AI Can Deliver Clarity At Scale, But Strategic Decisions Still Need Human Judgment

Companies keep pouring money into AI and struggling to see returns. The frustration often comes from a fundamental mismatch: organizations are buying isolated features when what they actually need is a system. Real progress comes from treating AI as part of a holistic decision support system rather than a collection of disconnected capabilities.

Ryan Muir is a Product and Data Science Leader currently responsible for the Insights suite at Cresta. His background puts him at the center of this problem. He previously served as Cresta's Head of Data Science, where his team's research on generative AI was featured in the Harvard Business Review. His career has focused on connecting deep technical work with business outcomes. For Muir, the problem is a breakdown in what he calls the "insights-to-action loop."

Modern AI has solved for the first two stages, identifying what is happening and understanding why, but most organizations stall at the third step: deciding what to do about it. "Legacy tools could tell you what was happening, and modern AI has unlocked why it's happening. But everyone is stuck on the third step: 'What do I do about it?'" says Muir. "That's where humans should be, doing the real thinking and the strategic decision making. AI needs your enterprise knowledge. It doesn't know you changed your policy yesterday or that a pricing change went into effect. You don't want AI making that strategic decision for you in 2026."

• Walled off from reality: Muir's systems engineering background shapes how he frames the challenge. He views it through the lens of building a decision support system, a concept that has existed for decades but is now being supercharged by AI. True value comes from systems that provide broad access to context, an element often missing from isolated features. Without that systemic approach, even the most advanced AI tools risk operating blind. "I don't want AI point solutions," he says. "A random app with an AI chatbot doesn't know about my life; it only knows about what's in its own walled garden of data. I want a system that has access to look at context and take actions on your behalf."

• AI finds, human decides: To illustrate a better approach, Muir points to a case study with a leading Fortune 500 brand. AI identified both a spike in calls about a certain topic and the root cause: customer confusion about a new policy. "That's where you absolutely need the human who has the context. They have the background from three months of debating this policy, so they can recognize what to tweak about the policy. That person can then go to the product team and clearly explain the gap, why it's happening, and how to close it, using AI as a tool and not as the actual decision maker." After the system diagnoses the what and why, it falls to a human to determine the what to do.

• Menial vs. macro: Muir draws a line between high-stakes insights that belong to leadership and low-stakes, repetitive tasks that are ripe for full automation. The goal is to free up human capital for work that matters. "There are many easy use cases where AI should be the decision maker. Someone calling to change their appointment date? That's a monotonous task for any human to handle. Let the AI be the decision maker there," Muir explains.

The growing wave of AI agents in business software reinforces the divide. Routine transactions move to machines. Strategic judgment stays with people. To make the need for systemic change concrete, Muir uses a familiar example: the post-interaction customer survey. He describes it as a classic illustration of a disconnected feature operating at the speed of a broken process, failing to keep pace with the customer.

• The Amazon effect: "We can't move at the speed of surveys, because by then the customer has already churned. We need to know they're about to churn from signals like a low CSAT score or the mention of churn, and then intervene within a minute after the call ends," says Muir. His alternative is a system built for what he calls decision velocity. "The people that move at the speed of their customers will be farther and farther ahead from a customer lifetime value perspective, and everyone else will get left behind. It's like Amazon. They started doing two-day shipping, and now everyone has to do two-day shipping because that's the new standard for customer experience."

• A losing game: Implementing such a system comes with real obstacles. Muir notes that many companies are hampered by institutional inertia and legacy workflows. "Leaders are often compensated on gamed metrics, so they think their CSAT is at an all-time high. For example, they will rephrase a neutral question about the quality of an experience to be a leading one, such as asking how great the experience was. This creates incentives to preserve workflows that probably shouldn't exist anymore, which is how companies get trapped in a 1995 mindset."

• Lightbulb moment: Breaking free from that trap requires a fundamental change in how value is measured from day one. Muir's vision is a system that delivers value so immediately it creates a new baseline for business intelligence. "From day one, you should be able to see if a customer's issue was actually resolved," says Muir. "This is a metric I've found that virtually no enterprise truly knows, and you get it without having to survey anyone. It creates a before and after moment, a shift so fundamental you realize you were operating in the dark and now you have a flashlight."

The contact center sits at a crossroads as organizations decide how to divide labor between humans and machines. Muir's argument is that winning CX teams in 2026 will be those that redesign for decision velocity rather than shipping more features. Skilled human agents remain essential for the moments that change customer loyalty. The rest can move faster, as long as the system is built to support it.