Score five gates, not your tech stack: 5,000+ contactable records, 200+ new leads a month, a median lead response slower than 30 minutes, 60%+ of records with a valid phone and a last-activity date, and a 20%+ close rate on meetings held. Two points a gate, one if marginal: 8 of 10 means deploy, 4 or fewer means fix the business first.
At a glance
- The five gates: contactable records, new lead volume, speed to lead, CRM hygiene, close rate on held meetings.
- Scoring: 2 points if you clear the gate, 1 if you are marginal, 0 if you are not there. Maximum 10.
- 8–10: deploy now. 5–7: one fix first — your lowest-scoring gate names it. 0–4: this is not an AI problem yet.
- The inverted gate: a slow lead response is headroom, not a disqualifier — the only gate where scoring badly makes AI more likely to pay.
- Time to complete: about 20 minutes with your CRM open.
How it works
How to score your own AI readiness in 20 minutes
Pull five numbers
Export contactable records, monthly lead volume, median first-response time, phone-field completeness and close rate on meetings held. All five live in your CRM.
Score each gate
Two points if you clear the threshold, one if marginal, zero if not. Gate 3, speed to lead, scores in reverse: slow is headroom.
Read the total
8-10 means deploy now. 5-7 means fix one gate first. 4 or fewer means this is not an AI problem yet.
Remediate or scope
Fix the lowest-scoring gate, or scope a single use case and instrument the baseline before launch so the result can be proved.
MAKE MORE SALES.
Pay-Per-Result pricing — We scale sales HARD aligned to your interests, better than anyone else.
“Is my business ready for AI?” — the five gates that decide it
Readiness is a measurement, not an opinion, and every input sits in a system you already own.
| Gate | What you measure | 2 points | 1 point | 0 points |
|---|---|---|---|---|
| 1. Contactable records | Records holding a mobile or email and a recorded lawful basis to contact | 5,000+ | 1,000–4,999 | Under 1,000 |
| 2. New lead volume | New leads entering the CRM per month, averaged over 3 months | 200+/month | 50–199/month | Under 50/month |
| 3. Speed to lead | Median minutes from enquiry created to first contact attempt logged | Slower than 30 min | 5–30 min | Under 5 min already |
| 4. CRM hygiene | % of records with a valid-format phone and a last-activity date | 60%+ | 40–59% | Under 40% |
| 5. Close rate | Sales ÷ meetings actually held (not booked), last 90 days | 20%+ | 10–19% | Under 10% |
Read gate 3 backwards. It is the only gate where a zero is good news about your business and bad news about the opportunity: if your median response is already under five minutes, that lever is pulled. If gate 3 is your only zero, score out of 8 and read the bands proportionally.
These are the thresholds we score an account against before we take it on, drawn from our own deployment record — 50,769+ AI-booked sales appointments since 2017 and 1M+ leads generated — not from published research. Treat them as a decision rule, not an industry benchmark.
Want this done for you? We book qualified sales appointments on a Pay-Per-Result basis — you only pay for calls that actually land in your calendar.
What an AI readiness assessment actually measures — and what it does not
An AI readiness assessment is a scored audit of whether your existing operating data — volume, reachability, response time, record quality and conversion — is sufficient for an AI system to produce a result you can measure. That is the whole definition.
It is not a technology maturity model: no gate above asks what is in your stack, because a data warehouse does not book meetings. It is not a strategy exercise: it produces a go/no-go on one use case, not a roadmap. And it is not a vendor evaluation — you score yourself before you speak to anybody.
The condition that changes the answer is the use case. There is no such thing as a company that is ready for AI; there are only use cases that are ready. The same business can clear every gate for outbound follow-up and fail at demand forecasting, which needs five years of clean transactional history where this one needs a phone number and a consent flag.
The blocker is almost always data, and it is not just you. Cisco’s AI Readiness Index 2025, a double-blind survey of 8,000 senior IT and business leaders responsible for AI strategy at organisations with more than 500 employees across 30 markets and 26 industries, reports that 64% of them struggle to centralise data — and that the “Pacesetters” who outperform their peers on every measure of AI value have held at about 13% of organisations for three years running. That denominator is large enterprises, not businesses your size.
