Are You Piloting, Implementing, or Operating AI?

The question is not whether your organization has invested in AI. It has. The question is what that investment is producing right now, this week, and whether anyone is accountable for that answer.

Most leaders, when pressed, cannot answer that question cleanly. They know the platform is deployed. They know there was a go-live. They are less certain about what the AI is doing on a Tuesday afternoon in month eight.

That uncertainty is not a technology problem. It is a maturity problem. And it maps to a predictable state in the journey from pilot to operated.

The three states

Piloting. AI features are enabled. A proof of concept is running in a controlled environment. The platform vendor or an implementation partner helped stand up a demo. Leadership has seen the technology work. The conversation in the room is whether to commit and what it will take to go live.

Piloting is not a failure state. It is a necessary step. The problem is organizations that remain in pilot indefinitely, running new proofs of concept rather than committing to production, treating the exploration as the destination.

Implementing. AI went live. There was a launch announcement. The agents are deployed and technically active. The team is using the platform. Usage metrics are being reported to leadership. This is where most organizations are in 2026, and where most organizations stay.

Implementing is a success relative to piloting. It is not a success relative to the business case that justified the investment. The implementation delivered the capability. It did not deliver the outcomes. Those require operation.

Operating. AI agents are running in production with a defined accountability structure. Performance is reviewed weekly. Drift is caught before it compounds. New use cases advance on a roadmap with outcomes defined before work begins and measured after it closes. The organization can show a board exactly which agents are running, what each one produces, and what the next quarter looks like.

Operating AI is the state where the investment produces results that show up in business metrics: reduced ticket volume, improved cycle time, higher utilization, faster close. Fewer than 10% of organizations that implement AI reach this state without an active operator.

The ACT maturity model

L5 mapped five levels of AI maturity from 600+ production environments. The levels define exactly where an organization is, what it is producing at that level, and what it needs to progress.

1. Pilot. AI features enabled or in proof of concept. Not in production. No accountability for outcomes. Signal: "We are exploring what AI can do for us."
2. Use. AI deployed in production. Individuals are using it. No operating cadence. No outcome measurement. Signal: "Our team uses AI every day." (No one tracks what it produces.)
3. Build. Multiple AI use cases live. Agents deployed across processes. Performance is tracked but not actively managed. Signal: "We have agents running but we are not sure how well they are doing."
4. Govern. AI performance is measured. Someone is accountable. Operating cadence exists but is not yet consistent or systematic. Signal: "We review AI performance monthly and act on the results."
5. AI Operated. Weekly Drive cadence. Outcomes defined before work begins. Measured after it closes. Agents tuned continuously. Roadmap advances every week. This is where the investment produces results that show up in business metrics.

The CFO conversation

Every leader using this maturity model has a CFO or a board asking the same question: what is the AI investment producing? The answer depends entirely on which level the organization is at.

At Level 1 or 2, the honest answer is: we are not measuring yet. The platform is live and teams are using it. We have not defined what success looks like in business terms.

At Level 3, the answer is: we have agents running across several processes. We have some data on usage but we cannot yet connect it directly to cost reduction or cycle time improvement.

At Level 4, the answer gets stronger: we are measuring performance and we are acting on the results. We have a governance function. We are seeing early indicators that the investment is producing, but we do not yet have a rigorous quarterly cadence to report against.

At Level 5, the answer is clean: here is what each agent produced this quarter. Here is the trend line. Here is the roadmap for next quarter, with defined outcomes for every Drive. Here is the business case for the next phase of investment, grounded in what the first phase actually produced.

"Leaders using ACT to communicate upward are not defending spend. They are presenting a roadmap where every level was defined before work began."

Bipin Paracha, CEO, L5

What moves organizations from level to level

The gap between Level 2 and Level 5 is not a technology gap. The platform at go-live and the platform in month twelve is the same. What changes is whether anyone is actively running it.

Moving from piloting to implementing requires an implementation partner with the discipline to define done before starting and deliver within the scope of a Drive. Moving from implementing to operating requires something different: an operator who runs a weekly cadence, owns outcomes in the SOW, and measures against defined baselines every Drive.

These are different functions. Most implementation partners do not become operators. They implement and leave. The engagement ends at go-live. What happens to the AI after the handoff is not their accountability.

The question worth asking this week: Do you know every agent running in your organization right now, what each one is doing, and what outcome it is accountable for? If the answer is no, you are at Level 3 or below. The operating cadence that produces a yes is Level 5. L5 builds that cadence as the first Drive of every Operate engagement.

Where are you?

Three questions that will tell you.

Who reviewed your AI agent performance last week, what did they find, and what did they change? If the answer is nobody, you are at Level 2 or below, regardless of how many agents are deployed.

Do you know your current deflection rate, automation rate, or cycle time improvement attributable to AI, and is that number trending up or down over the last two quarters? If you cannot answer this, you are at Level 3 or below.

Is there a Drive backlog with defined outcomes for the next four weeks, and is someone accountable for shipping one output per week against that backlog? If not, you have not started operating yet. You have implemented.

The gap between implementing and operating is real, it is measurable, and it is where the investment stops producing results. Assess is the one-week engagement that scores where you are, maps the gap, and gives you the business case to close it.