Agent production ratio
measurementproduction-to-pilot ratio · agents in production ratio
How many of your agents actually run live processes, divided by how many you've built. It's the number that separates an AI programme from a folder full of pilots.
This is an Implementa editorial formulation — not an industry-standard metric, and no analyst publishes it. We coined it because the market counts agents built and nobody counts agents operating, and those two numbers have drifted so far apart that they no longer mean the same thing. How to calculate it: agents in production divided by agents built. An agent counts as 'in production' only if it meets all four conditions: it has a named business owner, it runs on a real process without someone kicking it off by hand, it has a declared service expectation, and there's a rollback plan when it fails. Miss one and it's a pilot — however well it demos. Why it matters: it's the only number that answers in one line whether your AI investment is working or waiting. A low ratio with many agents doesn't signal missing technology. It signals missing integration, missing ownership and missing governance.
How it differs from
- AI adoption rate
- Adoption rate measures people using AI. The production ratio measures processes AI executes without anyone launching it. You can have sky-high adoption and a zero ratio: lots of people in a chat window, not one automated process.
- Pilot to production
- 'Pilot to production' names the journey. The production ratio turns it into a number you can compare over time and across business units — which is what makes it manageable.
Examples
- 18 agents built, 4 owned and running unattended on live processes: ratio 0.22.
- Two companies with 12 agents each: the one at 0.8 has a programme; the one at 0.1 has a spent budget.
FAQ
- Is this an industry-standard metric?
- No. It's an Implementa editorial formulation. We defined it because there are public anchors for the numerator and the denominator separately — how many agents get deployed, how much decisions are trusted — but nobody publishes the quotient, and the quotient is exactly what tells you whether the programme works.
- What counts as a good ratio?
- We don't publish a target: there's no public basis for one and we're not inventing a threshold. What is useful is your own series. If you build faster than you productionise, the ratio falls, and that fall is the early warning regardless of the absolute level.