Don't Forget Your Frontlines: Lessons On The Human-AI Partnership

Everyone’s racing to deploy AI. The winners in customer experience will be the ones who make it work on the frontlines — not just in the tech stack. At the CX BFSI USA East Exchange, leaders from major banks and insurers, like M&T Bank, The Hartford, Allstate Canada, JPMC, Regions, and Truist, showed what that looks like in practice.

1. The Most Valuable CX Data Loop Starts With The Frontline

A leader from M&T Bank shared their "sense-and-respond" model, turning customer and employee feedback into operational action. The system works because it's grounded in frontline involvement:

• Frontline staff help shape insights.
• Complaints are treated as valuable input to identify where service or product experience needs adjustment.
• What matters most is how behavior changes as a result of data, not just whether feedback is reviewed.

The bank compared two regional markets with different usage patterns. In one region, the average user accessed customer feedback data 69 times over the year and saw a 14% increase in customer satisfaction. In another area where usage declined, satisfaction also dropped.

The finding was clear: it wasn’t the breadth of who used the data that mattered — it was the depth. Users who logged in frequently and engaged deeply with customer feedback drove measurably better outcomes.

Those results came from tools built with frontline input. The same principle applies more broadly: every CX system, dashboard, and process has to start with the people who use it.

2. Build CX Tools For The People Who Use Them

Leaders across multiple sessions emphasized the importance of knowing who you're designing for. It’s not just the customer. The teams responsible for acting on the insight, the frontlines, are also priority users.

• CX programs fail when they overlook how frontline teams will use or support them in practice.
• Customer-facing teams require access to relevant insights and the authority to act on them promptly.
• Helping frontline teams develop habits that make data useful in real decisions matters more than just increasing access to reports.

The same tool can produce dramatically different outcomes based on how deeply teams engage with it. That means building the right tool isn’t enough. Teams need to integrate the tool into their standard working practices, or its efficacy will quickly diminish.

If CX programs depend on frontline action, then frontline teams need to be part of the design, rollout, and reinforcement.

3. AI Finds Patterns, People Make Decisions

Leaders are approaching AI adoption with both caution and a sense of urgency. In a session featuring representatives from The Hartford, Allstate Canada, and Self Financial, one speaker shared details from a comparison study of AI tools. When comparing internal AI tools against external vendors, the internal solution achieved 96% accuracy while the external vendor’s reached only 66%. The lesson: you can’t just trust vendor promises. You must test, monitor, and verify that the outputs meet your standards.

Multiple speakers described how their organizations are utilizing AI to review large datasets while retaining final decisions in human hands. At one organization, they process 6 million pages of medical records annually. AI-powered summarization could dramatically improve efficiency, but they drew a clear line. They will never deny a claim based exclusively on AI-generated summaries.

To make those calls with confidence, one speaker at CX BFSI shared his CAPE framework:

• Copyright: Make sure generated content doesn't infringe.
• Accuracy: Verify results before acting. After all, AI will confidently tell you things that are not remotely true.
• Privacy: Keep customer data secure at every step. Never send sensitive information to external portals.
• Explainability: Make sure everyone understands how the AI arrived at its conclusion. “You can’t use AI as a black box,” noted the presenter.

When frontline teams trust the process and the protections, they’re more likely to use AI insights consistently and effectively.

The same need for connection and context came up in conversations about silos.

4. Fix The Silos So The Frontline Teams See The Whole Customer

AI insights lose value when they’re trapped in departmental silos. A customer doesn’t care which system or team owns the data. They just expect the next person they speak with to know what happened before.

A speaker from JPMC illustrated this with a common scenario: a client emails their salesperson with a concern, then calls the service desk 24 hours later. The service person is unaware of the email. When the client says, “I already emailed your salesperson,” frustration escalates.

The solution doesn’t require sophisticated AI agents. It simply needs basic analytics that highlight recent interactions. Even basic acknowledgement changes the dynamic: “I understand you already reached out via email. I apologize that you’re contacting us a second time.”

