From Experiment to Engine: How a Circular Electronics Retailer Turns AI into Measurable Growth

At E-Connect Europe 2026 – Power of 50, Martin Roulund discussed how clear data and governance help GreenMind harness the value of AI in customer journeys, replenishment, and sales assistance.

The experience highlights how even relatively small organisations can move faster and more effectively with AI than many global industrial leaders, provided they focus on fundamentals like customer experience, data quality, narrow use cases, and governance that enables rather than blocks experimentation.

Why B2B Needs to Think Like B2C  

Industrial and aftermarket leaders serve professional buyers who are also everyday consumers. The same person who orders complex equipment or spare parts on weekdays is, in the evening, buying clothes, electronics, or groceries via frictionless, intuitive digital experiences. Those expectations do not simply switch off at the office door. 

• Customers expect simple, self-service digital experiences;
• They still value human interaction for complex or high-stakes decisions;  
• They want the freedom to move seamlessly between channels without losing context.

This is the “rule of thirds”, often appearing across industries, cultures, and markets: roughly a third of customers want pure digital self-service, a third prefer human interaction, and a third expect a genuine omnichannel mix.

B2B does not need to copy consumer e-commerce interfaces one-to-one, but to translate B2C principles, such as simplicity, consistency, transparency, and contextual relevance, into the industrial buying journey, from trigger and research through evaluation, purchase, and after-sales.

Mapping the Journey: From Trigger to Retention  

A simplified customer journey model is the backbone for much of the thinking:

1. Trigger – a need emerges (a device fails, a fleet must be upgraded, a project launches);  
2. Research – suppliers are identified, information is gathered, options are shortlisted; 
3. Evaluation – trade-offs are assessed (price, quality, sustainability, service, trust); 
4. Purchase – the transaction is executed through a chosen channel;
5. Retention – after-sales, service, and support determine whether the relationship deepens or ends.

Most organisations recognise these stages. However, when data and AI are introduced at every stage:

• Enhanced presence: Being visible and relevant at each stage significantly increases the likelihood of being chosen. When a supplier is present throughout the journey, the probability of purchase can rise significantly compared to competitors who appear only at isolated touchpoints.

• Reduced fragmentation costs: Connecting customer interactions across channels enables organisations to unify e-commerce and sales-driven transactions into a single relationship, preserving context, unlocking upsell opportunities, and strengthening loyalty.

• Maximised after-sales value: Supporting retention and expansion through digital tools such as portals, self-service, account-based marketing, and AI-powered assistance allows organisations to grow wallet share more efficiently and profitably than relying on constant acquisition.

Organisations can map the B2C journey in detail, including all digital touchpoints, and then apply those lessons to B2B. This helps understand behaviour, friction points, and what “good” feels like for the same human in a different buying role.

Data as the Centre of Gravity  

All of this can be constrained or enabled by data.

GreenMind’s operating model is not manufacturing, but sourcing, buying used electronics from end-users and brokers across Europe, refurbishing them, and reselling them. Profitability is built by buying and stocking the right products, at the right time, in the right locations, and then selling them through the right channels. Data about products, locations, transactions, and customer behaviour is therefore central.

• Centralisation without illusion: All key data, including e-commerce, point of sale, CRM, customer records, and financials, flows into a single ERP platform. This provides a unified view, but does not guarantee high quality.

• Data quality is non-negotiable: Inconsistent naming conventions, disconnected store and stock identifiers, mixed use of upper/lower case, and other basic hygiene issues can make data unusable for AI. Before models can be effective, organisations need to invest in cleaning, normalising, and governing data.

• Data must be actionable: It is not enough to accumulate records. Data has to be structured and accessible in a way that can power specific use cases, such as replenishment algorithms, chatbots, recommendation engines, or account-based initiatives.

Despite its size, the company uses an ERP suite comparable to typical enterprise deployments and leverages it as an operational backbone, intertwining ERP, CRM, e-commerce, POS, and product data. This architectural choice enables relatively quick development of AI agents that sit on top of existing data flows.

Internal vs. External AI: A Simple But Powerful Distinction  

GreenMind focuses on two perspectives for AI:

• Internal AI: Aimed at operational efficiency, process optimisation, and decision support. These are typically back-office tools that make teams more productive or help extract more value from existing resources.

• External AI: Customer-facing or market-facing solutions that directly shape the customer experience, capture new demand, and drive revenue.

This distinction guides:

• Prioritisation: External AI use cases are closer to revenue and experience, while internal AI often delivers faster cost and efficiency benefits. Both are needed, but the balance depends on strategy.  

• Governance: Customer-facing AI demands strict guardrails around data exposure, accuracy, and brand impact. Internal AI, while still sensitive, may allow more experimentation.  

• Communication: internal AI requires change management with employees, while external AI requires careful message control with customers.

Instead of pursuing dozens of scattered experiments, the company builds a long list of potential use cases and then applies a prioritisation model, such as RICE (Reach, Impact, Confidence, Effort), to select just a few to start with. This enables fast learning without diluting resources.

