You Don't Have to Boil the Ocean: How Banks and Credit Unions Should Actually Start with AI

Most financial institutions are stuck in the same place with AI. They know they need to do something. They're just convinced that something has to be enormous, risky, and customer-facing on day one. So they wait.

That instinct is backwards. The way to win with AI right now isn't a giant leap. It's a small, safe, high-value first step that builds confidence and compounds from there.

Posh CEO and co-founder Karan Kashyap made the case for this on the Amplitude of Tech podcast, and it's worth unpacking, because the "start small" path is the one most institutions overlook while they argue about the big swing.

The myth of all-or-nothing AI

There's a belief floating around that AI only counts if it does everything. That you haven't really adopted AI until it's running your entire contact center, handling every customer interaction, replacing whole workflows end to end.

Karan's response to that is blunt. As he put it, the idea that AI has to be all-encompassing to count is simply not true. You don't have to replace your entire IVR to get real value. You can start by handling something specific, like after-hours account lockouts, and solve a high-volume, high-ROI problem that people are perfectly happy to let AI take off their plate.

That's the unlock. Customers locked out of their account at 11pm don't want to wait until morning. They just want back in. An AI agent that resets a password or unlocks an account is solving a real problem, in a low-stakes lane, where almost nobody is upset that a human didn't pick up.

It's not glamorous. It's not the headline-grabbing voice AI demo. But it's the smartest possible place to begin.

Crawl, walk, run

Karan keeps coming back to a crawl-walk-run model, and it maps cleanly onto how the smartest institutions are moving.

Crawl is starting with employee-facing tools, where the stakes are lower and the upside is immediate. Roughly 80% of the market is most excited about AI here right now: knowledge assistants, training, agent support, the things that make a frontline employee faster and sharper without ever touching a customer directly.

Walk is bringing AI into self-service channels that customers already use and trust. Everyone fixates on voice AI because it feels cool and futuristic. But almost everyone still uses a search bar, and most people who end up calling tried to find the answer themselves first. Making search and digital self-service genuinely good is some of the lowest-hanging fruit in the entire institution, and hardly anyone is doing it well.

Run is the customer-facing voice and conversational AI that most institutions are still cautious about. That caution is reasonable. It's also temporary. The point of crawl and walk isn't to avoid run forever. It's to earn your way there with proof instead of hope.

Why starting small actually works

The reason this approach wins isn't just risk management. It's adoption.

AI only delivers value if people give it a chance. That's true for customers and it's true for employees. You can build the smartest tool in the world, but if a frontline team doesn't trust it or doesn't use it, the ROI never shows up. Forcing a tool on people who've had one bad AI experience just hardens their resistance.

A small, obviously useful first win does the opposite. It gives people a reason to believe. A customer who gets unlocked in thirty seconds at midnight walks away thinking the AI is competent. An employee who gets an instant, accurate answer from a knowledge assistant starts reaching for it on the next call. Confidence compounds. So you build the next thing on a foundation of trust instead of a foundation of skepticism.

The real risk is doing nothing

Here's the part most leadership teams get wrong. They treat waiting as the safe option. It isn't.

As Karan put it, the riskiest thing you can do right now is nothing. The institutions moving early, even in small ways, are building competencies and confidence that compound. The ones waiting for the perfect moment, or the perfect all-in deployment, are quietly falling behind while telling themselves they're being prudent.

You don't fix that with a moonshot. You fix it by picking one problem worth solving, solving it well, and letting the win earn you the next move.

Start small. Start safe. Start now.

Hear the full conversation between Posh CEO Karan Kashyap and Shawn Cordner on the Amplitude of Tech podcast: Build What Differentiates, Buy What Accelerates. Listen on Apple Podcasts or Spotify.