Ethical AI · 3 min read
You can’t afford a Chief AI Officer. You don’t need one. You need a 30-minute meeting with the five right questions.
Responsible AI is mostly about communication
Big companies treat responsible AI like a compliance problem. They write 80-page policy documents that nobody reads. Small companies don’t have that luxury, and honestly they don’t need it. What you need is for your team to understand the tools, to feel comfortable asking questions, and to know there are alternatives if a tool doesn’t fit.
That’s responsible AI for a 12-person company. Run this test once a quarter. If you fail any of them, you have a clear next step.
The 5-question test
1. Can every person on your team explain, in plain words, what AI is? Not in marketing terms. Not in academic terms. If they can’t say “it’s pattern recognition software trained on huge amounts of data, and it makes guesses based on patterns it learned,” they’re using a tool they don’t understand. That’s a training gap, not a policy gap.
2. Does your team feel comfortable raising concerns about an AI tool? If using AI feels mandatory, you’ve made it harder for people to flag problems. The person who notices the tool is getting things wrong is doing you a favor. Make sure they can say so without it looking like they’re refusing to keep up.
3. Does your team know there are alternatives to the big models? ChatGPT and Claude aren’t the only options. There are smaller models, open-source models, models you can run locally, tools designed for specific industries. If your team thinks AI means OpenAI, you’ve trained them into a single-vendor mindset that costs you flexibility later.
4. Has anyone audited what’s getting better and what’s getting worse? AI helps with some tasks and hurts with others. Without checking in, you only see what’s working. The harder, more important question is what’s quietly getting sloppy. Pick three workflows and actually compare AI-assisted output to manual output once a quarter.
5. Are you talking about the costs? Environmental impact. Job impact. Bias and accuracy. If these conversations only happen at the leadership level, your team isn’t part of decision-making about technology that affects their work every day. Make space for the harder conversations.
Score it honestly
If you can say yes to all five, you’re doing responsible AI at a small-business scale. If you can’t, the failures tell you exactly where to focus. That’s a useful test.
If you scored low on any of these and want to talk through what to do, we run sessions on this for small companies. Practical, plainspoken, no slides about disruption.