In 2025, every SaaS product roadmap seems to include an AI feature. Investors ask about it. Customers expect it. Competitors announce it. The pressure to add artificial intelligence to your product is immense and comes from every direction. But the question that product teams should be asking, and often are not, is whether AI actually solves a real problem for their users or whether it is being added because the market expects it.

At PR_D_CT DAY 2025 in Louvain-la-Neuve, Arthur Geeraerd, co-founder of Troople, brought a refreshingly honest perspective to this question. Troople works at the intersection of technology and business operations, giving Arthur a practical vantage point on where AI creates genuine value and where it creates expensive complexity without corresponding benefit.

For the Belgian product professionals in the audience, many of whom are under pressure to integrate AI into their products, Arthur's talk provided a decision framework rather than a sales pitch. The result was a nuanced conversation that the AI hype cycle rarely allows.

slideshow

Original Presentation Slides

Download the slides from this talk as presented at the gathering.

A Framework for Evaluating AI in Your Product

Arthur presents a decision framework that product teams can use to evaluate whether AI is the right approach for a specific problem. The framework is deliberately simple because he argues that the complexity of AI technology should not infect the simplicity of the product decision about whether to use it.

Problem clarity: can you describe the user problem in one sentence without using the word AI? If not, you are starting from a solution rather than a problem

Data availability: does your product already generate or have access to the data that an AI model would need? If not, the data acquisition cost may exceed the feature value

Determinism tolerance: does the user context tolerate probabilistic outputs? In some domains, a response that is right ninety percent of the time is valuable. In others, it is dangerous

Alternative comparison: have you evaluated whether a well-designed rule-based system, a good search algorithm, or simply better UX could solve the same problem more reliably and at lower cost?

Maintenance budget: do you have the engineering capacity to monitor, retrain, and maintain an AI system over its lifetime? AI features do not ship and forget. They degrade without ongoing attention

lightbulbThe rule-based litmus test

Arthur suggests a practical litmus test: before building an AI feature, build the simplest rule-based version of the same capability. If the rule-based version solves eighty percent of the problem at ten percent of the cost, it may be the better product decision. AI should be reserved for problems where the complexity genuinely exceeds what rules can handle.

He notes that this framework is particularly relevant for Belgian and European SaaS companies, which often have smaller engineering teams and tighter budgets than their American counterparts. An AI feature that a well-funded Silicon Valley company can build and maintain with a dedicated ML team may be unsustainable for a twenty-person Belgian startup.

Does Your SaaS Really Need AI?

Original Presentation Slides

The Pressure to Add AI

A Framework for Evaluating AI in Your Product

lightbulbThe rule-based litmus test

Where AI Genuinely Belongs in SaaS

Pattern recognition at scale

Building a Practical AI Strategy

This article is for Consortium members

Key Takeaways

Related Editorials

Don't Just Listen: Invent

The Hidden Price Tag: Pricing Strategy