From Experiments to Impact: Early AI Use Cases in Practice

A couple of early use cases illustrate how GreenMind is translating these ideas into tangible outcomes.

The first is a replenishment tool designed to optimise how products are distributed across stores. The use case is relatively simple, combining internal data, such as sales, stock levels, and product attributes, with a rules-based model to recommend where inventory should be placed to maximise revenue. By using AI-assisted coding, teams were able to move quickly from idea to working prototype, bringing business and technical stakeholders closer together in the design process. Instead of lengthy requirement cycles, they iterated directly with users, refining the tool in short loops. Now in production, it is delivering measurable impact by better aligning inventory with local demand. It could easily extend to areas such as spare parts or regional stock optimisation in industrial settings.

The second use case started as a typical chatbot initiative, aimed at reducing pressure on customer service teams. However, as prototypes were tested against real user journeys, its potential quickly expanded. By capturing user needs in natural language, such as budget, intended use, or preferences, and combining this with structured product data, the assistant can guide customers through large catalogues, recommend relevant products, and even support availability checks and reservations. It begins to replace traditional search and filtering with a more intuitive, conversational experience.

This changes how the organisation views the tool, not just as a way to deflect support queries, but as a potential driver of conversion and revenue. While still in testing, it already highlights the importance of strong data foundations and of grounding responses in reliable internal sources.

Relatively focused initiatives can create broader momentum, moving from isolated tools to capabilities that reshape both internal ways of working and customer interaction.

Governance as an Enabler, Not a Brake  

Across both use cases, governance is treated as a living system rather than a rigid checklist.

• Executive sponsorship: AI initiatives are explicitly sponsored from the top, with senior leaders involved in risk assessment and go/no-go decisions for each use case. This avoids shadow experiments and ensures alignment with strategy.

• Cross-functional AI team: A small core team combines AI expertise, business analysis, and IT/data governance. Their role is to translate business problems into AI use cases, assess impacts, and work with operational teams.

• Business-led ownership: Each use case has a business lead who acts as an internal client. This is critical for adoption. Tools are not thrown over the fence from IT, but co-created with the teams who will use them.

• Iterative risk framing: For every project, key risks (data exposure, hallucinations, operational disruption) are identified early and periodically revisited as tools evolve. Governance is updated as experience grows.

Data governance, in particular, remains conservative on sensitive topics:

• Customer data is not exposed through the chatbot; only product and transactional information already available in public reports or catalogues is used.  
• IT retains final authority on data usage, with recurring feedback loops throughout development.  
• Guardrails between internal data, the AI model, and external interfaces are continually refined.

Governance should provide a safe framework within which experimentation can happen quickly, with risks consciously accepted, mitigated, or rejected.

Start Small, Learn Fast, Scale What Works  

Perhaps the most important message for industrial and manufacturing leaders is methodological rather than technical.

Several practices underpin the speed and impact seen in this case:

• Start with narrow, well-scoped use cases: Instead of chasing transformational AI, focus on specific problems where value can be measured. Success in small areas builds political and organisational capital.

• Accept that data work is unavoidable: AI exposes the quality or lack of underlying data. Cleaning, standardising, and connecting data sets is foundational, not optional.

• Use AI to design AI: AI-assisted coding and prompt engineering can fundamentally change how requirements are captured and prototypes are built. Business experts can engage far more directly in solution design.

• Embrace continuous iteration: Interfaces, prompts, and features should be tested directly with users, adapted, and redeployed in short cycles. This is as true for internal tools as it is for customer-facing assistants.

• Recognise that AI is evolving weekly: Tools such as AI coding assistants or LLM platforms change rapidly. Governance, skills, and architectures must stay flexible enough to adapt.

For larger manufacturers and B2B organisations, the agility of a smaller player may seem hard to replicate. But the underlying principles are transferable. Breaking down silos, forming small cross-functional squads, and allowing micro-units to operate with more autonomy can deliver similar speed within a bigger structure.

Implications for Industrial Transformation  

GreenMind’s achievements carry several broader implications for industrial and aftermarket leaders:

• Circularity and AI are complementary: In circular models, profitability often depends on data-driven decisions about sourcing, refurbishing, and stocking. AI can enhance visibility, prediction, and utilisation across those loops.

• B2C and B2B are converging in practice: Learning from consumer journeys and applying those insights to industrial contexts is no longer optional. Customers expect frictionless, omnichannel experiences backed by intelligent support.

• AI is already practical at small scale: Organisations do not need massive budgets or proprietary models to start. Off-the-shelf platforms, when combined with strong data discipline and focused use cases, can deliver measurable business value.

• Governance must evolve alongside capability: As AI tools mature and new use cases emerge, policies, risk frameworks, and decision-making structures need to be updated continuously, not once.

AI in industrial and aftermarket environments should not be approached as a monolithic transformation programme. It is better understood as a series of targeted, governed experiments that steadily reshape how decisions are made, how customers are served, and how value is created, from the replenishment of a refurbished device to the recommendation of the next best product.