Aisling McCaffrey (Crystallized Intelligence) discusses AI’s effect on small-team hiring, the compliance gap behind tool choice, and why the real edge is a good agent, not a big budget.
Crystallized Intelligence founder Aisling McCaffrey surveyed 25 small Luxembourg teams on their AI habits. Her report uncovers a surprising hiring effect, a myth about budget and technical skill, and a compliance blind spot many small businesses and NGOs share without realizing it.
“AI can create jobs by letting small teams do more with less.”
Aisling McCaffrey, Founder and CEO of Crystallized Intelligence
What surprised you the most in the results of this survey?
It’s not one thing but two. First, AI’s effect on hiring is more positive than negative: three of our 25 respondents said AI helped their team grow faster, against just one who paused hiring. Most saw no change yet or said it’s too early to tell, but the ratio among those who did see an effect points the right way, and it supports something I’ve believed for a while: AI can create jobs by letting small teams do more with less. Second, there’s a myth that AI only pays off with a big budget, deep technical skill, and dozens of workflows across large teams. Our data says otherwise. Some small teams here are saving more than 10 hours a week, and over half are saving at least an hour, including that 10-plus-hours group. It’s especially effective for individuals or small groups who already know their job end to end, can recognize good output, and use AI to help make it happen.
64% of respondents say they factored in compliance when picking AI tools, yet Mistral usage stays low compared to Claude and ChatGPT. How do you explain that gap?
I think most people simply don’t realize what actually makes an AI tool compliant. Paying for Claude Pro or the ChatGPT equivalent and toggling “do not train on my data” feels responsible, but it doesn’t address GDPR on its own. What matters is the Data Processing Agreement, the document that sets out how the data controller, meaning you, and the data processor, meaning the AI company, handle and protect data. Mistral includes a DPA from its lowest paid tier. Brevo, a French email marketing company, offers one even on its free tier, because they need it to compete in the EU. American providers vary a lot: Google offers it for Gemini through Google Workspace, Claude through Team plans with a two-seat minimum or Enterprise, and ChatGPT through Business, Enterprise, and API tiers. And a DPA alone isn’t the whole story either. You still need a way to audit who accessed what data, and often, explicit consent to process client data this way. Options like EURouter, Mistral directly, or Scaleway in Paris exist, but most solopreneurs and small teams find that technical setup tedious, so they default to the tool everyone already uses.
“Think about what computer problems AI can solve for you.”
Aisling McCaffrey, Founder and CEO of Crystallized Intelligence
If you had to give one piece of advice to a small Luxembourg team still hesitating to adopt AI, what would it be?
Think about what computer problems AI can solve for you, not what thinking problems it can solve. You should be the one coming up with the ideas and the problems; let an agent figure out the execution. It’s never been easier or cheaper to customize your setup to do exactly what you need. I’ve seen small teams use AI agents to consolidate three organizations’ budgets into one agreed format ahead of a merger, purely mechanical work that AI can script and run. I’ve seen someone build a driver for an old printer that stopped working with a new computer, and I run daily tasks myself on an 11-year-old MacBook. Organizing files is another good one: an agent can write code that reads a file’s metadata, meaning its context and structure rather than the content itself, and sorts everything automatically, running entirely on your own machine. You describe what you want in plain language, and the system does the repetitive part.
