The insurance industry has largely framed AI as a productivity story.
Help an underwriter summarize a submission faster. Extract data from documents. Find information more quickly. Draft communications. Give employees copilots that help them get through more work.
There’s real value in all of that.
But I think it’s also a relatively narrow view of what AI can change.
If every underwriter gets faster, but submissions are still received, evaluated, prioritized, routed, and managed through the same fragmented processes, the carrier itself hasn’t fundamentally changed.
The bigger opportunity is to make the operation more intelligent, not just the individual.
Intelligence needs to move upstream
Today, a lot has to happen before an underwriter can actually underwrite.
A submission arrives with applications, loss runs, schedules, supplemental forms, and other documents. Someone needs to understand what came in, identify what’s missing, determine whether the risk fits appetite, decide where it should go, and figure out how urgently it deserves attention.
AI can help with each of those tasks individually.
But what becomes possible when intelligence is applied across the process instead?
The workflow itself can understand the information that came in. It can identify missing or inconsistent data. It can apply the carrier’s appetite and business rules. It can recognize exceptions, trigger the appropriate reviews, and help determine which submissions deserve attention first.
That changes the role AI plays.
Instead of waiting for an underwriter to open a submission and ask AI for help, intelligence starts shaping the work before it ever reaches them.
The question shifts from “How can AI help an underwriter process this submission?” to “Should this submission be taking up an underwriter’s time in the first place?”
The people closest to the business should have more control
There’s another part of this shift that I think is just as important.
Carriers change constantly. Appetite evolves. Authority levels change. Products expand. Review thresholds move. New risks emerge.
The people running the business understand these changes immediately. The systems supporting them often don’t.
Changing a workflow can still mean submitting an IT request, configuring another system, engaging professional services, or waiting for a vendor.
AI creates the potential to change that relationship.
Imagine an underwriting leader deciding:
“For coastal properties over $10 million in TIV, add a senior underwriter review before the submission moves forward.”
Instead of translating that requirement into a technology project, the business can increasingly communicate its intent directly to the software.
That’s a fundamental shift: technology adapting to the way the carrier wants to operate, rather than the carrier continually adapting to its technology.
What this can look like in practice
This is the direction we’re building toward with Robin, Feathery’s AI operations assistant.
Teams can describe how they want a workflow to operate in plain English and adapt it as business requirements change. Once that work is running, they can also interact with Robin to understand what needs attention and why.
This isn’t about removing the underwriter
As AI becomes capable of doing more, it’s easy for the conversation to turn to autonomy.
I think that misses the point in insurance.
Underwriters bring judgment, experience, and context to complex risk. AI doesn’t make those things less important. It creates an opportunity to apply them more deliberately.
If AI can handle more of the work around the decision, understanding incoming information, applying rules, identifying exceptions, coordinating handoffs, and prioritizing what needs attention, underwriters can spend more of their time on the decisions that actually require underwriting expertise.
The goal isn’t autonomous underwriting.
It’s an operating model that knows when human judgment matters and makes better use of it.
The real divide won’t be between carriers that use AI and those that don’t
In a few years, nearly every carrier will be using AI somewhere.
So simply adopting AI won’t be much of a differentiator.
The more interesting divide will be between carriers that use AI to make existing tasks faster and those that use it to rethink how work happens.
And submission intake is only one example.
The same opportunity exists across renewals, claims, servicing, portfolio analysis, enrollment, and other processes where information, business rules, systems, decisions, and people have to come together.
That’s where I think the next era of insurance AI is headed.
Not more AI tools layered onto existing processes.
Intelligence built into the way the carrier operates.
