A Simple Framework for AI in Your Business
Running a business often feels like a never-ending race to get more done in less time. AI sounds like the perfect shortcut to boost our productivity - until your team tries it and finds it’s falling short of expectations.
Sound familiar? What’s going on?
At Viabl, we believe AI can be a game changer. But, like any powerful tool, it's simple and effective but only if used the right way. If you don’t apply it to the right types of tasks, it won’t live up to the hype.
So how can you use AI effectively? Here’s our answer:
Flowchart for evaluation of AI tasks
First, Check Your Data-Sharing Agreements
Would you sign an NDA, then turn around and email that protected info to OpenAI? No? Then you probably shouldn’t put it into ChatGPT either.
Before you dive into AI, make sure you’re allowed to share the data you’re about to give it. This is something many employees at small and medium businesses skip over. We’re not lawyers and this isn’t legal advice. But that one’s pretty obvious, y’all.
Viabl offers services to ensure no protected information is shared to third parties. Get in touch if you want to know more!
Next, Ask These Three Key Questions
How could AI mess up, and what are the consequences?
Can a human step in to help if things go wrong?
If so, will it be obvious to the human when the AI has made a mistake?
Notice that question 3 says “when” the AI makes a mistake, not “if.” AI isn’t perfect and it’s extremely important to be aware of that.
Depending on how you answer, you’ll have a good sense of whether AI is a good fit for your specific task or if you need to be very careful.
Let’s dig into the implications of each question:
1. How could AI mess up, and what are the consequences?
First off, “messing up” can mean different things. You should think of the worst-case scenarios:
AI could give inaccurate information (it might “hallucinate” or make things up).
AI might generate responses that conflict with your brand guidelines, sometimes severely.
AI could perform a different task than you ask it to.
But even if the AI occasionally makes things up, it’s not automatically a deal-breaker. In some tasks, the AI making things up can actually be good (we’ll share a real-life example in a future post).
2. Can a human step in to help if things go wrong?
If the consequences of the AI messing up aren’t bad, then we don’t need to put a human in the loop.
But if errors could be costly, either financially, legally, or for your brand reputation, you’ll probably want someone to review or approve the AI’s output before it’s final.
How easy it is to add a human reviewer depends on the situation. For example:
If it’s a chatbot that’s live 24/7, it might be impossible to have people monitor every conversation, especially if you get thousands of them at once.
On the other hand, if it’s an AI tool that drafts responses for your help desk, having a human read them over might be easy.
3. If a human is in the loop, will it be obvious when the AI has made a mistake?
If an employee can quickly tell when the AI has messed up, then they can catch issues before they reach the customer. For example, if the AI suggests something that violates brand standards, that’s probably easy for a staff member to notice.
However, if the AI is suggesting complex solutions (like tech support instructions), it might be harder to know right away if the advice is correct.
What’s Next?
In upcoming posts, we’ll walk through real-life scenarios where AI works brilliantly…and not so brilliantly.
Got a usage of AI you want us to examine? Comment below! I’d love to hear your ideas and help you figure out where AI fits best in your business.