Never before has market research played a more critical role in the health of your business. Whether it’s navigating the political environment, understanding the economy, or even dealing with ill-advised lawsuits against advertisers from a certain ad platform (looking at you, X), knowing your audience to the core has become an essential tool for survival. Which is a big reason so many people are now pinning their hopes on artificial intelligence.

AI gives us a whole new way to deal with online research. Now, in the name of faster and more efficient data collection and assessment, we can compose a query that explains the scope of our project, then set our AI assistant loose to uncover the details. Within minutes it can return a summary of what it found or even a fully complete report that’s ready to go.

On the surface this looks like a pretty good deal. AI offers a faster, more efficient means of unpacking the market’s true needs and intentions. And, in turn, it can allow us to delve more deeply into competitive analyses, better understand our historical impact, and develop more relevant messaging and creative strategies for today.

But is it wise to put all your trust in a black-box technology?

For instance, what are the sources behind your AI’s conclusions? Can you really trust the raw data used to develop the confident and authoritative answers AI assistants typically provide you?

Efficiency is great, but when it comes at the price of accuracy, it’s important to know if your research partners also appreciate this fact or are simply letting their AI tools run amuck. After all, it’s your business that hangs in the balance.

Where AI can fail us

The issue with relying too heavily on AI is pretty much the same as relying too much on any single researcher—if you don’t have adequate peer review, you risk introducing biases and mistakes into the project.

AI still struggles with determining authority with the data it finds. Yet there’s no obvious indication of this struggle from the user’s perspective. All we see is the AI confidently proclaiming an answer to your query, with all the citations in order.

What’s worse, many of the AI search platforms make you dig to find the source links responsible for a query’s answer, leaving the impression that vetting is inconsequential.

Now admittedly, a seasoned researcher will usually suss this out and help the AI arrive at more accurate conclusions. However the vast majority of organizations passing themselves off as market research companies are not going to admit to the problems. They’re just going to see the efficiency of the work process. And the result could be catastrophic to you.

A plan for moving forward

It’s simply not acceptable to work with a research partner with a laissez-faire approach to AI. You need to know how they plan to develop and refine its output. So with that in mind, here are a few thoughts:

• Embrace the technology judiciously: There’s massive pressure to dive in head first with AI and sort out the problems later. However a better approach would be to treat the tech as just another thing you’re researching. Work with your partner to develop the most beneficials ways to integrate it first and then test the most promising applications before rollout.

• Look for evolution rather than downsizing research staff: The temptation for cost cutting is real among research companies, as are the promises from the AI vendors implying their tools can replace human capital completely. This simply isn’t the case. Instead, look for a partner who encourages a vision for collaboration between their staff and these new tools.

• Demand transparency about AI oversight: There’s absolutely no question that AI-driven market research needs human verification. Demanding this as the gold standard is non-negotiable. Always ask about your market research partner’s process for working with AI and be clear on the role their human staff plays.

Artificial intelligence undoubtedly offers incredible opportunities for market research to parse huge amounts of data into actionable insights. And the value it could bring to the industry is inestimable, to say the least.

Which is exactly why we need to take seriously the need to establish best practices. Because as promising as these opportunities seem now, only researchers exerting proper guidance of their AI tools will be able to reap the kinds of benefits that brands are looking to find.

Tim Ringel is the founder and group CEO of Meet The People.