Customer Experience
Building an AI chatbot for personalized sustainability guidance
As sustainability regulations grow more complex, businesses need guidance that reflects their specific context rather than generic checklists. Agilytic partnered with a company helping businesses build sustainability plans to design an AI chatbot prototype that would turn a manual recommendation process into a faster, more personalized experience.
To protect confidentiality, we may alter specific details while preserving the accuracy of our core contribution.
Context and objectives
Our client helps organizations across Belgium on their sustainability journey, combining coaching, training, and practical tools to turn ambitions into action plans.
The team wanted to automate the process of identifying and recommending sustainability practices for its own clients. The existing approach relied heavily on manual expertise: turning each client's context and priorities into actionable sustainability recommendations took significant time and effort.
The challenge was to make this recommendation process more efficient and consistent, without losing the tailored, expert feel that made it valuable in the first place. The objective was to build a working prototype of an agentic AI chatbot that could provide actionable guidance, and define a roadmap to move from discovery to a deployable solution.
Approach
1. Assessment and solution design
The project began with an assessment phase to review the client's existing data and tooling, paired with stakeholder interviews to capture pain points and expectations. This grounded the technical design in a clear understanding of what the AI chatbot actually needed to do and for whom.
Based on this assessment, Agilytic defined the target user journeys and recommendation logic, then selected a technical architecture suited to the use case.
2. Building the agentic AI chatbot
Rather than a standard question-and-answer chatbot, Agilytic developed the solution into a full agentic AI chatbot, including a user-friendly interface, coordinating several specialized components:
One for searching external, up-to-date information
One for retrieving relevant material from the client's own knowledge base
One for generating reports and structuring the final recommendations
These components were coordinated by a central orchestrator, allowing the chatbot to handle three distinct use cases:
Guided self-evaluation
Full implementation planning
General questions about the client's sustainability offering
3. Iteration and refinement
The team tested and refined the prototype based on usage by real clients and stakeholder feedback, improving both the quality of recommendations and the overall user experience. Along the way, the client's own knowledge base was cleaned and restructured to better support the chatbot: a byproduct with lasting value on its own, independent of the chatbot itself.
Results
The project delivered a working agentic AI chatbot prototype, along with a validated roadmap for further development. The solution came with a custom user interface, extending its use beyond a single-user demo.
Client feedback was very positive, with the team exceeding the initial scope of the engagement. The project also surfaced a valuable insight for the client: much of what made the recommendations useful came from the client's own accumulated experience and past reports, reinforcing the strategic value of that internal knowledge.
To safeguard confidentiality, we may modify certain details within our case studies.