AI is powerful but it makes mistakes. Is your business prepared to respond when something goes wrong?
Modern AI is powerful, but more than a little bit scary.
We’re not talking Skynet-style apocalypse scary — we’ll leave that kind of writing for others to do. We’re talking about business-risk scary.
(By the way, by “modern AI” we’re referring to what sometimes gets called generative AI or LLMs, though the category is already much broader. Tools that can take a text prompt and dynamically write, summarize, code, interact with customers, create graphics — that sort of thing.)
The more deeply AI is embedded into the ways you work, the more you need a plan for what to do if it goes off the rails. Because, if you think about it, that’s not really the right metaphor. The whole thing with AI is that you’re intentionally letting it go off the rails. You need to be able to trust that it will move properly on its own. And you need to know how to pull the plug if it doesn’t.
Understanding AI Risk
Why do these tools create business risk? Because they can be really powerful, but they aren’t always right. They get things wrong, make “bad decisions,” respond in unexpected ways.
That’s not a huge deal if an employee accidentally includes a prompt or “ChatGPT commentary” in an internal email (“write an email to my jerk of a boss”, “if you’d like, I can rephrase this…”).
But when an AI chatbot promises a customer something you can’t provide? That’s a bit of a deal. What about when an AI agent deletes your database or exposes customer data? Huge deal.
AI As Accelerator, Not Magic Potion
Another helpful way to think about AI that relates to risk is this: AI is an accelerant or accelerator, not a magic potion.
If you take a bad or broken process and throw AI at it, AI won’t magically fix that process.
Instead, AI will just move you faster through that bad or broken process. You’ll usually reach the same bad outcomes, just ten, twenty, or a thousand times faster (or at scale). One mistake becomes 20,000 faulty customer emails faster than you can hit “send”.
So if AI has the potential to get you to the right destination so much faster and at truly epic scale, then it has equal potential to get you to the wrong destination just as fast and just as wide.
A Quick AI Triage Plan
If you’re concerned about AI running amok at your business, here’s a quick 3-step triage plan.
1. Inventory AI Use
You can’t fix what you don’t know is happening, so start by inventorying every place in your business where people are using these AI tools. This ranges from frontline staff using Copilot or ChatGPT to refine emails all the way to behind-the-scenes AI agents doing things with data.
Make an inventory of the tools in use, what those tools are doing, and what they have access to.
2. Assign Ownership
Your analog business processes all have owners, right? Sales, customer support, and marketing all know which processes they own.
You need to do the same with AI tools so that when something goes wrong, there’s no debate over who needs to take action.
If during this process you discover AI tools in use that no one really knows how to “own,” that’s a red flag. You may need an IT assist from a provider like us.
3. Document Governance and Backup Strategies
It’s increasingly important to establish governance for all IT systems, including AI tools. “Governance” means essentially who has access to what. Depending on where you operate (and what industries you serve), you may need to show compliance with various regulations (AI-related or otherwise), and documented governance is a key piece.
With AI, you also need to document backup strategies:
- What does a business process look like if you need to temporarily disable an AI tool or system?
- How do you do that disabling?
- Who will do it?
In the end, the principles are similar to good IT governance. They just become even more important because of how quickly AI can move you — either for good or bad.
We help companies work through issues like this every single week. If you could use an assist, reach out anytime.