What is AI readiness?
AI readiness is the degree to which a company has the foundations needed to deploy AI in production and keep it running. An AI readiness assessment examines those foundations against the use cases the company actually plans. Readiness is always relative to a use case. A company may be ready for an assistant that searches published policies and unready for an agent that posts entries to the general ledger.
Assessments usually cover five areas.
- Data. The records the use case needs are accessible, reasonably clean, and documented in one place.
- Infrastructure. There is an approved place to run models and a way to connect them to systems of record through application programming interfaces (APIs).
- Identity and security. Single sign-on, role-based access control, and audit logging exist and reflect who may see what.
- Governance. The company has an acceptable use policy, an inventory of AI tools, and a process for approving new use cases.
- People. Staff have basic AI literacy, and someone has the skills and time to run the system after launch.
An AI maturity model describes stages of capability across the whole company over several years. A readiness assessment has a narrower purpose: to decide whether the next specific project can start.
AI readiness in practice
A credit union plans an assistant that answers representatives’ questions from policy manuals and member account history. The assessment finds policies spread across three file shares, with several outdated versions of the same procedure. Single sign-on is in place, but its groups do not separate branch staff from lending staff. Staff are also pasting member details into consumer chatbots, a case of shadow AI. The credit union retires stale policies, maps roles in its identity provider, and publishes an acceptable use policy before building the assistant. Kirk Tech Solutions typically checks identity and data access first in readiness work, because an assistant can enforce only the permissions that the identity provider already expresses.
What to watch for
- A single readiness score for the whole company says little about whether a particular project can start.
- Fix the data the first use case needs instead of waiting for every system to be clean.
- Treat shadow AI as evidence of demand, and offer a governed alternative alongside the policy.
- Be cautious with vendor assessments that conclude the company needs the vendor’s own product.