How to get your company started, not stuck, with AI

A version of this article originally appeared at Forbes. 

Are you and your leadership team thinking too much about AI? I don’t mean considering its impact or its potential, which will be significant and unpredictable. I mean: Are you thinking at the expense of doing?

The great thing about AI is that it is very approachable. While the technology behind it is fascinating, and its hype cycle is in full spin, finding your first AI applications doesn’t require much expertise. The real-world use cases are evident enough for you to start developing your own, so don’t waste time kicking tires while your competitors make progress.

I offer the following vision, process and guardrails to get you started.

1. Be open-minded, but don’t think too big.

AI is a time-saving productivity tool, so ask your leaders to explore repetitive tasks that take their teams a lot of time. I recommend an old-fashioned time analysis, complemented by qualitative conversations with the people doing the work.

For example, at my company, we discovered that one of our own customer success managers was using generative AI to summarize customer call transcripts, saving between 10 and 15 minutes every time. Our time analysis confirmed that this was a substantial opportunity for efficiency across their team and almost all customer success teams by extension. Not only did we start using AI summarization internally; we also incorporated it into our product.

Other areas to look at include content creation, task summarization and personalization for customer engagement at scale. Down the line, consolidating knowledge with internally trained AI models will be a big opportunity once the technology is there—but for now, it’s okay to think small. AI’s capabilities still leave much to be desired, and it needs to have its work checked by humans.

2. Establish essential guardrails to protect trust.

AI’s forerunner, automation, is now so prevalent in today’s business technology that we’re past the point of customers or employees feeling “creeped out” by AI. Today, they’re cautiously optimistic. However, there are real-world situations where the careless application of AI could impact trust.

My company applies three AI guardrails to protect customer information: The AI vendor we work with doesn’t use our prompts for training. We swap out personally identifiable information so AI doesn’t see the original data. If a customer’s compliance doesn’t allow the use of our AI model, we disable it for their account.

Your team should understand that AI always needs oversight. It’s a great starter and finisher—able to write an email draft and do a final cleanup, for example, but should always have a human editor. If expectations are unclear, consider an AI policy, similar to your social media policy, to ensure compliance and confidence.

3. Isolate your AI initiatives’ impact with focused metrics.

What goals should your leaders set for your first AI initiatives, and how should they measure the impact?

To isolate AI’s impact, I suggest to focus on productivity goals and broad efficiency gains. It’s still early for AI, so it’s fair to look at intermediate metrics as a proxy for success.

For example: How many more opportunities can a salesperson handle simultaneously with AI email assistance? Can a customer success manager who handled 40 accounts last year now oversee 44 accounts thanks to AI productivity gains? Can your support team manage a 10% increase in tickets? These are the goals I recommend prioritizing initially—and they should lead to positive end results.

Final note: Start from the top.

Whatever your first AI initiative is, it needs leadership buy-in from the top down. If everyone is on board, it’s easier to identify areas of productivity and easier for different teams to experiment and find their own use cases. Departmental leaders don’t have to be ardent supporters, but they can’t disregard or be cynical about AI—which, after all, is a tool, not a belief system.

For larger corporations, there’s a second need for a top-down approach: If your departmental leaders don’t take ownership of AI, your legal and compliance teams will. As with social media 15 years ago, you could end up with a risk mitigation approach that stifles organic innovation.