How to score yourself in 20 minutes
Five queries, in this order, all inside your CRM.
- Gate 1. Filter contacts where mobile or email is not empty and a consent, source or opt-in field is populated. Export the count. Do not count the whole database; most of it is not contactable.
- Gate 2. Records created per month over the last three months, averaged. Every source, not just paid.
- Gate 3. The median minutes between lead-created timestamp and first logged outbound activity, over 90 days. Use the median: one rep clearing a backlog at 2am drags the mean into fiction.
- Gate 4. Percentage of records where the phone field matches a valid format and a last-activity date exists. Two fields, not a full audit. Deeper work belongs in a CRM data hygiene project.
- Gate 5. Closed-won divided by meetings held, over 90 days. Booked-but-no-show belongs to a different metric, not this one.
Every number here comes out of your own system. If a readiness assessment does not ask for these five, it is scoring your enthusiasm.
If we can’t make you money, we don’t deserve yours.
Pay-Per-Result pricing — performance-based alignment.
Why a slow lead response makes you more ready, not less
Most readiness frameworks treat a weak operating metric as a reason to wait. For lead response it is the opposite: the gap between your median response time and five minutes is the upside, and a business answering in four hours has more of it than one answering in four minutes.
Size it before you buy anything: count leads arriving outside staffed hours and divide by total leads. At 200 leads a month with 35% after hours, 70 leads are already going cold — that is what an always-on responder competes for, not the whole 200. Our speed-to-lead conversion rate benchmarks show what the recovered portion looks like.
On magnitude, be careful with the arithmetic. In our own client work we typically see speed to lead worth roughly 3x on its own, doubling contact rate worth about 2x, and doubling set rate about another 2x. These do not multiply. 3x × 2x × 2x is 12x on paper and nothing like that in practice, because the three overlap — the same after-hours lead gets counted as a speed win, a contact win and a set win. Model them as one improvement with three names.
What to do if you score below 5
The failing gate names the fix, and the order matters: out of sequence, the spend is wasted.
- Gate 5 under 10%. Stop. Do not add meetings to a sales process that loses nine in ten. AI multiplies your close rate; it does not repair it. Tighten written qualification criteria first — the cheapest fix on this list.
- Gate 4 under 40%. This is a data job, not an AI job: a dedupe and phone-validation pass, then re-score. It is measured in weeks, and it is the one remediation almost everybody underestimates.
- Gate 1 under 1,000. Do it by hand. Below roughly 1,000–2,000 contactable records, one person with a phone and a spreadsheet beats anything you can buy: setup cost is fixed and your list cannot amortise it.
- Gate 2 under 50/month. You have a demand problem sitting upstream of your conversion problem. Fix supply first or you will automate an empty pipe.
- Gate 3 your only zero. Nothing to fix. Re-score out of 8 and read the bands proportionally.
The readiness bar for AI outbound is lower — here is what we do not require
Readiness is per use case, so this page has to say which prerequisites a narrow use case skips. Conversational outbound, qualification and database reactivation need far less than a broad AI programme.
| Prerequisite | AI outbound / reactivation on your own list | Typical broad AI programme |
|---|---|---|
| Unified data warehouse or CDP | Not required — one CSV export with phone, email and consent | Usually a prerequisite |
| Training or fine-tuning a model on your data | Not required | Often required |
| Dedicated AI or data engineering hire | Not required | Usually a dedicated hire |
| Field-complete CRM records | Not required — 4 fields: name, phone or email, consent, last activity | Required for analytics use cases |
| Integration with every system of record | Not required — write-back to one CRM object | Usually multi-system |
| Documented lawful basis to contact | Required, no exceptions | Varies by use case |
That last row is the one people skip and the only prerequisite we will not waive: consent provenance, suppression handling and opt-out propagation have to exist before a single dial, which is what our AI outbound compliance checklist for enterprise buyers checks.
