones most companies fail while knowing it perfectly well.
Data: can the records be trusted? (Questions 1–4)
Cisco’s 2025 index found only 19% of organizations have fully centralized data, against 76% of the ready minority. That single number accounts for most of the distance between the two groups, and the four questions here are how to find out which group you are in.
1. Can you name the single system of record for customers, deals, invoices and support cases, and is it the same answer from sales, marketing, finance and service? Why it matters: an agent pointed at three versions of the customer picks one, and it picks without telling anyone. Demand Gen Report’s 2026 survey found only half of B2B companies have a unified sales and marketing database; in 2024, 57% were still keeping customer data in spreadsheets.
2. Do you know what share of your CRM records are duplicated, incomplete or stale, as a measured number rather than a feeling? Why it matters: Validity’s 2025 State of CRM Data Management found 76% of CRM users believe less than half of their data is accurate and complete, and 37% had lost revenue because of it. Industry estimates put B2B contact decay at roughly 30% a year. If nobody measures it, assume the worst; you will usually be close.
3. Are the definitions of lead, opportunity, customer, churn and revenue written down and version-controlled, or renegotiated every time the board deck is built? Why it matters: every model and every agent inherits the disagreement silently. A Salesforce org where the Opportunity StageName picklist means three different things to three sales managers cannot produce one forecast, with or without AI.
4. Is there a governed reporting layer between your operational tools and your dashboards, so a number on a slide can be traced to the records that produced it? Why it matters: Salesforce’s 2026 State of Data and Analytics found 49% of organizations occasionally or frequently draw incorrect conclusions from their data, and insightsoftware’s 2024 survey of 500 finance professionals found 75% spend five to six hours or more a week recreating reports. An AI layer on top of that makes the wrong conclusion arrive faster.
Systems: does context move without people? (Questions 5–7)
MuleSoft’s 2026 Connectivity Benchmark counts 957 applications in the average enterprise, 27% of them integrated, and reports that 96% of IT leaders say agent success depends on data integration. Half of the agents already deployed run in silos. Mid-market estates are smaller, but the shape is familiar to anyone who has drawn one on a whiteboard.
5. Do your core systems exchange data through maintained integrations, or through exports, imports and a person who “knows the process”? Why it matters: every manual hop is a place where context is lost and a place where an automated workflow will fail at 2 a.m. on a Sunday. If a CSV export from Salesforce to the billing system sits inside lead-to-cash, AI will make the export faster and the workflow no more reliable.
6. Have you inventoried the tools you actually run, and retired the ones you do not? Why it matters: Zylo’s 2025 SaaS Management Index found the average company operates about 275 SaaS applications, with more than half of purchased licenses unused. Every redundant tool holds a partial copy of the customer that an agent may read and trust.
7. Does every integration and automation have a named owner who is notified when it breaks? Why it matters: a HubSpot workflow built in 2023 by someone who has since left is still running, still writing to the Lifecycle stage property, and nobody will know it stopped until a quarter’s numbers look wrong. The test is whether you can produce the list of integrations with a name beside each.
Process and people: is anyone accountable? (Questions 8–12)
Cisco’s ready minority differs from everyone else most sharply here: 99% have a defined AI strategy against 58% overall, and 91% have a change-management plan against 35%. Lucid’s 2025 survey of roughly 2,200 knowledge workers adds the view from the floor. Only 16% describe their workflows as extremely well documented, and 61% say their company’s AI strategy is only somewhat aligned, or not aligned, with what the operation can actually do.
8. Is the workflow you want AI to touch documented step by step, including the decision rules a human applies today? Why it matters: AI automates documented process. It cannot automate one person’s intuition about which cases to escalate, and it will not discover the rule on its own. If the document does not exist, the pilot is where it will be written, expensively.
9. Can you state in business terms what a successful outcome looks like for that workflow: hours returned, cases deflected, forecast error reduced? Why it matters: Precisely found only 31% of organizations have AI metrics tied to KPIs. Without a number the pilot cannot fail, which also means it cannot succeed.
