- Market
- Saudi Arabia
- Written for
- CIO, Head of Digital, COO
- Decision
- Establish whether the organisation is ready to get value from AI before commissioning anything
- Languages
- EN · AR
The review
The five inputs an AI readiness review weighs
Readiness is not a score; it is a decision made from five inputs. The honest outcome is sometimes "prepare first", and a review that cannot produce that answer is a sales process.
Five inputs — data availability and quality, a named owner, process documentation, integration surface, governance appetite — converge into a readiness decision that produces one of three outcomes: a credible first use case, a preparation plan, or a deliberate deferral.
Data availability and quality
Can the data be reached, and does it reflect reality?
A named owner
One accountable person with time and authority.
Process documentation
Is the process written down, or does it live in heads?
Integration surface
Can systems be connected to, or only looked at?
Governance appetite
Will leadership own the rules for errors and oversight?
What we do
The readiness decision
The inputs are weighed together, because strength in four does not compensate for a void in the fifth.
A credible first use case
Scoped where the inputs are strongest.
A preparation plan
Named data, ownership or documentation work first.
A deliberate not-yet
Cheaper decided now than discovered mid-project.
The checklist
What the AI readiness checklist examines, item by item
Each item comes with a test you can apply this week, without a vendor in the room.
Data you can actually reach
Ask for an export of the data the use case would depend on. If producing it takes a committee, permissions archaeology or a supplier's goodwill, the project inherits that friction on day one.
What passing looks like: The relevant data, exportable on request by your own staff.
Data that tells the truth
Sample recent records and check them against reality — duplicated customers, fields filled with placeholders, statuses nobody updates. A model learns the data's habits, including the bad ones.
What passing looks like: A known, honest picture of quality, even if imperfect.
One accountable owner
Name the person whose objectives the initiative sits in. If the answer is a committee, or someone with no time and no budget, the project has no owner — it has an audience.
What passing looks like: A named individual with authority, time and a stake in the result.
A process that is written down
Ask two people who run the target process to describe it independently. Where their answers diverge is where an AI system will behave "wrongly" by faithfully following the version it was shown.
What passing looks like: A documented process, exceptions included, that both would sign.
Systems that accept connections
Establish whether the systems the use case touches expose interfaces your team may use, or whether integration means screen-scraping, vendor tickets or contract renegotiation.
What passing looks like: A known integration path for each system in the use case.
An appetite for governing it
Put the uncomfortable questions to leadership now: what error rate is tolerable, who reviews the system's decisions, what happens when it is wrong in front of a customer. Silence here becomes paralysis later.
What passing looks like: Written answers on tolerance, oversight and escalation.
The first bet
Picking a first AI use case you can afford to learn from
The first use case carries a weight beyond its own value: it sets what the organisation believes about AI for years. A visible, high-stakes, customer-facing first project maximises the cost of the inevitable early errors, and one bad public failure buys years of internal resistance. The better first use case is internal, measurable and reversible — a process where the AI's output is reviewed by a person before it acts, where the metric that should move is agreed in advance, and where switching the system off costs embarrassment rather than revenue.
Choose it where your checklist answers are strongest, not where the ambition is largest. A modest use case with reachable data, a real owner and a documented process will ship, move its number and earn the mandate for the next one. A flagship use case sitting on unreachable data and an undocumented process will spend its budget discovering this page the expensive way.
Our role
Our part once the AI readiness picture is honest
- 01
A review that can say not yet
We run the readiness review as a diagnosis, not a pitch — and where the inputs point to preparation first, the recommendation is data, ownership or documentation work, some of which needs no vendor at all.
- 02
The preparation itself
Where the gap is technical — data that needs consolidating, systems that need interfaces — we do that engineering, and it holds its value whichever AI direction you take afterwards.
- 03
A first use case built to be judged
When a use case is chosen, we build it with the metric, the evaluation and the human fallback agreed before launch, so the organisation learns from a measured result rather than an impression.
What the organisation is ready to change is a leadership judgement; the checklist puts facts under it, and our work starts wherever the facts say it honestly can.
Common questions
What buyers ask
How much data do we need before an AI project makes sense?+
The useful measure is not volume but fit: whether the data you hold actually describes the decisions the use case would touch, whether it is reachable by your own team, and whether it reflects how the process really runs. Modern approaches have also shifted the question — systems built on large pre-trained models often need your data for grounding and context rather than for training from scratch, which lowers the volume bar considerably. A small, honest, reachable dataset supports a good first use case; a vast dataset that is scattered, stale or misleading supports nothing. Judge fit and access first, and size only within that.
Who should own an AI initiative — IT or the business side?+
The owner should be the person accountable for the number the use case is meant to move, which usually places ownership on the business side, with IT as an essential partner for data, integration and security. An initiative owned by IT alone tends to optimise for a working system rather than a changed outcome, and it struggles to alter the business process around the technology — which is where most of the value is realised or lost. What matters more than the department is that it is one named person, with the authority to change the process and the standing to defend the project through its early, imperfect months.
What does it mean if we fail most of the AI readiness checklist?+
It means the highest-return work available to you is not an AI project — it is the preparation the checklist just named, and knowing that is the checklist paying for itself. Consolidating data, appointing an owner, documenting the target process and opening integration paths are investments that improve the organisation regardless of AI, and they convert a would-be failed project into a later successful one. The expensive alternative is proceeding anyway: the project then performs this same discovery at delivery rates, under deadline, with a vendor whose commercial interest is to keep building.
In this cluster
A working session on your checklist answers — what they say about a credible first use case, or about the preparation that should come first.
Review your readiness with us
