Agentic AI production readiness assessment
Ten questions, five scored dimensions, and the one dimension holding the other four back. Answer for a single agent, not for your organisation.
What is this agent actually for?
How this is scored, and the one rule that overrides the total
Each question scores 0, 1 or 2, so every dimension is out of 4 and the total is out of 20. The bands are our own judgement, not an industry benchmark, and they are set deliberately hard. Gartner expects 40% of enterprises to demote or decommission autonomous agents by 2027 over governance gaps that surfaced only after a production incident, which is what an easy scorecard would have missed.
One rule overrides the total. If you score 0 on verification, the verdict is capped no matter how well the other four dimensions score. An agent whose output nothing independently checks is not production ready, and no amount of data access, cost control or documentation changes that. This is the Greenlight position expressed as arithmetic rather than as an argument: work is proposed, it is independently checked against the standard it has to meet, and a person clears what matters. Remove the middle step and the other two stop meaning anything.
The autonomy dimension is scored the way Gartner recommends rather than the way most enterprises govern. Uniform controls across every agent score lower than controls set per use case, because an agent that reads a stock level is not the same risk as one that raises a purchase order.
Gartner, Applying Uniform Governance Across AI Agents Will Lead to Enterprise AI Agent Failure, 26 May 2026, and Over 40% of Agentic AI Projects Will Be Canceled by End of 2027, 25 June 2025. Scoring bands are Softobiz's own and are not derived from those reports.