What gets audited
Implement Agentic's two-week assessment covers six dimensions, each scored 1–5 with evidence.
- Data quality — coverage, freshness, ACL propagation, table parsing.
- Identity & permissions — SSO, scoped service accounts, audit logging.
- Workflow maturity — documented SOPs, KPI definitions, exception paths.
- Integration surface — APIs, MCP servers, RPA, event streams.
- Governance — model risk, privacy, NIST AI RMF, OWASP LLM Top 10.
- Team capability — engineering, ops, change management readiness.
What you receive
Output is a 30–50 page report plus a working session: ranked use-case backlog, target architecture, remediation plan, vendor recommendations, ROI model, and a 90-day execution plan.
What the assessment is actually testing
A readiness assessment is not a maturity questionnaire. It tests one candidate workflow against the conditions an agent needs in order to work at all: is the context retrievable, is the decision reconstructable from recorded data, can the action be executed through a permissioned path, and is the outcome measured anywhere.
Most workflows fail on one of those four before any model question arises — usually permissions or outcome measurement. Finding that out in two weeks is the point of the exercise.
- Data: does the context the decision needs exist in a retrievable form?
- Permissions: can access be resolved per requesting user, at retrieval time?
- Actions: is there a safe, reversible write path into the system of record?
- Measurement: is the outcome recorded, and by whom?
- Ownership: is there a named operational owner after go-live?
What you receive at the end
The deliverable is a decision package rather than a report: a workflow and data inventory, an access-control and governance review, a closed-loop agent design for the workflow, an ROI model with the assumptions exposed, and a buy-versus-build recommendation with a tool shortlist by loop stage.
It is explicitly allowed to conclude that no agent should be built. A recommendation to fix data ownership, buy a product, or change the process is a valid and common outcome, and it is cheaper to reach it here than after a build.
How to prepare, and what slows it down
Assessments move fastest when the people who actually run the workflow are available for an hour each, when a sample of recent real cases can be reviewed, and when someone can answer authoritatively how access is granted in the source systems.
The common delays are access approvals for source systems, and discovering that the workflow as documented differs from the workflow as performed — which is itself a useful finding.