AI is no longer just something insights teams are testing or talking about. AI is actively hitting everyday workflows across organizations. As teams race to improve speed, efficiency, and access to information, many are also confronting new questions around trust, interpretation, governance, and decision quality.
Following a recent AI summit and a series of conversations with leaders across industries, Monika Rogers, SVP of Growth Strategy at CMB, began noticing a pattern. While much of the discussion focused on what AI can do, there was far less conversation about how organizations should adopt it and where they may be creating new risks in the process.
In the discussion below, Monika shares her observations on the future of insights, the role of AI in decision-making, and the traps leaders should avoid as they build more AI-enabled organizations.
Q: You recently wrote about three traps organizations should avoid when adopting AI. What prompted that perspective?
A: I recently came back from an AI summit at UW-Madison genuinely excited, but also a bit concerned. There were terrific examples of teams using AI to save time, improve workflows, and make the technology feel practical rather than theoretical. The concern comes from seeing how quickly organizations are moving to adopt these tools without thinking through how they fit into actual decision-making. Right now, the momentum is heavily weighted toward what AI can automate, rather than focusing on what it can help an organization do better.
Q: One of the traps you describe is focusing on automation rather than enablement. What’s the difference?
A: Automation focuses squarely on efficiency and moving faster. Enablement is a higher bar; it’s about changing the quality of the decisions you make because the technology is there. Many organizations begin by applying AI to repetitive tasks because the benefits are easy to measure: things like summarization, reporting, coding open ends, or drafting communications. Those are all valuable applications. But the risk is assuming that productivity gains automatically translate into better business outcomes. An organization can absolutely become faster without becoming any smarter. For insights leaders, the real question is: What becomes better because AI is part of the workflow? Does it improve judgment? Does it improve decision quality? That’s a much tougher standard than simple efficiency.
Q: You also raise concerns about organizations making knowledge more accessible without making it more interpretable. Why is that important?
A: A lot of companies are building internal knowledge environments and loading past research, reports, transcripts, and data into AI-enabled systems. Making information easier to access is incredibly useful, but access is not the same thing as understanding. The real value of research lives in the interpretation layer: the context, judgment, tradeoffs, caveats, implications, and recommendations wrapped around the data. If those elements are lost, teams can retrieve information very efficiently while completely misunderstanding what that information actually means.
Q: Your third trap focuses on “ignoring drift until trust breaks.” How does drift manifest for an insights team?
A: Drift sounds like a technical machine-learning term, but for insights teams, it shows up in very human, organizational ways. There’s context drift, where yesterday’s consumer reality becomes stale faster than the organization realizes. And then there’s interpretation drift, which occurs when a nuanced body of learning is repeatedly summarized, repackaged, and reused until the original context and critical nuances begin to disappear. When drift takes over, a nuanced finding gets stripped down into a simplified takeaway, which eventually gets flattened into a bullet point on a summary slide. Eventually, people are working from a version of the learning that may no longer reflect what the original research actually showed.
The recent GRIT report highlights why this is becoming a massive issue. It shows that formal insight groups are becoming less consolidated even as total insight staff grows, and the new “insights operations” role has emerged strongly across both research and analytics. When insights work is spread across the enterprise, more people are touching it, using it, and likely reshaping it. Without a shared understanding of where the learning came from or what boundaries were attached to it, different teams can use AI differently and pull completely contradictory conclusions from the exact same data.
Q: How does that ultimately impact trust within the organization?
A: Trust becomes entirely foundational as AI adoption expands. If teams begin encountering conflicting outputs, outdated information, poorly interpreted findings, or recommendations that cannot be traced back to reliable evidence, confidence starts to erode. Once trust breaks, adoption stalls, regardless of how sophisticated the technology may be. That is why governance, validation, transparency, and data quality are not side issues; they are baseline requirements for long-term success.
Q: How do you see the role of insights evolving over the next few years?
A: The role becomes more strategic, not less. AI will increasingly help with drafting, summarization, synthesis, pattern detection, and information retrieval. Those capabilities will continue to improve. But the human value shifts toward framing the right problem, interpreting nuance, understanding context, evaluating evidence, and translating learning into business action. In many ways, the future of insights is not about producing more information. It is about helping organizations make better decisions with greater confidence.
Q: What should insights leaders be doing right now to build a better path forward?
A: My advice is to think beyond efficiency and focus on how learning moves through the organization. To build AI into our workflows in a way that strengthens that movement, leaders should approach adoption in four deliberate stages.
First, look at how you generate insights: use AI to enable better, deeper research for critical decisions and context, rather than simply trying to replace the research process. Second, work to extend that learning by training, testing, and validating AI agents on past research so the right strategic synthesis is embedded and carried forward to prevent drift. From there, you need to integrate secondary, always-on data sources to blend with periodic research, so newer, more continuous data streams become part of a broader evidence system. And finally, activate those insights by partnering internally to ensure the output is purpose-driven, accurate, and actively guiding the tradeoffs, decisions, and business next steps that matter.
Organizations that approach AI as a decision-enablement tool rather than simply an automation tool will be in a much stronger position over the long term. Winning here won’t mean deploying the highest number of AI tools. It will mean being the best at pairing those tools with human judgment to drive actual business choices.