AI Readiness Assessment & Strategy

Proceed: permitted data, named owners and a measured baseline are in place, and the candidate route meets the agreed evaluation cases; pilot authorization and remaining specialist approvals are separate. Repair: a remediable gap has an owner, such as an incomplete permissions map or missing baseline; resolve it and recheck the affected evidence.

Narrow: users, sources or case types exceed the available permissions, evidence or review capacity; define a smaller boundary and evaluate it again. Stop: permitted data or an acceptance owner cannot be secured, or an existing tool already meets the need without this pilot. A useful non-AI route may continue.

These are working labels, not a score or automatic go/no-go test. The sponsor and responsible reviewers agree the evidence and thresholds for the actual use; passing the listed examples alone is not pilot acceptance.

Start with the least complex route that could meet the same agreed acceptance cases. Unless the evidence already rules a route out, compare an improved manual process or checklist, then existing search, rules or enterprise tools. Consider a hosted model with retrieval when source-backed drafting is a named requirement those routes cannot meet, and a bounded custom change only for a specific remaining gap, such as a required access boundary or integration.

This is a comparison heuristic, not a mandatory sequence or validated selection algorithm. At a workload plainly beyond manual capacity, document that reason rather than run a pointless manual trial. If existing search finds approved passages but cannot produce the required draft, evaluate source-backed drafting against the same cases. State what to retain, change and exclude, and what evidence would change the recommendation.

For this assessment, data readiness for AI means that the required sources are permitted for the intended use, have named owners and identifiable applicable versions, respect required role or customer boundaries, and can support representative evaluation cases. It does not mean a company-wide data clean-up or establish answer accuracy by itself.

The proposed pack covers the workflow and baseline, a data and permissions map, options and a recommendation, an evaluation and acceptance plan, a prioritized next milestone, a monitoring and fallback plan, and a handover record of decisions, evidence locations, versions, outstanding risks, acceptance responsibilities and who owns the next step.

Agree which outputs need work: this is not a fixed bundle included in a curator call or founder session. Missing measurements remain marked as unknown. The options and recommendation state what to retain, change or exclude, the alternatives considered, trade-offs, dependencies, assumptions and reasons to stop.

Against the agreed scope: the recommendation, supporting evidence, alternatives, unresolved assumptions and bounded next milestone. Distinguish performed checks from planned tests and identify each reviewer's approval boundary.

A later pilot needs its own demonstration; pilot acceptance covers that record, and the AI adoption roadmap covers sequencing.

If the task and fix are already understood, such as a known prompt fix, scope the work directly. Use an assessment when a material unknown remains, such as data rights, the acceptance owner or whether an existing tool is enough.

If the decision is which product to buy, use technology vendor selection; if an existing system keeps missing releases, start with a technology assessment. Use AI governance consulting when authority to approve or stop work is unresolved.

The number and depth of workflows, available code and failure evidence, data permissions, integrations and implementation route shape the scope. Timing, commercial terms and participation are confirmed after that fit-gap review, not inferred from a generic assessment checklist.

Marton Csernai's profile describes RegTech RAG with Temporal orchestration and reports 100% citation provenance for regulated document analysis. That is attributed document-workflow experience to discuss against your proposed scope; citation provenance concerns traceability, not answer accuracy.

Ask which decisions the proposed CTO personally designed or operated, and agree participation through the curator.

It establishes the evidence for one pilot decision, not a company-wide maturity score, audit certification or data clean-up. The first discussion establishes whether the question fits, not a commitment to build or release the pilot.

Grand Rounds connects you with CTOs in its network through the curator. CTO leadership and implementation require their own agreed scopes and evidence; domain, legal, security and data approvals stay with their responsible owners. Grand Rounds is an initiative of Dinarys, and a conversation may lead to cooperation with Dinarys.