Almost every project we start at Nyxora begins with an audit. Not because it's the procedure, but because most of the failures I've seen on AI projects come from the same place: something impressive got built before anyone truly understood the problem.

An AI audit, on our side, takes one to two weeks. Here's what we actually look at during that time.

1. Not your tooling, your process

The first question is never "which LLM do you use" or "is your data already clean." It's: how does the work actually get done today, step by step, by whom, and where does it get stuck. Technology comes after. An audit that starts with the tech stack misses the real problem.

2. Where the human time actually goes

We systematically look for repetitive, well-defined tasks: a role that's stayed open for months because nobody wants to do it, a team handling the same type of request over and over, reports rebuilt by hand every week. That's where the first realistic use case lives, not in the most ambitious idea raised in a meeting.

The best signal an audit can find isn't an AI opportunity. It's a task nobody enjoys, that repeats every week, and whose current cost we can actually measure.

3. What an audit actually delivers

At the end of the two weeks, you get three things: a clear map of your candidate processes, two or three use cases prioritized by frequency, data availability and risk, and a realistic estimate of what each would cost to build. Not a 40-page report that gathers dust: a document you can actually use in a leadership meeting the following week.

  • A map of candidate processes, ranked by impact and feasibility
  • 2 to 3 prioritized use cases with cost and gain estimates
  • A clear recommendation on what's next: a PoC, automation, or nothing for now

4. Why two weeks, not two months

An audit that drags on becomes a project of its own. We deliberately keep this format short: enough time to understand the terrain, not enough to get lost in it. If a topic deserves deeper investigation, it becomes a clearly identified next step, not a silent extension of the audit.

In summary

An AI audit isn't there to convince you to do AI. It's there to tell you, with numbers, whether it's worth it, where to start, and what it will actually cost. It's the most honest entry point we know for starting a project that actually holds up.