10. Is there one accountable owner for the AI initiative whose goals contain that metric? Why it matters: a steering committee owns nothing. In our experience a named owner is the single strongest predictor of whether a pilot reaches production, and it is the only item on this list that costs nothing but a decision.
11. Are access, permissions and data-handling rules defined well enough that an agent can be granted exactly the rights of the employee it assists, and no more? Why it matters: an agent runs as a user. If you cannot describe what a support rep is allowed to read and write in Salesforce today, you cannot describe it for the agent either, and the default is usually too much.
12. Is there a governance rhythm: who reviews what the AI did, how often, and who can switch it off? Why it matters: Salesforce’s 2026 study found only 43% of organizations have formal data governance in place. Governance is a launch requirement, and it is also what keeps a consumption-priced agent from running loops nobody reads.
Count the yes answers. The score sheet below turns that count into a next step, and it is deliberately blunt about the bottom band.
Scoring, and what the score is telling you
Ten to twelve: you are AI-ready. Choose the use case with the sharpest metric and take it to production, not to pilot. The foundation will hold, and the remaining no answers are the first two items on the production backlog.
Six to nine: pilot-ready, not production-ready. This is the most dangerous band, because the pilot will demo well. It will then stall on precisely the questions you answered no to, usually around month four, when the sponsor asks why the agent updated the wrong Account. Close those questions before scaling, not after.
Five or fewer: stop evaluating AI vendors. Every dollar spent on the AI layer now will be spent again after the foundation is rebuilt, and a consumption-priced agent will charge you for every action it takes on bad data in the meantime. This band is also the fastest to improve from, because the first fixes are decisions: an owner, a definition set, a declared system of record. None of them needs a purchase order.
What we actually see
A pattern from first conversations, with the caveat that it is what we observe across engagements, not a controlled sample. The typical mid-market technology company scores between five and seven. It answers yes to question 1 for deals and no for customers, because Salesforce is the system of record until the deal closes and nobody agrees what happens after. It answers no to question 2, and the admin in the room could produce the duplicate number in an afternoon if anyone asked. It answers yes to question 5 for the two integrations someone built carefully and no for the rest. It has never been asked question 10, because the AI initiative reports to a committee.
The most common surprise is question 6. Companies routinely discover a second marketing automation tool, a forgotten enrichment subscription still writing to Contact records, or a dashboard product that three people use and finance pays for. Each of those is a partial copy of the customer. The second most common surprise is that question 12 gets a confident yes from the CIO and a puzzled look from everyone who would actually do the reviewing.
What to do with the result
Score it in writing, with names. For every no, write the name of the person who would own the fix, even if the fix is not yet scheduled. A no with no name beside it will still be a no next quarter.
Fix questions 3 and 10 this month. A written definition set for lead, opportunity, customer, churn and revenue, and one accountable owner. Both are meetings, not projects.
Then questions 1, 2 and 7. Declare the system of record per object, measure duplicate and staleness rates on Account, Contact and Opportunity, and produce the list of integrations with an owner beside each. This is a few weeks of operations work.
Then question 8, for one workflow only. Document the decision rules for the single process you intend to put AI into first. Do not document everything.
Repeat the checklist each quarter. The score should move by two or three points a quarter while the foundation is being built, and it should not slip afterward.
The twelve questions are deliberately short. Behind them sits a longer instrument, the AI Readiness Health Check, with 30-plus concrete checks we run inside Salesforce and HubSpot environments: field-level completeness, duplicate rates, ownership maps, integration health and reporting lineage. We publish it as an open checklist so any operations team can run it alone. If you would rather have it run for you, that is a one-week engagement, and the output is the same scorecard, with the evidence attached.
Related reading
The AI Readiness Health Check: 36 checks to run on Salesforce or HubSpot - the 36-check instrument behind these twelve questions, with thresholds and sources.
