From imagining software to experiencing it before it exists
For more than two decades, Agile has transformed the way software is built. Its principles emerged in response to a world where requirements evolved continuously, long development cycles created unnecessary risk, and organisations needed mechanisms to learn while delivering value. By breaking work into small, working increments, Agile enabled teams to reduce implementation risk, gather feedback early and continuously refine their products.
Prototyping has followed a similar objective. Long before generative AI, paper sketches, wireframes, clickable mock-ups and interface design tools were widely used to explore ideas, capture requirements and help stakeholders understand a problem before committing to implementation. These techniques have always been valuable because they encourage discussion at a stage where change is still inexpensive.
The limitation was never the concept of prototyping itself. It was the cost of realism. Building a prototype capable of reproducing the behaviour of an entire system quickly became almost as expensive as building the system itself. Once realistic interactions, business rules, workflows and representative data had to be incorporated, organisations naturally limited prototypes to isolated journeys or interface concepts. They were sufficient to validate individual features, but rarely rich enough to allow people to experience how the complete solution would operate.
Generative AI is changing that economic balance. Today, complete functional simulations can be assembled in hours using AI-assisted development, mocked services, in-memory data structures and synthetic datasets. These simulations are not intended to become production software, yet they behave realistically enough for stakeholders to interact with an end-to-end experience. As the cost of creating and modifying these simulations falls, it becomes possible to use them for something that was previously difficult to justify: understanding the organisation before building the software.
A Practical Experience
This idea became much clearer during an internal initiative carried out over the past week. In roughly fifteen hours, we developed a functional simulation of a complete platform intended for discussion with members of the executive team. From the outset, the objective was not to optimise architecture, security or scalability. The prototype only needed to be sufficiently realistic for people to interact with it naturally and recognise their own processes within it.
The initial discussions were exactly what one might expect. Participants commented on screens, workflows and individual interactions. However, once they had spent some time navigating the complete experience, the conversation immediately moved away from the software itself. Instead of discussing how a particular feature should behave, the focus shifted towards organisational questions: Why is this capability necessary? Who should own this process? Does this workflow reflect how the organisation actually operates? Which department benefits from this information, and who is responsible for maintaining it?
That change in perspective proved to be the most valuable outcome of the exercise. Requirements that initially appeared obvious became open to discussion. Hidden dependencies between departments emerged naturally. Some responsibilities were reassigned as people recognized they belonged elsewhere, while several seemingly important requirements disappeared because nobody could clearly explain the value they created once the entire process became visible.
Looking back, the prototype itself was not the main deliverable. Its real contribution was providing a shared environment where different perspectives could be explored against the same representation of reality.
From Validating Features to Exploring Organisations
This experience led to an interesting reflection. Traditionally, prototypes have been regarded as tools for validating requirements or testing interface designs. AI does not change that purpose, but it significantly expands what a prototype can represent. Instead of showing isolated parts of a solution, it becomes feasible to simulate an entire organisational experience and observe how people respond to it.
That difference changes the nature of discovery. Previous generations of prototypes required stakeholders to imagine how the complete system would behave by looking at fragments. Today's functional simulations allow them to experience that behaviour directly. As a result, discussions become grounded in observation rather than speculation. Participants notice inconsistencies because they encounter them while completing realistic activities, not because someone described them during a workshop.
This also changes the kinds of questions that emerge. Rather than concentrating exclusively on functionality, stakeholders begin discussing governance, decision-making, ownership, organisational boundaries and the way value flows across the institution. The conversation naturally moves from "What should the software do?" to "How should the organisation work?" That shift is subtle, but it has important consequences for projects where success depends on coordinating multiple actors rather than satisfying a single decision-maker.
Cheap to Change, Easier to Learn
One consequence of AI-enabled prototyping receives relatively little attention. When creating and modifying realistic simulations becomes inexpensive, it also becomes inexpensive to broaden the conversation.
Traditionally, incorporating additional stakeholders late in the discovery process carried a cost. New perspectives often implied redesigning prototypes, revisiting requirements and reopening decisions that teams considered settled. For understandable reasons, projects tended to stabilise the participant group early and progressively narrowed the discussion as implementation approached.
That trade-off is beginning to disappear. If adapting a prototype takes hours instead of weeks, incorporating another department or governance body becomes a relatively small investment. As new stakeholders are identified, their reality can be represented in the simulation, allowing them to react to the complete experience instead of isolated documentation. Their feedback enriches not only the prototype but also the organisation's understanding of the initiative.
This creates a very different dynamic for stakeholder management. Alignment no longer depends exclusively on meetings, documents or requirement reviews. Instead, it emerges progressively through interaction with a shared experience. Responsibilities are negotiated while people observe the consequences of different decisions. Organisational bottlenecks become visible. Hidden dependencies surface naturally. Potential conflicts appear while they remain inexpensive to address, allowing the project to enter later phases with a much stronger foundation.
In this sense, the prototype stops being merely a representation of the future solution. It becomes a temporary space where the organisation negotiates what that solution should become.
A Complement to Agile, Not a Replacement
None of this suggests that AI-powered prototyping replaces Agile. In fact, both approaches pursue remarkably similar objectives: reducing uncertainty through iterative learning. The difference lies in the type of uncertainty they address and the moment in the project lifecycle where they create the greatest value.
Agile assumes there is sufficient understanding of the problem to begin delivering software. Its strength lies in reducing implementation uncertainty through short iterations, continuous feedback and incremental delivery. Experience prototypes operate earlier. Their purpose is to reduce uncertainty about the problem itself by exposing organisational assumptions, clarifying responsibilities and helping stakeholders converge on a common understanding before development accelerates.
Seen from this perspective, the two approaches complement each other naturally. Experience prototyping helps organisations understand what they should build and why, while Agile provides an effective framework for learning how to build it successfully. One optimises organisational learning during discovery; the other optimises product learning during implementation.
Where This Approach Creates the Most Value
This approach is unlikely to be equally valuable in every context. Products composed of loosely coupled capabilities, where teams can deliver independent increments with limited organisational interaction, may obtain little additional benefit from simulating the entire system. In these environments, Agile's traditional delivery model remains highly effective.
The picture changes in organisations characterised by distributed governance, multiple decision-makers and complex stakeholder ecosystems. Universities, public administrations, healthcare organisations and large enterprises frequently operate in this way. In such environments, technical implementation is often not the greatest source of project risk. Achieving a coherent understanding of responsibilities, priorities and value across the organisation is usually far more challenging.
For these projects, whole-system experience prototypes offer an opportunity to surface disagreements much earlier. Rather than discovering conflicting expectations during implementation, organisations can explore them while change is still inexpensive and while the prototype remains an effective instrument for discussion rather than a commitment to a technical solution.
Final Reflections
The most significant contribution of generative AI to project management may not be that it helps us produce software faster. Its greater contribution may be that it enables organisations to learn together earlier and at a much lower cost than before.
Prototypes have always existed to support discovery, but realistic simulations were traditionally too expensive to justify except in a handful of projects. That constraint is rapidly disappearing. As a result, prototypes can evolve from being simple validation artefacts into environments where stakeholders explore processes, negotiate responsibilities, incorporate new perspectives and progressively construct a shared view of the value they want to create.
Perhaps this is where AI will have its greatest impact. Not by replacing Agile or changing the principles of iterative delivery, but by strengthening everything that happens before implementation begins. If organisations can experience a future solution together, they are far better positioned to align expectations, resolve disagreements and make better decisions before writing production software.
In many complex procjects, that shared understanding may ultimately prove to be the most valuable deliverable of the entire discovery phase.
Comments
Post a Comment