dp.cx blog

A few weeks ago I wrote about what happens when companies stop hiring junior engineers and replace them with AI. The argument was straightforward: if you don't hire juniors today, you don't have seniors in five years. The post was about software engineering because that's the world I work in, but I've been thinking about it in a broader context since then.

My wife works for an insurance broker in the commercial space. Not consumer auto or homeowners - business insurance. The kind where a manufacturing company needs coverage for a plant with 200 employees, or a construction firm needs certificates of insurance for every job site they're running across the country.

Business insurance isn't simple. Every client's policy is different depending on their industry, their size, their risk profile, and the specific things they do. A tech startup's coverage looks nothing like a general contractor's, which looks nothing like a restaurant group's. The people who handle these accounts spend years learning the differences - what riders matter for which industries, which exclusions to watch for, how to structure coverage for a company that operates in twelve states with different regulatory requirements.

It can also be high-volume. A construction company doing jobs nationwide might need dozens of certificates of insurance issued in a week, each one specific to a particular job site and general contractor. The junior who handles those requests learns something with every single one - which requirements are standard, which ones signal a problem, which contractors are easy to work with and which ones will call back three times. All of that builds the foundation for understanding how commercial insurance actually operates on the ground.

And then there are the calls that aren't routine at all.

A CEO calls because there's been a fatality at one of their job sites. They're facing potential OSHA investigations, wrongful death lawsuits, grieving employees, and their own guilt about what happened. The person on the other end of that call needs to know how to handle it - not just the claims process, but the human side. When to listen. When to act. What to say to someone who's in crisis and looking to their broker for guidance on what comes next.

That judgment doesn't come from a script. It comes from years of handling progressively more complex situations, starting with those certificate requests and working up through routine claims and coverage questions until one day you're the person a CEO trusts to help them through the worst day of their professional life.

The relationship dynamic matters here too. This isn't a consumer calling a 1-800 number. A CEO expects to talk to someone who understands their business - someone who's been to their facilities, who knows their operations, who remembers that they expanded into a new market last year and what that meant for their coverage. That trust gets built over time, starting when the junior helped with a straightforward request and gradually took on more responsibility. Replace that pipeline with AI and the senior brokers retire with nobody behind them who knows the clients.

The failure mode is also different from what I described in the tech version of this argument. In software, losing your junior pipeline means you can't staff projects in five years. In commercial insurance, it means a mishandled or delayed fatality claim, which carries real errors and omissions liability for the broker and potentially serious consequences for the CEO's business.

I keep coming back to the same basic problem. AI handles the average case well enough, and business insurance is full of situations that aren't average. Every unusual rider, every edge case in a multi-state operation, every claim that doesn't fit the standard process - those are the moments that require someone who's been trained through years of handling the simpler stuff first, and there's no shortcut for that.

This isn't just a tech industry problem. Any field that depends on deep expertise built through years of apprenticeship-style learning - where juniors gradually absorb institutional knowledge, client relationships, and professional judgment - faces the same pipeline question. If you stop bringing people in at the entry level because AI can handle the simple work, you're making a decision about who's going to be available to handle the hard work in a decade. And in some of these fields, the hard work involves real people in real crises who need a human on the other end of the line.