AI readiness assessment: score your organisation in ten minutes

What is an AI readiness assessment, and what does it measure?

An AI readiness assessment checks whether an organisation has the foundations to put AI to work on real tasks: a chosen use case with a baseline, knowledge an agent can read, written processes, people with time to review, systems it can reach safely, and rules for what it may do alone. It measures foundations, not the AI tools you already pay for.

Readiness is relative to a next step: a business not ready for an agent that replies to customers may be ready for one that drafts a weekly report.

Each question offers four answers you could check, such as "we measured it for two weeks", not an agree-or-disagree scale. The framework behind AI Singapore's AI Readiness Index makes the same choice: its answers are "descriptions of observable behaviour, not opinions".

What are the six dimensions, and why does each matter?

Six dimensions, three questions each. Strategy, data and governance carry 20% of the score each, processes and people 15%, technology 10%.

Technology weighs least on purpose: AI Singapore's AIRI page notes that smaller organisations, "even if they lack data, talent, or ML infrastructure, could adopt ready-made AI solutions".

How is the AI readiness score calculated?

Each answer scores 0 to 3. A dimension scores its points out of 9 as a percentage; the overall score is the weighted sum of the six, rounded. The stage follows from it, then two gates.

• Governance gate. Governance below 34 caps the stage at Building foundations: an agent that sends, pays or deletes needs approval rules first.
• Weakest-link gate. Ready to scale needs every dimension at 50 or more, because scaling exposes the weakest one.

AIRI version 3.2 has a similar rule: the overall level "can never exceed" the ethics and governance level. Next steps come from your three lowest dimensions (ties go strategy, governance, data, processes, people, technology) and their lowest answers.

AI readiness vs AI maturity: which frameworks are worth knowing?

A readiness assessment asks whether the foundations for a next step are in place; an AI maturity model places you on a ladder of levels showing how deeply AI is built into your work. Public frameworks cover different parts.

AI Singapore's own self-assessment still uses AIRI v2.0, while airi.foundation listed v3.2 when we checked. IMDA's agentic AI factsheet asks organisations to "trigger human approvals at significant checkpoints", the approach these next steps take. This is Olano's own method, not an AIRI score or a certification, and none of these bodies has endorsed it.

Worked example: a 25-person services firm

Press Try an example to load these answers. A 25-person professional services firm gets 100 to 500 client messages and spends 10 to 25 hours on email each week. Most answers are 2s, but governance is all 1s.

A score of 56 would mean Ready to pilot, but governance at 33 is under the gate, so the stage is Building foundations. Next steps go to governance, processes and people (the tie order puts technology last). The first agent suggested is a customer-message agent with human approval, with one caution: decide first that outbound replies wait for approval. A one-page policy and that rule lift governance to 56, the score to 60 and the stage to Ready to pilot.

How do you raise a low AI readiness score in 90 days?

Fix the three weakest dimensions your result names, for one use case only; most take weeks, not months.

1. Weeks 1 and 2. Pick the use case and count it: volume, hours, errors. Name a sponsor and a reviewer.
2. Weeks 3 and 4. Write the process down, put its answers in one shared document, and write a one-page policy with one rule: anything outbound waits for a person.
3. Weeks 5 to 8. Pilot with internal users or test data, on limited accounts, reviewing every output.
4. Weeks 9 to 12. Compare with the baseline, loosen approvals only where the record supports it, and retake this assessment.

See how to measure AI ROI with numbers you already have and where to start with AI in your business.

What are the most common AI readiness mistakes?

• Starting with a platform before a use case. The project becomes a search for a use.
• No owner. Without a named reviewer, approvals pile up or get quietly switched off.
• No human approval on outbound actions. Drafting is cheap to get wrong; sending, paying and deleting are not. Gate them from day one.
• Ignoring data access. Shared logins, administrator permissions, personal data sent to a service nobody checked. Give each agent its own limited access.
• No baseline. Without a before number, a good pilot cannot prove itself and a bad one cannot be stopped on evidence.

How Olano helps

Olano Studio, our Singapore-based implementation team, runs this sequence as a project. Its AI strategy and advisory service starts with an opportunity audit of your workflows, data and tools and a roadmap scored by effort and impact; delivery designs the agents and their approval points, connects your systems, and deploys with approvals and spending controls in place. See also AI for business.

The platform covers much of the governance dimension. Each deployment is isolated. Sending email, spending money, writing a file, using a secret and delegating work can each be held for a human, and high-risk operations need a human at every trust level. Every action is attributed on an append-only log, and hard spending caps apply per agent and per deployment. When you switch Cortex on, it runs Fully Autonomous by default; rewrites of an agent's own instructions, outbound sends and irreversible actions always wait for a human. See security.

You do not need Olano to act on your result, and no platform fixes a missing use case, owner or baseline.