Why AI Adoption Fails

AI for Advisors is a loose, funny, wonkish show, and hosts Mark Heinen and James Canwell open this one by ribbing each other about domain names and how many products Canwell has shipped this week. Then they bring on two guests who sit at opposite ends of an advisor's life. Vineet Mohan, founder of FastTrackr AI, works on the end of a client relationship — the transition, when an advisor moves and has to carry their book with them. Frantz Widmaier, who runs Altitude CRM out of the 45-year-old Bill Good Marketing institution, works on the front end, finding and keeping clients. Both run lean and move fast, which is the other thread the hosts keep pulling.

The result is a two-hour-feeling hour that is really about one question: why does AI so often disappoint the firms that adopt it? The answer both founders keep circling is that AI is not plug-and-play, and pretending otherwise is how adoption fails.

Why Mohan bet on transitions

Mohan's origin story is familiar to anyone who's heard him — India, an MBA, fourteen years at HSBC across Leeds, London, Hong Kong, New York, San Francisco, and Edinburgh, always drawn to the roles where he had to stand something up. What's new here is how he found FastTrackr's real wedge. He started with the "gateway drugs" — the meeting assistant, document extraction — but went hunting for a hairier problem. He found it in transitions.

Everyone quotes the numbers, he says: 15,000-20,000 advisors moving firms a year, 400-500 M&A deals. But the talk is always about the deals, rarely about what happens after the paperwork is signed. Moving a client book is extremely painful and extremely high-value, because the clients are the whole reason the advisor was attractive in the first place — and the process had barely changed in 10-15 years.

Canwell pushes on the pain, and Mohan confirms the ugly stats: transitions typically take 90 days or more, and 10-20% of assets can leak out the door depending on where you're moving from and to. Part of that is time and a bad experience.

"No client who's been with an advisor for ten or fifteen years wants a blank intake form saying, give me these hundred data points so I can keep working with you. It's a you problem, not a me problem — the client didn't decide to change."

He lays out the three pipes feeding the funnel: aging advisors — median age around 55-56 — looking to exit; PE-backed aggregators buying up smaller teams; and advisors breaking away from wirehouses to go independent or join an RIA. Each move has its own nuances, and that's the point.

An operational problem with AI all along the way

Asked for the big unlock, Mohan is careful not to oversell. The hardest problem is re-acquiring client data when you've resigned and can carry almost nothing with you — so instead of a blank form, AI leverages existing sources and only goes back to the client as a backstop. Then there's the end-to-end paperwork, which changes by destination firm and custodian. And there's the piece nobody manages well: project management across hundreds of households living in hundreds of spreadsheets, where no one knows which household is where.

Canwell, who has consulted on "firm changes," vouches for the framing — other tools handle the pipes, the data aggregation to the custodian, but leave the human orchestration on top. He calls FastTrackr close to the first firm he's seen try to tackle the whole thing end to end. He tells his own war story from the consulting side: a colleague who pioneered the firm-change playbook back when advisors weren't even using DocuSign, hand-building the sequences and the all-important client letters and then hounding the team through the paperwork. Their most successful firm change, he says, arrived with 105% of the assets it left with — proof that the mechanics, done right, don't have to leak. Mohan's favorite proof point is smaller and more human: after a recent transition, an advisor told him the win was not having to work his weekends, which is unheard of when people move.

"I'd be lying if I said the AI is plug-and-play and it'll do the entire transition for you. It won't. But it takes away a lot — and those small wins are what people remember."

The founder who can't stop building — and won't let AI write the letters

Widmaier's half is a great counterweight. He came up through sales, then marketing, then product, married into the Bill Good family, learned the business, and took it over. His hardest-won lesson: running a service company and a software company at the same time is nearly impossible. Professional-services consulting needs deeply trained people who learn wealth management's terminology and nuances; the economics only work at very high revenue, and consultants you spend years training tend to leave. That's why the industry splits into consulting firms, software firms, and content firms — nobody does all three. AI changes the math, letting Altitude encode what its consultants do into an agentic system for a client base that already exists.

The most quotable stretch is about content. Altitude is AI-native, with an assistant called Pathfinder built from the application's own UI components so it surfaces data instead of dumping text. But Widmaier draws a hard line: the marketing content stays human-written, always, because there's an emotional quality to how they write letters that lands differently from anything AI-generated. Advisors regularly report clients arriving with a stack of those letters, convinced the advisor is a brilliant writer. The panel riffs, half-joking, about em dashes getting a bad name and clients accusing firms of using AI simply because their punctuation is correct.

There's a genuinely thoughtful middle section on the system of record versus the system of action, the "harness layer" everyone in the Bay Area is suddenly saying — Mohan notes he heard it constantly at New York Tech Week — and whether SaaS tools get used more than ever as agents become their power users. Widmaier bets on independent, personalized agents owning the data; Heinen bets that a breach at a major AI lab pushes people toward local, open-source models running for the cost of electricity.

The panel also spends real time on the human side of adoption, which is where a lot of AI projects quietly die. The recurring advisor objection — "my clients will never do that" — gets picked apart. Canwell's retort is that those same older clients already fill out iPads at the dentist and get a text form before a doctor's visit; the industry keeps underestimating them. Mohan brings a matching story from his HSBC days, a 75-year-old who refused online banking until paper instructions were simply withdrawn. The lesson both draw is that you build for both worlds at once — a slick data-collection flow for the clients who'll use it, and a path for the advisor who says one segment of their book will only ever DocuSign in the office, in front of them. There's friction at the start, then everyone flips, the way no-takers went from feared to default.

The misconception that sinks adoption

The rapid-fire close is where the episode earns its title. Asked the biggest misconception advisors hold about AI, the answers converge. Widmaier: that the industry can hand them a tool that does everything they want. Mohan: that things will be plug-and-play, which breaks down past a certain point — a meeting assistant just works, but the harder multi-step flows with edge cases need a human playing a real role. Canwell adds the sharpest version: AI can do a lot, but like anything it needs effort and, in AI terms, context — about your business, your clients, your processes, your goals. The tools that advertise themselves as plug-and-play are exactly the ones that fail, while the good ones find ways to build that context on your behalf.

Two underrated things get flagged before they wrap. Mohan mentions using Claude for persona-based QA testing — telling it to take on the persona of a client of a certain age and place, walk through the tool, and report what feels off — which the hosts immediately admire. And the whole panel lights up on skills: not chat's input-output loop but building your own skills, which Widmaier calls the most life-changing thing, period, and the "gateway drug" to asking what else AI can do. For a show about why AI adoption fails, it's a fitting note to end on — the firms that win aren't the ones with the flashiest tool, but the ones willing to give it context and do a little work.