At NexusOS, our signature fits in four words: we challenge, then we build. In practice, before writing a single line of code, we run every AI project through the ROI sieve. Out of three requests coming through our diagnostic funnel, one leaves with a reasoned no rather than a quote. This isn't a sales stance but a conviction: an AI project that doesn't pay for itself in under twelve months should, in most cases, never have started.

The ROI filter: payback under 12 months

Our first criterion is purely arithmetic. We estimate the expected annual gain (hours saved, errors avoided, revenue unlocked) and compare it to total cost: development, integration, API subscriptions, maintenance. If the return on investment exceeds twelve to fifteen months, the project enters the red zone.

  • Fuzzy or unmeasurable gain: we decline until it can be measured.
  • Volume too low: automating 20 cases a month rarely earns more than it costs.
  • The real problem is organisational, not technical: AI won't fix a broken process.
  • The required data doesn't exist, is dirty or inaccessible: no fuel, no engine.

An AI project that doesn't pay for itself within a year is almost always one that should never have started.

The signals that trigger a no

Beyond the math, certain patterns recur. The request is often framed as 'we want to do AI' rather than 'we have this problem and it costs us X per month'. When the goal is to tick a box or follow a trend, failure is almost guaranteed, because no one will own the use case once the demo is over.

Another classic signal in SMBs: no internal sponsor. Without someone whose daily work genuinely improves and who champions the tool, adoption collapses. We also decline projects where we're asked to replace critical human judgement with no control net, because the risk outweighs the gain.

Saying no is already doing you a favour

When we decline, we never leave the client empty-handed: we explain why, we often suggest a simpler alternative (a better spreadsheet, a no-code integration, a process change) and we keep the door open for the day the ROI conditions are met. That's how a no turns, six months later, into a healthy project.

This discipline has a virtuous side effect: the projects we accept succeed, because they were chosen for the right reasons. If you're unsure about the profitability of an AI automation idea, let's talk before coding: write to us at contact@nexus-os.fr.