Standardize before you automate
Is your model ready to automate — or just eager to?
An agent inherits whatever process you point it at — Deloitte's warning is that layering agents onto broken processes amplifies them. Rate your operating model against the five things that must be true before you automate, mapped to the GSDPI lifecycle. Score each honestly; the low scores are what an agent would inherit today.
The intake is standardized0 1 2 3
Work enters one governed front door — not side doors an agent would faithfully automate. A single, defined intake path exists and is followed.
Decisions are codified as rules0 1 2 3
The judgment that handles exceptions is written down as explicit, governed rules — not tribal knowledge the agent won't have.
Execution follows one governed path0 1 2 3
There is a single standardized route the agent runs, with every action logged — not a shortcut it discovers on its own.
Every output is traceable0 1 2 3
Any agent result can be followed back to the work and the rule behind it — an audit trail, not "reported, not proven" at machine speed.
Oversight is designed in0 1 2 3
Governance, monitoring, and human-in-the-loop roles exist before scale — not retrofitted after an incident.
Score /15. 12–15: ready — automation will amplify a model that works. 6–11: partial — agents will inherit real gaps; standardize the low scores first. Below 6: not ready — automating now scales the dysfunction at machine speed. A Zero-Based read across GSDPI closes the gaps; the Traceability Ratio becomes the agent's audit trail. Standardize, make it traceable, then automate.