AI transformation: the complete guide for mid-market technology companies - how readiness fits into the full transformation.
Why your CRM data quality decides whether AI works - the six CRM data numbers, with thresholds.
Is your Salesforce ready for Agentforce? A pre-flight health check - the Salesforce-specific pre-flight.
Sources: Precisely, 2026 State of Data Integrity and AI Readiness, January 2026; Cloudera and Harvard Business Review Analytic Services, March 2026; Cisco, AI Readiness Index 2024 and 2025; Microsoft, AI Readiness Advisor, May 2026; AWS, Cloud Adoption Framework for AI; Salesforce, “5 ways to measure data readiness for an AI agent,” September 2025; Gartner, “Lack of AI-Ready Data Puts AI Projects at Risk,” February 2025; Validity, State of CRM Data Management 2025; Demand Gen Report, 2026 unified data survey; insightsoftware, 2024 Finance Team Trends; Salesforce, State of Data and Analytics, 2nd edition, January 2026; MuleSoft, 2026 Connectivity Benchmark; Zylo, 2025 SaaS Management Index; Lucid, AI Readiness Report 2025; industry estimates of B2B contact data decay.
Frequently asked questions
What is an AI readiness assessment?
An AI readiness assessment is a structured test of whether a company’s data, systems, processes and people can support an AI initiative from pilot to production without a rebuild in between. The useful ones produce yes/no answers and a score rather than a maturity narrative, and they are answered by the people who run the systems, not the people who bought them.
What are the pillars of AI readiness?
Published frameworks vary. Cisco scores six pillars (strategy, infrastructure, data, talent, governance and culture); AWS’s Cloud Adoption Framework for AI uses business, people, governance, platform, security and operations; Microsoft splits readiness into technology and organizational dimensions. For a revenue operation on Salesforce or HubSpot, three layers cover the ground: data, systems, and process and people.
How do you measure AI readiness?
Answer a fixed set of questions in writing, score them, and repeat quarterly. The twelve questions above give a score out of twelve: ten or more means ready, six to nine means pilot-ready, five or fewer means the foundation comes first. Behind the score, the measurable inputs are duplicate rate, field completeness, record staleness, an integration inventory and a named owner per outcome.
What data is needed for AI readiness?
One system of record per business object; records that are measurably deduplicated, complete on the fields the use case reads and recently updated; written definitions for lead, opportunity, customer, churn and revenue; and a reporting layer where a number can be traced to its source records. Volume matters far less than ownership and consistency.
Is my company ready for AI?
If you can answer yes, in writing, to at least ten of the twelve questions above, yes. Most mid-market companies cannot yet. Cisco has found only 13% of organizations fully ready two years running, and Cloudera and HBR Analytic Services put completely AI-ready data at 7%. The cheapest gaps, ownership and definitions, are the place to start, and they close in weeks.
How long does an AI readiness assessment take?
The twelve-question version takes under an hour with the right people in the room. The full AI Readiness Health Check, which pulls actual numbers from Salesforce or HubSpot for duplicates, completeness, staleness, ownership and integration health, takes about a week. Repairing the gaps it finds typically fits within a quarter.
NEXT STEP
See where your organization stands before the next AI purchase.
A short conversation is enough to tell whether the AI Readiness Health Check, 36 checks on your Salesforce or HubSpot org, is the right first step. No decks, no obligations.
WRITTEN BY
Tal Oryon
MORE NOTES
Perspective
The world is changing. Is your organization part of it?
In one year the platforms your company runs on rebuilt their pricing, their products and their partnerships around agents. A note to executives on the one decision that is still theirs.
AI transformation
An AI transformation roadmap for a 300–2,000-person company (12 months)
Four quarters, one workflow at a time. The deliverables, the owners and the failure mode of each quarter, written for mid-market companies that cannot afford an enterprise timeline.
