Artificial Intelligence (AI) at Springly

Like many technology companies, we believe AI can help us serve you better: faster answers, and teams focused on what genuinely requires human expertise.

This article explains where we stand and the framework we hold ourselves to.

AI in our customer support

Our goal is simple: to give you a reliable answer, as quickly as possible. Specifically, when you write to our support team:

1. Your request is analyzed and qualified by our AI assistant, so it can be routed to the right person, with the right context.
2. An initial response may be sent to you immediately, drawing on our public resources: help center, documentation, product articles. Only if your question is unambiguous and the assistant has sufficiently reliable information. When in doubt, it sends nothing.
3. A member of our team then reviews your request, every time. If the answer needs to be corrected or clarified, they'll write to you again. No request is closed without this human review.

We know these assistants can sometimes answer with a lot of confidence, even when they're wrong. That's why we prefer a human review step. If an answer doesn't seem to address your question, simply reply to the ticket: you'll get someone from our team.

A conversational assistant in your platform: AI Assistant

A chat is now available, in testing, for some of our customers.

It lets you easily query our product knowledge base to move forward with your work on Springly.

Its scope today is deliberately limited: documentation lookup, and support on how to use the software and on nonprofit/association life.

If feedback is positive, we'll roll it out more broadly. In the medium term, we plan to integrate it more closely into your platform and allow it to carry out certain actions on your behalf.

What technologies do you use? How is my data handled?

The models and solutions we use evolve regularly: publishing a fixed list of our technical choices here wouldn't make much sense. What we can describe is the framework we hold ourselves to.

• Restraint. We limit what's shared with a model to the strict minimum necessary.
• Professional solutions only, under enterprise license, governed by a contract that defines and limits data use.
• No training on your data: this is a contractual condition, which we verify before putting anything into service.
• No consumer-grade tools: free solutions whose terms allow data reuse are off-limits to our teams — who are trained on this point.
• Internal, sovereign infrastructure: we host certain models on our own infrastructure, in Europe, and are looking to expand this.

Finally, use cases are clearly bounded: software development, and support for requests whose answers are already documented publicly. Details of our processing activities and subprocessors can be found in our privacy policy.

And the human and environmental impact?

These technologies are transforming our work, and we won't pretend otherwise. For our part, we're fairly enthusiastic — these tools let us improve our services faster.

We continue to hire (a lot!). This includes our tech and customer support teams. In the latter, we're now looking for somewhat different skills: more focused on live training and guidance.

We're also aware of the environmental issues tied to excessive use of these technologies (particularly in data centers: electricity and water consumption, the proliferation of electronic components to meet soaring computing and storage needs).

We moderate our usage, explore leaner solutions (in-house hosting, lighter models), train our teams, and avoid systematically giving in to the pursuit of "more, always more" technology.

Have a question or comment on this topic? Write to us: we'd much rather talk about it.