AI at Seyna: from assistant to operator

Over the past few months, AI was the world's best co-pilot. It would suggest, but humans would decide and act.

At Seyna, that model is changing. AI no longer simply helps our teams work: in a growing number of situations, it operates, under human oversight.

That is the defining shift of this quarter.

This Product Update is a little different from previous ones. No new feature to discover in the app, but something more foundational: a deep dive into how Seyna uses AI to build its platform, manage your portfolios, and prepare for the years ahead.

🏗️ Our starting point: an architecture built for AI

Many companies have grasped the potential benefits of AI. Few of them, unfortunately, can move as fast as they would like.

The reason is often the same: legacy information systems, fragmented, ageing and difficult to connect.

Every change runs into decades of accumulated technical debt.

At Seyna, we built our information system from scratch, just a few years ago. That changes everything.

In practice: when we want to connect Claude to our internal knowledge base, we build the connector in a week. When we want an AI agent to access our product data, the APIs are available and documented. There is no technical debt to work around, no legacy to manage.

That ease is what makes everything that follows possible.

⚙️ What AI is already doing, in practice

AI is now integrated across several layers of our operations.

‍
Broker data integration

‍
Every exchange with a broker (contractual documentation, bordereau specifications) now goes through an AI agent that structures the documentation required for data transformation. Previously, this could take hours of manual interpretation. Today, the process is automated.

AI also assists in writing data transformations and quality controls.

The next step, planned for this quarter: moving from a model where AI helps humans create controls, to one where AI writes them independently and humans validate. A seemingly small change but a fundamental one in its logic.
‍

Platform development

‍
On the engineering side, teams use AI to draft specifications, review code, and generate designs. The assistant has become significantly more capable.
‍
For the second half of the year, Seyna has set an ambitious objective: AI agents that write code independently, reviewed and validated by engineers. This is an evolution of the developer's role.

🔬 Deep Dive: bespoke analysis, for everyone

This is the flagship project of the quarter and probably the most consequential one for you, our broker partners.

The problem Deep Dive solves:

Until now, analysing an insurance portfolio in depth required either advanced actuarial expertise, bespoke Excel exports, or custom development.

The result: in-depth analysis was only accessible for large products and only through the actuarial teams.

‍
What Deep Dive changes:

Deep Dive is an AI-powered analysis tool that lets you query our data in plain language. No SQL queries to write, nothing to export. Data is always up to date, analysis is instantaneous.
‍
The power of the system comes from its embedded business knowledge: it knows, for instance, that a health product must account for non-linear acquisition, or that variable commissions affect the profitability calculation. That expertise never walks out the door.

Tangible gains:

• Customer Success Managers can access product metrics directly, without having to involve the actuarial team for every chart.
• Actuaries free up time for what truly adds value: interpretation, recommendations, and decisions.
• In-depth analysis can now be deployed across all products, not just the largest ones.
• Monitoring quality becomes consistent across the entire portfolio.

For brokers: faster, more accurate visibility into the real performance of your products. And steering committees focused on the decisions that matter, not on reconciling figures from previous months.

🚦Moving forward with our eyes open

The enthusiasm for AI is real but so are the risks. At Seyna, we have chosen to progress in a controlled manner.

A few principles we apply:

• No sensitive data leaves without oversight.
  We scrupulously adhere to GDPR principles across every use of AI.
  ‍
• Human in the loop, systematically.
  Whenever an action carries risk (a production release or a consequential decision) a human validates it. There is no full autonomy yet.
  ‍
• AI overseeing AI.
  We use automated verification systems ("guardrails") that self-regulate model outputs before they reach a user or a critical system. A first line of defence, on top of human oversight.
  ‍
• Business knowledge first.
  For an AI agent to be reliable, it must understand our business rules. Part of this quarter's work has been to structure and document that knowledge, making it accessible to the models.

‍

The result: we move fast, but we know exactly what AI is doing, why, and within what boundaries.

‍

🏁  Conclusion

What we are experiencing at Seyna is not an AI adoption story.

It is a shift in paradigm: moving from AI that amplifies humans, to AI that operates within defined boundaries, under oversight, and in a documented way.
‍
This is what we are building, concretely, quarter after quarter, with an architecture designed for it from day one.
‍
AI is becoming part of the Seyna team. Under control. In service of our brokers.‍

‍

See you in three months!‍

The Seyna product team

‍