Designing Brevo’s AI experience

Turning an AI-first ambition into a coherent product system.

I led the definition of Brevo’s AI experience framework, establishing a shared model for how intelligence assists, recommends, executes, and orchestrates across the product.

The framework connects product strategy, agentic architecture, and user experience, giving teams a common foundation to design AI-powered experiences at scale.

Company : Brevo — European SaaS CRM unicorn, serving 600,000+ customers globally across Marketing, Sales and Customer Experience.

Role : Unified UX and Design System Manager

Scope : AI Experience Strategy · Interaction Design · Design System · AI Product Foundations · Agentic Architecture

Timeline: 2025 to present

Collaboration: Product · Product Design · Content Design · Engineering · Brand

01 – The challenge: AI features don't create an intelligent product

As Brevo entered its next phase of growth, AI was becoming a company-wide ambition, but the market was already saturated with AI promises, assistants, and isolated “magic” features.

Across the product, teams were exploring opportunities to generate content, surface insights, automate actions, and introduce more intelligent workflows.

Individually, these initiatives could create value. But together, they introduced a new risk: adding more AI features wouldn't necessarily make Brevo feel more intelligent.

Without a shared model, each product team could define its own capabilities, interaction patterns, level of autonomy, and relationship with the user, creating a collection of disconnected AI experiences rather than one coherent product intelligence.

The opportunity was therefore bigger than designing an assistant. We needed to rethink how intelligence could connect context, data, insights, decisions, and actions across the product, helping users do things better, faster, and with more clarity, without taking control away from them.

Our problem became:

How might we enable teams to build AI capabilities at scale while making Brevo feel like one coherent intelligence?

02 — We stopped designing features. We started designing capabilities.

The vision became the Augmented CRM: an intelligence designed to assist and upskill users, helping them understand more, decide better, and act faster.

To make that vision tangible, we introduced Aura, Brevo's intelligent layer. Rather than a standalone assistant or AI embedded independently within each product, Aura was conceived as a shared system that could operate across the entire Brevo experience.

At its core, the system connects two layers:

  • Intelligence layer
    Reusable capabilities defining what the system can understand and do.

  • Experience layer
    Global interaction surfaces making those capabilities accessible wherever users work across Brevo.

Intelligence layer

The intelligence layer is structured around four types of skills:

  • AI primitive skill
    Shared atomic capabilities that power the system and can be reused across experiences.

  • Execution skill
    Turns intent into action, changing or generating outputs in the product.

  • Insight skill
    Watches, thinks, and recommends by analyzing context and data to surface risks, opportunities, and suggested actions.

  • Orchestration skill
    Combines capabilities into missions and determines what happens next, within defined rules and guardrails.

Together, they create one intelligence that can be reused, composed, and surfaced across the entire product.

Technically, these capabilities can be implemented as skills or agents depending on their complexity and autonomy.
The framework was intentionally technology-agnostic: its purpose was to give a broad audience across Product, Design, and Engineering a shared mental model of how the system understands, recommends, acts, and orchestrates, without requiring everyone to understand the underlying agentic architecture.

03 — One intelligence, wherever users need it.

The experience layer makes Aura available across Brevo, adapting its presence to the user’s context and the complexity of the task.

Aura was designed to meet users where the work happens, rather than becoming another destination they need to visit.

Its presence adapts to the task, from lightweight assistance to focused work, embedded creation, or proactive recommendations.

Compact mode A persistent assistant in a drawer alongside the product experience. Users can get support without leaving their context, and seamlessly switch to Focus mode when a task requires more space.
Focus mode An expanded Aura experience for more complex tasks requiring deeper focus, such as generating and refining an artifact alongside its preview. Users can seamlessly return to Compact mode at any time.
Editor mode Embeds Aura directly within creation surfaces, making conversational assistance available inside editors where users create and refine content.
Contextual recommendations Proactive suggestions surfaced directly within the product when Aura identifies a relevant insight, opportunity, or next action. Users can act immediately or continue the interaction with Aura.

These surfaces aren't separate AI experiences. They provide different levels of engagement with the same intelligence, allowing Aura to move between proactive assistance, conversation, creation, and execution without breaking the user's workflow.

Users don’t have to go to AI. AI comes into the experience when it becomes useful.