Agentic maturity assessment

For those who need to decide whether and where to use an artificial intelligence (AI) agent. We start from a map, before writing code.

Let's talk about your agent

The agentic maturity assessment is used to decide where an artificial intelligence (AI) agent is useful and where it would do harm, before writing code. It often happens that a company feels it must "do something with agents" before knowing where they are needed. The assessment answers this question by looking at your journeys, your processes and the people who make them work. Sometimes the answer is an agent; at other times it is enough to clarify a rule or change a step in the service. We design and test reliable AI agents with real users, and the work starts here, before the first prototype.

Who it is for

For those who need to decide where to use an agent, when the question is still "where does it make sense, and where must control stay with people".

  • The Chief Information Officer (CIO), who is responsible for information systems and must put initiatives in order.
  • The Chief Technology Officer (CTO), who is responsible for technology and wants to understand which processes and data can be built on.
  • Process owners and those who lead customer service and Customer Experience (CX), the customer's experience.
  • Those who have already received proposals from technology partners and want a view from their own side before choosing.

If you already have an agent in pilot or in production, you probably need Testing AI agents with real users.

What we look at

We reconstruct how the people who use the service work today, which data and tools are in place, where an error would cause harm, and who can intervene.

  • People's journeys. How customers and operators move, where they get stuck, where judgement and exceptions are needed.
  • Processes. Written procedures compared with actual work: which steps can be delegated and which cannot.
  • Data and tools. Which information and which systems are really there, who can access them, what is missing.
  • Risks. Where an agent error would harm a person, the service or the company, and how hard it would be to go back.
  • Human control. Who can intervene, at which moments, with which information, and who is accountable for the result. On the handoff to a human operator we have a dedicated guide.

Where we start: the initial analysis

Before designing an agent we reconstruct the current state, with interviews with those who decide and those who work in the service, and with a reading of the processes and data already available. We also compare the alternatives without an agent. The result is a map of the current state and a map of opportunities: the situations in which an agent can help and those in which control must stay with people.

What you receive

You receive a map of the current state and a map of opportunities, to discuss with your team.

  • The situations in which an agent can help and those in which it would do harm, with the reasons.
  • For each useful situation: the role the agent could have, the human control needed, the tests to run before expanding its use, and the alternatives without an agent.
  • Priorities and the recommended first step.

Format, duration and ways of working are defined in the quote, on request. We do not show a price on the page.

How it continues

After the assessment there are three paths, and the choice depends on what emerges from the map. If there is an agent to design, an agentic design workshop on the Agentic Design Canvas, a method by Luigi Greco, is used to set objectives, context, delegations and limits before writing code. If an agent already exists, we test it with Testing AI agents with real users. If the whole service needs to be rethought, we continue with a project that holds together people, processes, agents and human control, with your team and those who build the systems: this is the work described in «Make an artificial intelligence agent reliable in production».

Frequently asked questions

Do you already need to have an agent to do the assessment?

No. It is designed precisely for those who still need to decide whether and where to use one.

How long does it take and what do you deliver?

We deliver a map of the current state and a map of opportunities, to discuss with your team. Format, duration and ways of working are defined in the quote, on request.

What is the difference from testing agents with real users?

The assessment is used to decide where to use an agent. The test is used to verify an agent that already exists, before expanding its use. The method of the test is also described in «Testing AI agents with real users».

Tell us which decision you need to take on agents and which processes are involved. We reply from our offices in Rome and Milan.

Let's talk about your agent