Running it in-house is doable and costs what it costs: list maintenance, script iteration, a reminder cadence for booked meetings, and someone reviewing transcripts weekly so the agent does not drift. In our own client work that last item quietly consumes about half a day a week, and it is the first thing teams drop. Handing it over changes the shape of that cost rather than removing it — our model is pay-per-result: a revenue share of 5–20% of sales we help generate, or a fee per booked qualified appointment, rather than a retainer or per-seat licence. AI appointment setting and the wider AI for business overview cover the delivered version.
What a readiness score cannot tell you
A readiness score tells you whether AI can help. It does not tell you whether your organisation will finish the project, and those are different risks with different owners.
RAND interviewed 65 experienced data scientists and engineers, each with at least five years building AI and machine-learning models in industry or academia, and found the leading root cause of failure is not data or technology: stakeholders misunderstand what problem is being solved, so models get optimised for the wrong metric or never fit the workflow. Their second cause — the organisation lacks the data — is the one the gates above test.
Worth knowing before you quote it: the line “more than 80 percent of AI projects fail—twice the rate of failure for information technology projects that do not involve AI” appears in that RAND report as cited background. Its footnotes point at a 2022 Fortune interview and a 2023 Harvard Business Review article, not at anything RAND measured. The 65 interviews are RAND’s own work; the 80% is not.
Two things no score catches: whether anyone owns the system after go-live, and whether your judging metric is instrumented before launch. If you cannot state today’s baseline close rate and response time to one decimal place, you cannot prove the result later, whichever way it goes.
Frequently asked questions
What is an AI readiness assessment?
A scored audit of whether your current data and operating metrics can support a specific AI use case. Vendor indexes measure the same idea at company level — RAND’s report on the root causes of AI project failure cites one such survey finding only 14% of organisations fully ready to integrate AI — but a company-level index cannot tell you whether your next use case will work. Score the use case, not the company.
How long should an AI readiness assessment take?
Self-administered, about 20 minutes: five CRM queries and one score out of 10. Vendor-run assessments run for weeks rather than minutes, because they add stakeholder interviews and a roadmap on top of the same numbers. If you are deciding on one use case, the 20-minute version answers the same question, and you should run it before paying for the long one.
What percentage of AI projects actually fail?
Nobody knows precisely, and the circulating figures measure different things. The most-quoted — over 80% of AI projects failing, twice the rate of non-AI IT projects — appears in RAND’s The Root Causes of Failure for Artificial Intelligence Projects and How They Can Succeed as cited background rather than as RAND’s own measurement. Other widely stacked numbers count pilots, production rates and adoption, which are three different denominators. Treat any single failure rate as directional.
Do I need clean data before I can use AI?
For a narrow outbound or reactivation use case, no — you need four fields populated, not a clean database: name, a phone or email, a consent record, and a last-activity date. For analytics and forecasting use cases, yes. Cisco’s AI Readiness Index 2025, surveying 8,000 senior IT and business leaders at organisations with more than 500 employees across 30 markets, found 64% struggle to centralise their data — which is why scoping to a use case that does not need centralised data is the faster route.
Is my business too small for AI?
Under about 1,000 contactable records, almost certainly yes for automated outbound — the setup cost is fixed and your list cannot amortise it, so a person and a spreadsheet wins. Under 50 new leads a month, the same holds. Small is not the disqualifier; small and already fast on lead response is.
Should we run the assessment ourselves or have a vendor do it?
Run it yourself first. All five inputs live in your CRM, the scoring rule is above, and a vendor-run version scored before you know your own numbers cannot be checked. Bring your score to a vendor conversation and ask them which gate they think is wrong — that question sorts the good ones quickly.
Pay-Per-Result appointments
See if we’re a fit
We book qualified sales appointments for you and you pay on results, not retainers. Our booking page asks a few quick questions so you find out in two minutes whether that model suits your business.
- 50,769+ appointments booked without cold calling.
- Pay-Per-Result pricing — you pay for booked, qualified calls.
- Pick your own time on our live calendar, no phone tag.