The trust dividend

By Stephen Wunker

Managing Director, New Markets Advisors

Stephen Wunker is a top thinker and doer in the worlds of innovation and AI. After leading AI product and business model transformations for nearly 15 years, he’s written over 20 articles on AI advances for Forbes and recently co-authored the book AI and the Octopus Organization with Jonathan Brill, the Futurist-in-Residence at Amazon. A longtime colleague of legendary Harvard Business School Professor Clayton Christensen, who coined the term “disruptive innovation,” Wunker had responsibility early in his career for one of the world’s first smartphones. As Managing Director of New Markets Advisors since 2009, he has advised some of the world’s leading organizations – including Meta and the Mayo Clinic – on innovation and AI strategy. He has worked throughout financial service industries. Wunker holds an MBA from Harvard Business School, a Master’s in Public Administration from Columbia University, and a BA from Princeton University.

If you sat beside a customer service representative at Amica Mutual Insurance three years ago, you would have watched a complex split-screen performance.

One monitor showed the policyholder’s file. Another held a blank notepad where the rep typed frantically throughout the call, in a format of their own invention, capturing details for the record while also trying to listen to the person on the line. Every rep could improvise differently, leading to inconsistent customer service. And every call carried the same challenge: balancing the need to deliver the right customer service while also capturing the necessary documentation.

Then Amica, a mid-sized US insurer, deployed AI to draft call summaries automatically. Calls got faster as reps had to pause less to type. But the more consequential change was harder to measure: The reps themselves were now able to actively engage with their customers: to react and respond, rather than focusing on basic data input. Net Promoter Scores have risen since. The machine took on the typing and the human got the conversation back.

This story inverts the narrative dominating the press, which can cast AI as an acid dissolving trust in finance through deepfake fraud, opaque algorithms and chatbots standing between customers and anyone who can actually help. Sadly, the public mood seems to agree. A YouGov poll in December 2025 found that trust in AI is lower in finance than in any other industry, with just 19% of Americans trusting it in that context.¹ TD Bank’s 2026 AI Insights Report found that only 18% of consumers would let AI make financial recommendations on its own.²

But look more closely at the TD data and a paradox appears. The same survey found 55% of respondents now use AI to help manage their finances, up from 10% a year earlier, and 85% still trust their banks as information sources. Consumers distrust AI in the abstract while embracing it in the particular, and they extend deep trust to their financial institutions. The question for financial services leaders is how to convert that institutional trust into an even stronger customer bond, using the very technology that headlines cast as trust’s destroyer.

Trust in employees at the edge creates trust from customers

The answer starts with recognizing what trust in financial services actually is.

It is earned in moments when, for example, a customer feels heard, an institution anticipates a need before it becomes a loss, and a claim gets paid accurately and the customer can see why.³ AI, deployed well, can produce more of those moments than any purely human operation can.

Yet, none of this can happen without foresight, planning and a clear and transparent strategy for AI deployment. The connecting thread is the principle that intelligence must live at the edges, where the customer relationship actually exists, with the center providing platforms, standards, and guardrails. I call this AI model the ‘Octopus Organization.’ An octopus has an incredible two-thirds of its neural tissue outside of its central brain and residing in its arms, and it happens to be one of the most intelligent creatures. Eight arms could become a tangled mess but it turns these limbs into a great advantage through the way that the arms connect to each other, sharing their signals instantly, all while the mini-brain located in each arm can direct local activity without always needing central instruction. It is a model of distributed intelligence where the bits closest to the action move relatively autonomously, with full context of what is happening elsewhere. The central brain issues instructions only when necessary. That is exactly how an AI-infused organization should work – a model now made possible by AI’s capabilities.

Trust, likewise, is won or lost arm by arm, interaction by interaction. And there are three mechanisms drive the trust dividend.

AI enables attention

Decades of growth, regulation, and specialization have added layer upon layer to how financial services organizations operate. Think of call scripts, data entry and compliance documentation. Each layer was rational, but it pulled the human further from the customer. The Amica example shows a reversal. When AI absorbs the clerical load, empathy stops being a soft aspiration and becomes an operational capacity. Reps have the cognitive room to hear what a customer is actually saying, including the things customers may never state directly.

