tioga.ai — free readiness checklist

ERP Agent-Readiness Checklist

22 questions across 6 areas — for a CIO, finance-transformation lead, or ERP owner sizing up whether an environment is actually ready for a governed AI agent, before scoping any engagement.

This is a self-assessment tool, not an audit. Unchecked items aren't disqualifying — they're exactly what a discovery sprint scopes out before any build starts.

01 — System & API surface

Can an agent actually reach your system through a sanctioned path, or only the database underneath it?

Known version and edition — e.g. Oracle EBS R12.2, SAP ECC vs. S/4HANA — confirmed, not assumed.
An application-logic API or BAPI layer exists for the target workflow (not just direct table access).
A non-production instance (test/sandbox) is available to build and validate against.
API rate limits and auth model for the relevant module are documented somewhere.

02 — Approval & control workflows

Every governed write needs a policy to enforce — most orgs already have one, just not written down in one place.

A delegation-of-authority matrix exists — who can approve what, up to what dollar amount.
Escalation paths are defined for exceptions above the autonomous-approval threshold.
A named owner for the workflow's approval policy (not just "finance" in the abstract).
Current exception volume and type for the target workflow, from at least the last 90 days.

03 — Compliance & audit requirements

Whatever framework applies, the agent's decision trail needs to be legible to it from day one.

Applicable framework(s) identified — NIST AI RMF, ISO 42001, EU AI Act, SOX, or internal audit standard.
Evidence-retention requirements known (how long, what format, who reviews it).
Internal audit or compliance has been looped in before any agent touches a live write path.
A rollback/reversal requirement is explicit, not assumed to be "someone will fix it manually."

04 — Data quality

A governed policy can't save a decision built on bad inputs.

Master data (vendor, customer, chart of accounts) for the target workflow is reasonably current.
Known data-quality issues in this workflow are documented, not just "we know it's messy."
A sample of real (or realistically shaped) records can be shared for prototype-building under NDA.

05 — Integration points

Most exception workflows touch more than one system — worth mapping before scoping.

Upstream/downstream systems for this workflow are identified (e.g. OCR/AP capture tool, bank file, CRM).
Existing middleware or iPaaS (MuleSoft, Boomi, etc.) that a governed agent would sit alongside is known.
No conflicting automation (RPA bots, legacy scripts) already writes to the same path unmonitored.

06 — Organizational readiness

The best governed write-path in the world doesn't matter if no one owns adopting it.

An executive sponsor is willing to own the pilot's outcome, not just approve budget.
The team executing this workflow today has been part of scoping, not just informed after the fact.
Success criteria for a pilot are defined before it starts (hours reclaimed, error-rate change, cycle time).
A realistic decision timeline exists for moving from pilot to production if it works.

Scoring this yourself: most orgs check 10–15 of these 22 on a first pass. That's normal — the gaps are exactly what a 5-day discovery sprint is scoped to close before any larger commitment, not a reason to wait. See a real example of the sprint's actual scope: sample discovery sprint document, or watch the governed write-path pattern itself run live: AP exception workflow demo.

Tioga AI — governed AI automation for Oracle EBS, SAP, and other complex enterprise systems. This checklist is a free self-assessment tool, not an audit or a substitute for a scoped engagement.