People of Product

The episode contains several genuinely useful observations - customers not wanting AI to do their taxes, experimenting with completing appointments inside ChatGPT rather than redirecting, and using dark web scans as a zero-cost value moment during onboarding - but they are surrounded by considerable repetition, host summaries, and well-worn CX truisms. Insight rate is moderate, not dense.

“they don't want H and R Block to use AI to do their taxes”

“if somebody goes to ChatGPT and starts asking tax questions and it's clear somebody's trying to maybe do their own taxes or figure out how they're going to do their taxes. Can we get them to set up an appointment with H and R Block in chatgpt”

The framing of 'using technology to create a human connection more than technology to replace the human' is a modest reframe of a conventional position, and the geographic-proximity-as-differentiator angle is underexplored. Most of the AI thinking (trust issues in financial services, human-in-the-loop) follows conventional wisdom rather than offering contrarian or first-principles arguments.

“we are using technology to create a human connection more than technology to replace the human”

“there is sort of this just natural mistrust of technology by many people, maybe most people”

Mike Halvorsen is a genuine operating VP running CX at a major consumer brand through live tax seasons, with 14 years of prior e-commerce leadership at Kohl's - clearly a practitioner, not a career conference speaker. His credibility is real but the role is functional/mid-tier rather than a transformational C-suite voice with cross-industry impact.

“I've got teams in Chicago doing testing with customers this week, trying to figure out what do they expect from H and R Block and how we use AI”

“we are much more open as an organization to testing and learning in tax season”

There are useful concrete specifics - 5% test cohorts, dark web scans as a named tactic, 70-year brand history, within-5-miles office density, average tax-pro tenure of 10 - 11 years - but no conversion rate data, no revenue or NPS figures, and no before/after metrics from any of the live tests, leaving most claims qualitative.

“put a new experience out in front of 5% of, of customers or a small subset of tax pros”

“somebody gives us a Social Security number and their email address and we can do a dark web scan for them in the moment”

The host asks a few legitimately useful setup questions (strategic planning cycles given model velocity, how product teams get close to field realities) but repeatedly paraphrases the guest back to him with approval, shares long personal anecdotes, and never pushes back on an unchallenged claim - resulting in a warm but low-friction PR-adjacent conversation.

“I really appreciate everyone that I've ever known that's been on the team on the, either the design side of the technology side there at Block”

“Love that. Because when you're thinking about taxes, mistakes happen”