Travelers, a large US-based property and casualty insurer, has pushed the same logic deeper into its organization. They have trained large language models on specialized domains so frontline staff can synthesize complex information and resolve issues without escalating up the hierarchy. Managers have been redefined from gatekeepers into contributor-coaches who vet AI outputs and unblock progress.⁴ The customer experiences this as speed and competence. That is how trust at the edges of an organization translates into trust from the outside.

The question for financial services leaders is how to convert that institutional trust into an even stronger customer bond, using the very technology that headlines cast as trust’s destroyer.

Listening becomes anticipation

The second mechanism is subtler. Every financial institution sits on an ocean of customer signals it has never had the bandwidth to comprehend. Amica records millions of calls a year. Until recently, the knowledge in those conversations could not be analyzed at scale. Now the transcripts feed an AI research agent that Amica’s product teams query to learn what customers want, what confuses them, and where coverage gaps are hiding.

The results are strikingly practical. Analysis of aggregated customer conversations and third-party data revealed that many policyholders don’t fully understand the nuances of their policy agreements. To take one example of the impact, Amica then launched a proactive outreach program to close potential coverage gaps for jewelry insurance. The extra premium paid is a benefit to the organization but not the primary objective. The goal is to ensure customers are properly protected in the event of a loss. The insurer is not selling harder. It is listening at industrial scale and acting on what it hears.

BBVA, the Madrid-based global banking group, applies the same principle to relationship banking, where the mathematics of attention have always been brutal. BBVA’s enterprise business asks roughly 2,000 relationship managers to serve about 107,000 mid-market client groups. That meant most clients didn’t receive so much attention. BBVA's answer is an “AI Banker” that prepares relationship managers with client analysis, sector context and next-step recommendations, freeing up scarce human judgment to reach clients who previously received a fraction of it. Crucially, the bank inverted its development process to protect quality. Rather than writing functional specifications and hoping for the best, business leaders define “golden standards,” real validated examples of what a good output looks like, and the AI is tuned to them. The grounding principle of this approach: start from the outcome you want and then work backward.⁵

Reliability you can inspect

The third mechanism addresses trust’s oldest foundation, which is: getting it right. Consider, for example, what an insurance claims adjuster faces after a serious accident. A hundred pages of medical bills may arrive from multiple providers, each in a different format and containing distinct demands with deadlines. Reviewing and evaluating this volume of information is incredibly time consuming for an unaided human.

To address this challenge, Amica is piloting AI that extracts the details of the claim from those documents, captures the loss information and links the findings directly back to its source page to assist the adjusters in their assessment. The payoff is a higher standard of care rather than a headcount reduction. Note the design choice deep inside that pilot, with every AI conclusion linking back to the underlying document. Traceability turns a black box into an audit trail, which matters enormously in a regulated industry. The role of the employee moves from an administrative one to a decision-making one, leveraging the technology to help them along the way.

Why trust is first and foremost a people process

That is why the internal dimension matters so much. Amica made nearly every AI tool opt-in, keeping humans as the final judges on how to handle customer-affecting decisions. People remain the decision makers, using AI to help them review information more efficiently, improve consistency, and better serve customers. When an early pilot underperformed, the business shut it down ahead of schedule because employees said it was not working, a decision that initially confused staff accustomed to top-down rollouts. That ultimately won their confidence.

Two years on, Amica’s internal surveys show that employees are proactively identifying new use cases with a positive outlook on the benefits. Likewise, at Travelers, the goal of AI deployment is not a cost-cutting exercise but rather extending the company’s lead in risk expertise and delivering great experiences for customers, agents and employees.

The Octopus Organization AI model, shown in the institutions winning the trust dividend, has a signature. These organizations measure success in attention restored, needs anticipated, errors eliminated and creative use of AI by employees at the edge, rather than focusing on headcount removed. That builds trust inside the organization. These companies also make machine conclusions inspectable and allow humans to focus on relational skills such as empathy and listening.

The press will keep writing about AI as trust’s destroyer, and for firms that automate carelessly, it could be. But for leaders who treat trust as a critical design objective, this technology promises to be the most powerful trust-building tool that financial services has seen in a generation. The machines are finally doing the typing, and the humans are finally free to listen.

Stay closer to the people, ideas and conversations shaping what's next. Join now