Every software vendor is talking about AI. The more important question for utilities and energy providers is what actually changes for customers. At exnaton, AI isn't replacing engineers, designers, or product managers. It's helping them move from ideas to validated solutions faster without lowering the quality, security, or review standards our customers expect. Across product design, engineering, data science, and product management, AI now helps us prototype faster, automate repetitive work, document releases more clearly, and evaluate new ideas in days instead of weeks.
To better understand how this looks in practice, we spoke with four members of the exnaton team to understand where AI makes a real difference, where human expertise remains essential, and what this ultimately means for the utilities using our platform.
Every new feature begins with an idea but before customers can provide meaningful feedback, they need something tangible to react to.
That’s where AI has fundamentally changed the way our UX design team works.
Instead of spending days refining static wireframes, designers can now create interactive prototypes within hours. This allows colleagues and customers to experience workflows much earlier in the design process, making conversations more concrete and feedback significantly more valuable.
"Before, we relied on sketches and wireframes, and stakeholders often misunderstood them," explains Irene Muñoz, UX Designer at exnaton. "Now we can go straight to prototyping. We can explore several working prototypes in the time it used to take to polish one static design."
The UX design team uses Claude to structure ideas and brainstorm before generating and iterating on prototypes in Figma Make. Behind the scenes, engineers maintain shared Markdown documentation containing colors, typography, and design tokens. This gives AI a consistent understanding of how exnaton products should look.
AI also helps the team synthesize interviews, adapt content for different languages, and create reports more efficiently.
But despite these improvements, AI cannot replace the honest, sometimes challenging feedback that comes from real users.
"LLMs are too agreeable," Irene says. "They don't give you the constructive friction you need to design a good user experience. We still validate every concept with real users, not synthetic ones."
For our customers, this means product ideas can be tested and refined much earlier, reducing misunderstandings before development begins.
Once a concept has been validated, engineering takes over. Interestingly, AI hasn't reduced the importance of planning.
Rather than immediately writing code, engineers now spend more time defining exactly what should be built and how every component should behave. Once those specifications are reviewed, AI can help generate implementation, testing, and documentation.
"A lot of the time has moved into the specification," says Sergi Gimenez, Software Engineer at exnaton. "You invest in a precise plan of what will be built and how. Once that plan is reviewed and right, the AI can implement it, write the tests, and document the result — and two people still review every change before it reaches our codebase."
AI also supports communicating product changes more clearly. Every merged software change is automatically summarized in plain language, providing the basis for the bi-weekly release notes shared with customers.
The same acceleration is visible in data science.
Hanna Paulava, Machine Learning Engineer at exnaton, has seen a similar shift in forecasting work.
"Last year I rolled out our first larger forecasting pipeline in two to three months, essentially as a team of one. Eight years ago, a comparable project took a team of three about a year and a half."
For Hanna, the larger change is that generated code can be evaluated and replaced much more easily without changing the standards it must meet.
"The code is an artifact now, not the product. If we don't like what the AI wrote, we throw it away and generate it better. What never changes are the reviews, the tests, and the pipelines every change has to pass."
The outcome for customers isn't simply faster development. It means improvements, fixes, and new capabilities can move through development more efficiently while maintaining the same engineering standards.
AI is also changing how product managers spend their time.
Instead of manually producing documentation, comparing feature requests with technical capabilities, or preparing exploratory analyses, much of this work can now be completed in minutes, allowing product teams to focus on customer conversations and product strategy.
One example is an internal AI that navigates the exnaton Backoffice, captures workflows – such as automated bill runs – and produces documentation drafts using exactly the same German interface terminology customers see inside the product.
"Documentation used to be full days of work," says Gabrielle Every, Product Manager and Energy Market Expert at exnaton. "Now I have a solid draft in 15 minutes and can spend my time on things only humans can do."
Those drafts still require careful review and refinement, but they provide a strong starting point that is available within minutes rather than days.
AI has also become a useful way of identifying opportunities to improve the product itself.
When Gabrielle found herself using it to configure complex Time-of-Use tariffs in our exnaton tariff editor, it became clear that the workflow needed to become more intuitive.
"If we need AI to operate part of our own product, that's a signal the UI needs to improve."
That insight directly led to a redesign of the tariff editor.
Other internal AI skills are connected to exnaton’s OpenAPI specification, allowing product and sales teams to compare customer feature requests with existing platform capabilities and provide realistic feasibility assessments much earlier in the sales process.
AI also makes it faster to explore and communicate data. For example, a dashboard visualizing the CO₂ intensity of customer consumption can now be created in an afternoon rather than requiring an entire development sprint.
Throughout all of this, one thing in product management remains unchanged.
"AI is not a replacement for people. It gives us the space to be creative but understanding a customer's real pain point, and the root cause behind it, is still our job." – says Gabrielle.
Using AI every day internally has also shaped how we think about customer-facing AI.
Many of the lessons learned while developing internal AI workflows are reflected in Wattson, exnaton's AI assistant for utilities. Whether helping users navigate complex workflows, answering product questions, or making platform knowledge easier to access, the same principle applies: AI should reduce friction, not replace human expertise.
By using AI ourselves every day, we continuously learn how to build AI experiences that are genuinely useful for utility employees.
For customers, the biggest impact isn’t that we use AI. It’s what AI enables us to do.
Utilities benefit from faster prototypes and feasibility assessments, more frequent product improvements, clearer documentation and release notes. It also gives more time to solve customer problems, explore new customer facing experiences including personalized consumption insights and further developments around Wattson.
Importantly, faster doesn't mean less rigorous.
The same review, testing, access, and data-protection requirements apply regardless of how a change is produced.
Human review. Every software change still begins with a reviewed specification. Every implementation is reviewed by another engineer. Nothing is merged or deployed autonomously.
Deterministic quality checks. Automated tests, linters, and staging environments must all be passed before a change reaches production.
Limited access. AI agents receive only the permissions required for their task. Where they interact with databases, their access is read-only.
Data protection. Real customer data is not entered into third-party AI tools, and data used to train forecasting models is anonymized. These practices are part of our broader information-security approach, formalized through our ISO/IEC 27001:2022 certification.
These aren't special rules introduced because of AI, they're the same engineering and security standards that have always guided how we build software.
Ultimately, we don't measure AI by the number of prompts written or lines of code generated.
We measure it by outcomes: better software for utilities, faster delivery of customer-requested improvements, clearer documentation, and more time spent solving real customer problems.
AI helps us move from ideas to validated solutions more quickly but the decisions, accountability, and responsibility remain firmly with people. That's how we believe AI creates lasting value for utilities.
Want to explore how exnaton can help you launch modern energy products fast with AI-driven workflows? Contact us.
