Context

Enogrid is an energy company specializing in collective self-consumption: it helps electricity producers sell their output directly to consumers located nearby. Founded ten years ago, the company stands on two “legs”: on one side, software that makes it easier to study and manage electricity-sharing operations; on the other, a range of services (training, studies, support, turnkey management).

It is also a committed company, ISO 27001-certified, that handles sensitive data (billing, energy consumption, personal data of individual customers) and has followed a frugal development policy inspired by Green IT since it was founded. All good reasons not to roll out AI just any old way.

Challenges

  • Structuring what had started informally: with its strong tech culture, Enogrid wondered how to extend the use of AI to its business side and its engineering team, to work more efficiently. The topic was on the table, but the company had neither the time nor the skills to drive it internally. Enogrid realized that this was above all a matter of change management.
  • Guaranteeing security and compliance: with an ISO 27001 certification and personal data, every decision had to be approved from a security standpoint and from a GDPR standpoint, with the DPO (data protection officer).
  • Protecting the training of junior staff: with a young team, Enogrid wondered how AI would affect skill-building: if AI is “a super junior”, how do you keep training the seniors of tomorrow?
  • Staying competitive while keeping a reasonable use: in a market of volume rather than margin, AI is “a bit of an obligation” to bring more value to clients. But Enogrid wanted a use consistent with its culture of frugality.
“Shutting ourselves off from AI and acting as if it doesn’t exist could potentially hurt our employees’ careers.”
— Rémi Bastien, CEO, Enogrid

The solution: why Teamstarter?

Enogrid chose Teamstarter to support it along the whole chain: scoping, tool choice, policy, then identifying the tasks that could be delegated and creating skills* with the teams. All with a bottom-up methodology (one that starts from the front line), inherited from continuous improvement.

1. An inventory and an informed choice of tool

First step: map who uses what. The audit reveals limited but real shadow AI, that is, employees using personal licenses, the worst situation in terms of compliance. Teamstarter then provides a comparison of the solutions on the market, which lets Enogrid choose its LLM (AI model) on three criteria: the security guarantees, the maturity on agents to progress quickly, and the portability of skills to other providers if the market changes.

2. An AI policy and proper approvals

Enogrid writes its AI policy, which covers ethics, reasonable use and data. Every decision is approved with the DPO and from the ISO 27001 standpoint, looping back with its cybersecurity providers and its internal team to dissect the LLMs’ terms of use.

3. A step-by-step approach: master the skill before connecting it

Faced with the teams’ reluctance, Enogrid makes a pragmatic choice: not to connect the CRM and the other tools right away. Skills are first developed and tested with anonymized or manually extracted data, to master the technique. Once the skill is reliable and the doubts are cleared, it is connected to the tools.

The scope is framed too: AI is used on the business side (sales, engineering, back office), not on the core software business, where the most sensitive personal data lives.

4. Ideation workshops by team, to start from tasks rather than processes

Two teams take part: the sales team and the engineering team. Each team identifies its repetitive, time-consuming, low-value tasks, then breaks its processes down task by task to spot the pain points. A revealing exercise: after ten years in business, Enogrid rediscovers habits nobody had ever questioned, which could have been simplified even without AI.

The results

  • Back office / billing: automatic extraction and processing of the information needed to bill subscriptions and fill in quotes, an automation the developers kept deprioritizing. The result: time saved and fewer errors.
  • Sales / CRM: automatic enrichment of new contact records, fed by Livestorm and the mailboxes. Where a salesperson on duty used to spend two hours every week on it, the AI now produces and sends up a report. A person checks and corrects it, and the time saved goes into calls with clients.

* A skill is a reusable user guide you give to an AI: instructions and good practices describing how to carry out a specific task. Once it is created, there is no need to explain the method again: the AI applies it and produces a consistent result that meets the team’s expectations.