The People-to-AI Framework: How to Sequence AI and Technology Change

There is a persistent temptation in technology-led transformation to start with the technology. The platform is evaluated, selected and contracted before the people who will use it have been genuinely consulted. The result is predictable: technically sound implementations that the organisation never fully adopts. AI has made the temptation stronger, because the tools are impressive in a demonstration.

The People-to-AI framework is Ganexa's answer. It is a sequencing model: people and operating model first, then process, then data, then AI. Each layer is built on the one before it. This article explains the order, the four phases we use to apply it, and how to tell whether your own programme is running in the right direction.

Why sequencing matters

The framework rests on a simple insight: the value of any technology is determined by the humans who use it, not by the technology itself. A forecasting model is worth nothing if planners override it out of habit.

This means the sequence in which you engage people, build capability and introduce tools is not a project management choice. It is a value-realisation choice. Programmes that start with the tool spend their later months retrofitting ownership, redesigning process under pressure and cleaning data in a hurry. Programmes that start with people do the same work in an order where each step makes the next easier.

The People-to-AI framework: people, process, data, then AI

The framework has four layers, addressed in a fixed order.

• People and operating model. Who owns the outcome, who decides, who does the work and how their roles, incentives and skills will change. If accountability for an AI-assisted decision is unclear, no amount of engineering will fix it.
• Process. How the work actually flows today, where the friction lives and what the redesigned flow should be. Automating a poor process produces a faster poor process.
• Data. Whether the information the redesigned process needs exists, is accurate, is accessible and is governed. Only now do you know which data matters, because the process tells you.
• AI. The models, assistants, agents and platforms, chosen to fit the people, process and data already defined.

The order does not mean each layer must be perfected before the next begins. It means decisions in a later layer are never made without the earlier layers being understood. Prototype a model in week two if you wish. Do not contract a platform before you know who will own the process it supports.

AI is not a replacement for people. It is an amplifier. And like any amplifier, what it amplifies depends entirely on what you feed into it.

The four phases of the framework

The layers describe what to address. The phases describe how a programme moves through them.

Phase 1: Understand. Listen before you build

The first phase is discovery, not of systems but of people. Who are the key roles? What do they actually do each day? Where does friction live in their workflows? What are their legitimate concerns about change? This phase maps the operating model and the process as they really are, not as the procedure manual describes them. It produces a people-centred problem statement that the technology must address, and the KPIs by which success will later be judged.

Phase 2: Enable. Build capability before you deploy

Before any new tool is introduced, the people who will use it need to understand the why in terms that are personally relevant to their work. This is not a training programme. It is a capability and confidence-building programme. It includes awareness, hands-on exploration and the psychological safety to make mistakes. It is also where the redesigned process is agreed with the people who run it, and where data gaps found in Phase 1 are closed. Managers go first, because they will be asked to judge AI-assisted work before their teams produce it.

Phase 3: Deploy. Introduce tools into prepared soil

Technology deployment into a team that has been genuinely prepared looks different from a standard rollout. Resistance is lower, workarounds are fewer, and the feedback loop between users and the deployment team is richer. Iterations are faster and more targeted, because users can say precisely what is wrong and why. Deploy in slices: one team, one process, one measurable outcome at a time.

Phase 4: Sustain. Embed, measure and evolve

The final phase closes the loop between deployment and value. Are adoption rates meeting targets? Are the KPIs set in Phase 1 moving? What has the organisation learned that should shape the next wave of change? Sustainability is not about locking in a solution. It is about building the muscle to keep evolving, which matters more with AI than with any earlier technology because the tools themselves keep changing.

How to tell your programme is running backwards

Most organisations drift into running the sequence in reverse. The warning signs are consistent:

• The vendor was chosen before the use case had an owner. A platform is in place and the search for problems it can solve has begun.
• Success is defined as go-live. Nobody has written down which business measure should move, or by when.
• Training is scheduled for the week before launch. Capability is treated as a rollout task, not a precondition.
• Data problems surface during testing. The process was never mapped, so nobody knew which data mattered until the model needed it.
• Frontline staff first hear about the change in an announcement. Their knowledge of where the process really breaks has not been used.

None of these is fatal. Each is a signal to go back up the sequence. Many of the patterns behind why digital transformations fail are this sequence run in reverse.

Where to start: questions to ask your team

• Who owns the outcome? Name the business leader accountable for the result, not the project.
• What will people stop doing? If the answer is nothing, the change is an addition to workload, and adoption will suffer.
• Have we watched the work? Process maps are useful. Sitting beside the people doing the job is better.
• Which data does the redesigned process need? Assess that data, not the whole estate. An AI readiness assessment covers people, process, data and technology in one view.
• What is the smallest slice we can deploy and measure? Choose one team and one KPI, then widen.
• Who keeps it alive after launch? Agree ownership, measures and a review rhythm before deployment, not after.

Common questions

What is the People-to-AI framework?

It is Ganexa's model for sequencing technology change. It addresses people and operating model first, then process, then data, then AI, and moves a programme through four phases: Understand, Enable, Deploy and Sustain. The aim is adoption and measurable business value, not just a working system.

Does starting with people slow an AI programme down?

It moves effort earlier, which can feel slower in the first weeks. In practice the discovery and enablement work is modest compared with the cost of a deployed tool that nobody uses. Phases can overlap, provided no irreversible technology decision gets ahead of the people and process work.

How is this different from change management?

Traditional change management is often brought in after the solution has been chosen, to help people accept it. People-to-AI puts the people work before the solution is chosen, so that it shapes what gets built. Change management remains essential. It simply starts on day one.

Where Ganexa can help

People-to-AI is how Ganexa runs every engagement, from defining the AI business case and operating model to organisational change and AI adoption and AI strategy and implementation. If you want a second opinion on the sequence of your own programme, book a free consultation.