Another practical example is when a service agent can see that a customer spent two hours yesterday trying to submit a payment online. The agent can open with, “I’m sorry. I see what you went through yesterday. Let’s get this resolved.”

Even simple integrations, like surfacing past customer interactions during a service call, can transform disconnected touchpoints into consistent experiences for customers and employees. When AI connects data across service, product, and operations, frontline teams can anticipate needs instead of racing to problems.

That context also matters when interpreting what performance metrics really mean.

5. Context Still Belongs To Humans

Data, dashboards, and predictions can mislead if they’re not paired with human judgment. One speaker shared two powerful retail examples that illustrate how metrics can deceive:

Kmart’s satisfaction paradox: As Kmart entered its downward cycle, customer satisfaction scores kept climbing. In fact, the company posted a record-high satisfaction rating the quarter before its bankruptcy. Why? Because the only customers left were those addicted to rock-bottom deals — unprofitable customers the company was inadvertently scaling.

Walmart’s Project Impact. Walmart undertook a major customer experience initiative to improve store design. They widened aisles, removed cluttered end caps, and created more space. Test results showed customers loved the new experience. But year-over-year sales declined in test stores. It turned out that narrow aisles slowed food traffic and created more browsing time, leading to impulse purchases. The improved experience eliminated profitable friction.

Of course, sometimes, the reverse is true — numbers can look worse while things are actually improving. In contact centers, for instance, as AI handles routine calls, average handle time (AHT) rises even as first call resolution (FCR) improves. Without context, it can seem like productivity dropped. In reality, complexity increased, and customer outcomes got better.

AI and analytics can surface patterns, but people are needed to interpret what’s really happening. That’s why one speaker challenged CX BFSI attendees to measure differently, “Don’t just measure yourselves, measure competitors. Convert ratings into rankings. Look at novel metrics. Ask unusual questions like, ‘Did the associate exhibit a strong desire to serve you?’ rather than the generic, ‘Were they professional?’”

6. Internal AI Use Cases Are Gaining Ground

Early AI gains are coming from inside the organization, where tools that remove friction for employees also improve CX. Examples shared across sessions were practical and grounded:

• Converting meeting transcripts into training documents.
• Helping staff navigate complex benefit questions (like new parents calculating leave time, pay, and job protection).
• Surfacing recent customer actions before service calls.
• Enabling employees at large companies to find information across departments.

One speaker noted that freeing up salespeople – often the most expensive resource in a company – delivers multiple benefits: increased revenue when sales focuses on high-value activities, improved operational efficiency when service handles service requests, and happier employees who can focus on what they were trained to do.

Each of these removes friction from daily work, helping teams focus on higher-value interactions. When AI simplifies tasks, employees respond faster, learn faster, and serve customers better.

Frontline Teams Make AI Work

AI can transform customer experience, but only when it’s connected to the human moments that shape that experience. A recurring message across sessions: the biggest obstacle isn’t the technology, it’s the silos. Everyone is trying to integrate AI, but too many are still working in isolation. Meanwhile, customers are moving across channels and touchpoints.

The path forward requires CX leaders to:

• Involve frontline teams early in AI and CX design.
• Break down silos so frontline teams can see the entire customer experience.
• Focus on building habits and behaviors that turn insights into action.
• Preserve human judgment where customer trust and nuance matter most.

As AI handles routine tasks, the nature of frontline work undergoes a fundamental shift. Examples from CX BFSI illustrated this perfectly. AI is summarizing medical records, but humans are still needed to apply critical thinking and make final decisions. Meanwhile, an increasing number of service agents are being equipped with customer history and click patterns, enabling them to provide proactive support rather than reactive problem-solving.

This transition doesn’t happen automatically. And as the M&T Bank experience demonstrates, the ROI from AI and CX investments depends on how deeply frontline teams engage with new tools and insights. That’s where Pathstream helps. We motivate and equip your frontline teams with the AI readiness, empathy, judgment, and problem-solving skills they need to deliver better outcomes in this new environment. Learn more about how to build AI-ready customer contact teams here.