A framework for adopting GenAI

Generative AI tools tend to arrive at organizations before there is a clear idea of how to bring them into the work. First comes the access. Then an early trial. After that, the expectation that use will sort itself out.

It rarely does.

When a pilot loses momentum, the explanation is rarely only the quality of the model. The clarity of the use case also weighs in, along with how the tool fits into existing processes, the team’s trust in its results, and the organization’s capacity to supervise what happens after the deployment.

That is why we look at GenAI readiness as a set of conditions rather than an isolated technical decision. What follows is how we are ordering those questions inside a working instrument, designed to help decide what should be ready before moving forward.

A profile, not a score

AI readiness tends to get summarized in a single score. It is practical: it lets you compare, prioritize and report progress in a simple way. The problem is that one number rarely shows where the real friction lives.

A use case can be well defined and still fail for lack of trust, for a poor integration, or because there are no clear criteria for supervising the results. The reverse happens too: there is interest, technical capacity and internal support, but infrastructure or governance still cannot sustain the deployment. A composite or arithmetic-mean score smooths over those differences exactly when seeing them clearly matters most.

At Tressia Labs we propose a profile. Six dimensions observed in parallel, each with its own reading. The point is not a high score but understanding which condition needs work before moving forward.

Critical Moderate Enabling

Usefulness It solves a real clinical problem

Enabling

Ease of use Learning it fits into the routine

Moderate

Trust People act on the results

Critical

Literacy Users notice when something is off

Moderate

Governance There are policies, reviews and owners

Enabling

Climate The close team backs the change

Critical

The six dimensions

We group the dimensions into three blocks because each one observes a different part of adoption. The first looks at the experience of the people who will use the tool. The second, at the organization’s capacity to sustain it. The third, at the rules that allow it to be decided, supervised and corrected.

The user experience

Usefulness

The first question is simple: what problem does it help solve. A GenAI tool can save time, improve a task, support a decision, or open a more precise way of working. That usefulness has to show up in daily practice. When the use case only works in the pitch, adoption fades quickly. TAM · UTAUT

Ease of use

Adopting a tool also implies effort: learning to use it, fitting it into a routine, and understanding when it is appropriate to apply it. When that effort is too high, the problem is usually in the design of the process, before the tool itself. UTAUT · Effort Expectancy

Trust

Trust determines whether use is sustained over time. It covers trust in the system's results, in the organization that deploys it, and in the team's ability to review, question or correct what the tool produces. Without that base, use stays limited to isolated tests. Steerling 2023 · Cicin & Gürkan 2026

Organizational capacity

Literacy

Not everyone needs to understand how a model is trained. People do need enough criteria to interpret its results, recognize frequent errors, and know when to ask for review. That critical reading reduces dependency and improves the quality of decisions. As'ad & Faran 2025 · WHO 2021

Climate

Adoption also depends on the immediate environment. When the team feels supported, given time to learn, and permission to flag problems, the tool stands a better chance of being integrated usefully. When use is imposed without a conversation, resistance appears early. FAIIR-H · CFIR

Infrastructure

This dimension requires a technical review. Data quality matters, along with integration with existing systems, source traceability, security, and the mechanisms to detect performance degradation. Without those conditions, even a strong use case can become fragile. Marteau 2025 · FAIIR-H

The decision frame

This block defines how the organization responds when the tool stops behaving as expected, produces unintended effects, or demands a decision beyond the team using it.

Governance

The central question is who decides and with what authority. Before deploying a tool, an organization should have a live policy, a risk-assessment process, and a space responsible for approving, adjusting or suspending its use. Governance does not slow down adoption; it gives it the conditions to move with judgment. Wells 2025 · NIST RMF

Monitoring

A GenAI tool needs follow-up after deployment. Clear indicators, periodic review, channels to report problems, and an improvement process allow you to know whether it is still adding value. Monitoring turns adoption into a revisable decision rather than a permanent bet. Wells 2025 · RE-AIM

Ethics

Autonomy, safety, transparency and equity work as decision filters. They are not a decorative layer or a separate score. They serve to check whether the intended use respects the people affected, distributes risks well, and allows the way decisions are made to be explained. WHO 2021

Putting people at the centre

What the clinical case taught us

The clinical case was a specific vertical, but it left a broader lesson: a GenAI tool is best understood when you look at it from the person who will live with its consequences.

In that context it was the patient. In others it can be a student, a citizen, a worker, a customer, or an internal team. The category changes with the sector, but the question stays the same: what actually improves for the person affected by this decision, this process, or this recommendation?

At first we considered reserving a separate dimension for the patient. We then saw that the solution impoverished the framework. The affected person does not belong inside a box; they cross several. Inside trust, because the tool changes the relationship between the one who decides and the one who receives the decision. Inside literacy, because an undetected error rarely stays inside the system. Inside governance, because rules designed without those voices tend to arrive late to the risks that matter.

The real test of GenAI comes when it touches the experience of people.

For each use case, three things are worth doing from the start: include people with direct experience of the process, observe what changes for them while the pilot runs, and keep a clear human responsibility over the final decision. That centrality is not an ethical add-on. It is a more precise way of designing, evaluating and deciding.

How it is applied

We apply the framework as a diagnostic protocol. It works with three lines of evidence read in parallel: the user experience, the technical base, and the organization’s decision frame.

The first line gathers how the use case is understood: what problem it tries to solve, who will use it, what changes it introduces in the work, and what effects it can have on the people affected. Here perceived usefulness matters, alongside ease of integration, trust, and the team’s capacity to interpret the tool’s results.

The second line reviews the technical conditions. Available data, source quality, integration with existing systems, traceability, security, and mechanisms to detect performance loss. This part requires conversations with technical teams and documentary review. The goal is to separate a viable trial from a fragile implementation.

The third line observes the institutional capacity to decide and respond. Live policies, risk-assessment criteria, defined owners, supervision channels, and the authority to adjust or stop use when context requires it.

With those three inputs we build a readiness profile. The result is not an automatic approval but a situated reading: which conditions are ready, which need work, and which decision is worth taking before moving forward. Sometimes the profile enables a pilot. Other times it points to a dimension that needs strengthening first.

That is the value of the method: it turns a general intention to adopt GenAI into an informed, traceable, and discussable decision.

Where this stands

This framework is in a refinement and validation phase. It builds on consolidated literature about technology adoption, implementation, trust, governance, and AI evaluation, and combines it with criteria developed through our applied work with GenAI use cases.

The methodological agenda is open: expert review of the dimensions, content validation, cognitive testing with users, and application across different organizational contexts. We are interested in observing how the profile behaves when it moves from a conceptual conversation into an actual adoption decision.

We share it at this stage because it helps to formulate, more precisely, the question many organizations have in front of them: how to move a GenAI project forward with method, evidence, and organizational sense. We want to test it with teams working at that frontier who can contribute evidence, methodological judgement, or application cases.

If you work at this frontier of knowledge and want to try, discuss, or contribute to the validation of this instrument, let’s talk.