Checklist · Readiness · 5 min read
AI Systems Readiness Checklist
A vendor-neutral checklist for working out whether your organisation is actually ready to run AI systems in production — and what to fix first if it is not.
LB Labs · Updated Aug 2026
At a glance
- Readiness is operational, not technological — most blockers live in process and ownership, not in models.
- Clean-enough data beats perfect data: what matters is knowing where it lives and who maintains it.
- Every AI system needs a named owner before it goes live, not after.
- Score the four areas separately — your readiness is your lowest score, not your average.
- A working copy of the checklist is available on request.
Most AI projects that fail do not fail because the technology was not ready. They fail because the organisation was not — the process was undocumented, the data was scattered, nobody owned the outcome, and the team discovered all of this three weeks after go-live instead of three weeks before.
This checklist is the assessment we run before recommending any build. It is deliberately vendor-neutral: it does not assume a platform, a model or a budget. Work through it honestly and you will know where you stand — and what to fix first.
What readiness actually means
Readiness is not a maturity score or a slide in a strategy deck. It is a practical question: if an AI system started doing real work in your business on Monday, what would break?
Four areas answer that question — the data the system reads, the process it participates in, the governance that keeps it accountable, and the people who work alongside it. A serious weakness in any one of them will surface in production, usually at the least convenient moment.
Data and systems
An AI system is only as reliable as the information it can reach. This is rarely about volume — most businesses have more than enough data. It is about location, access and trust.
- You can name the systems where your customer, job and financial records actually live.
- The records an AI system would rely on are current, or you know exactly where they drift.
- Key systems expose an API, an export or an integration path — not just a login screen.
- Someone is responsible for each core dataset, and they know it.
- You know which data is sensitive, where it is allowed to travel and where it is not.
Process clarity
Automation amplifies whatever process it is given. A clear process gets faster; a messy one produces mess at scale. Before any system design, the process itself has to be visible.
- The process you want to automate is written down — not perfectly, but truthfully.
- You know its volume: how many enquiries, jobs or documents move through it in a week.
- The exceptions are known. The cases that do not fit the happy path are listed, not discovered later.
- Handoffs between people and systems are explicit — who passes what to whom, and when.
- You can say what the process costs today in hours, delay or missed work.
Governance and ownership
The difference between an AI system a team trusts and one it quietly works around is almost always governance: clear ownership, clear boundaries and a clear way to intervene.
- Every proposed system has a named business owner — a person, not a department.
- You have decided which decisions the system may make alone and which require a human.
- There is an agreed way to pause or switch off the system without an engineer.
- Someone reviews the system's outputs on a schedule, and the review has teeth.
- You know which regulations and consent obligations touch the process being automated.
Team capability
AI systems change how work feels day to day. Teams that were consulted early tend to make systems better; teams that met the system at go-live tend to test its weaknesses.
- The people who run the process today have been involved in describing it.
- Someone internal will be trained to operate, monitor and question the system.
- Leadership agrees on what the system is for — cost, speed, quality or capacity — and would give the same answer independently.
- There is appetite to run a pilot honestly, including reporting the parts that do not work.
How to score yourself
Score each of the four areas from one to five, where one means most of these questions could not be answered and five means every box is ticked and evidenced. Resist averaging.
- 01Work through each area with the people closest to the process — not just leadership.
- 02Score every area separately. Your readiness is your lowest score, not your average.
- 03For any area below three, fix the two cheapest gaps first — they are usually process documentation and ownership, and both cost time rather than money.
- 04Reassess after the fixes. Most organisations move a full point within weeks once the gaps are visible.