Case Studies

Commercial organisation — integrating two sales networks

Two sales networks built around different logics, channel-based responsibilities and conditions that were not always aligned were making territory management increasingly complex, both internally and in relationships with agents and customers.

The work started with the structure: moving beyond a strongly centralised model built around channel managers and shifting towards a model closer to the territory, with Area Managers given clearer responsibilities and an overall view of their respective areas.

In parallel, commercial conditions, discount structures and exceptions accumulated over time had to be addressed, seeking a balance between commercial continuity, sustainability and ease of management.

The objective was not simply to merge two networks, but to build a more readable system: fewer overlaps, clearer responsibilities and more consistent commercial terms for agents and customers.

The result was a model that was easier to manage internally and easier for agents and customers to understand, with less ambiguity in day-to-day operations.

E-commerce — building a sustainable assortment

Moving a traditional product catalogue into e-commerce does not simply mean publishing as many products as possible.

In this case, the starting point was a very broad assortment, with products that differed significantly in price, availability, logistics costs and economic sustainability. Some apparently attractive products became uncompetitive once shipping, handling and real margin were taken into account.

The work consisted of defining clear selection criteria, analysing products one by one and separating what could genuinely work online from what would generate unsustainable sales or operational problems.

Price, stock, delivery costs, margin and product characteristics were cross-checked to create a narrower assortment that was much better aligned with the channel.

The result was not more products online, but the right ones: an assortment that was easier to manage, more competitive and, above all, sustainable over time.

People — helping someone regain direction

When someone loses engagement, motivation is not always the problem.

In this case, a person who had previously shown ability and reliability had progressively reduced initiative, participation and the quality of their contribution. The simplest reaction would have been to increase control or interpret the behaviour as a lack of interest.

Instead, the work started with direct conversation and listening, trying to understand what had changed in the relationship between the person, the role and the context.

Elements emerged that would hardly have appeared through targets, reports or KPIs: loss of clarity, a reduced sense of impact and growing distance from decisions.

The path was then built together, redefining responsibilities, expectations and room for autonomy.

It was not about “motivating” someone, but about removing what was preventing them from working well again.

The most important result was not only the recovery of performance, but the return of initiative and personal responsibility.

Software / Beta testing — putting a new product in front of the market

A product can work perfectly from a technical point of view and still have problems when it meets real users.

In this case, the work involved a platform still under development, used in real conditions to identify issues that did not emerge in traditional testing.

Testing was not limited to checking whether the expected functions were present. It looked at the product as a whole: unintuitive flows, inconsistencies between interface and logic, updates that changed previous behaviours and situations in which a user could be left without understanding what was actually happening.

Each issue was documented from the user experience: what I was trying to do, what I expected to happen, what actually happened and what consequence that would have for a normal user.

The value of beta testing, in this case, was not simply finding bugs.

It was bringing the market’s perspective into the development process before the product truly reached the market.

AI — Narrative Red Teaming of conversational AI systems

Testing an artificial intelligence system is not just about trying to make it say something it should not.

In conversational Red Teaming, the interesting problems appear when the system is taken outside ideal conditions: long conversations, ambiguous requests, role changes, partial information, pressure from the interlocutor and attempts to progressively move the boundaries of the original rules.

The work was carried out across different conversational models, including Gemini and Claude, using realistic narrative scenarios rather than isolated prompts.

The objective was to observe how the system behaved over time: whether it maintained its rules, started contradicting itself, invented information to preserve conversational coherence, or could gradually be led towards behaviours that a direct request would not have produced.

The method involved building a credible context, progressively evolving the conversation and observing the points at which the model began to drift from its original constraints: not necessarily through an obvious violation, but through small concessions, increasingly elastic interpretations or loss of the initial constraints.

AI was also used on the other side of the table, supporting scenario construction, response analysis and the search for contradictions.

This turned testing into an iterative process: attack, response, analysis, scenario adjustment and another attempt.

The result was not simply understanding whether a model was vulnerable to a particular prompt, but observing how its behaviour changed inside a realistic and sufficiently long conversation.

Because, often, a system does not break because of one wrong question.
It breaks after twenty seemingly harmless questions.