Nothing torpedoes underwriting productivity faster than an inbox full of half-completed PDFs, stray spreadsheets, and email chains that stretch for pages.
That messy reality (and how to fix it) was the focus of our recent LinkedIn Live conversation with Lance Poole (Head of AI Underwriting, Next Insurance) and Zach Kahn (Co-Founder, Feathery). Below is a big-picture recap of the hour-long session and the content leading up to it: five ideas you can take back to your own data-intake headaches, plus a link to the replay if you want to dig deeper.
Despite years of buzz around “digital underwriting workbenches,” most carriers and MGAs remain glued to manual hand-offs:
Submission packets arrive in wildly different formats; ACORD PDFs, agent portals, even screenshots.
Critical fields (TIV, loss runs, payroll) often need re-keying multiple times before they hit a rating engine.
Every manual touch introduces delay, error risk, and compliance exposure.
Lance noted that Next began by mapping every single inbound data source, then ruthlessly triaging which documents had to be parsed automatically and which could wait. Zach echoed the point from the platform side: until you see the full inventory of intake “species,” you can’t prioritize the fixes that matter.
Checkpoint: Make a two-column list this week: structured vs. unstructured inputs. Put a dollar sign next to each one to see where the cost of chaos is highest.
Both speakers stressed that AI pays off only when it’s embedded inside a clear workflow. Feathery’s approach layers document intelligence directly onto a low-code form builder, so teams can:
Capture broker or customer input through dynamic forms that hide irrelevant questions.
Extract data from supporting docs (statements of values, loss runs) with a confidence score.
Route the structured output straight into policy or pricing systems without swivel-chair keystrokes.
Next Insurance takes a similar stance in its own stack. Rather than chasing the latest large-language-model headline, the team first asks, “Does automating this step shorten quote-to-bind time or improve loss ratio?” If the answer is yes, they slot AI where it removes the most friction.
Checkpoint: Before you spin up a proof of concept, diagram the end-to-end task on a whiteboard. If you can’t show exactly where the AI reduces clicks or errors, revisit the use case.
When you apply real numbers, many “cool” ideas fail the test. Two scenarios that do pass:
High-volume, paper-heavy lines (e.g., habitat, small commercial) where every submission looks different.
High-severity decisions (e.g., inland marine, E&S property) where a single mis-typed TIV can skew rate adequacy.
Automation in those zones delivers measurable ROI within months. Everywhere else, incremental improvements (pre-fill, smart validation) may be enough for now.
Checkpoint: Attach a simple spreadsheet to every AI request: current FTE hours vs. projected tool costs and error-rework savings. Greenlight only what beats the status quo on paper.
Software alone doesn’t cure bad habits. Lance and Zach both highlighted people issues that can sink even the slickest intake platform:
Change fatigue. Analysts who survived three system rollouts in five years will quietly revert to email unless you sell the “why.”
Control anxiety. Seasoned underwriters worry that algorithms will overrule their expertise; they need transparency and manual-override rights.
Metrics blindness. If leadership never publishes cycle-time or error stats, nobody feels urgency to improve them.
A few tactics that worked for our panelists:
Roll out the tool on one narrow workflow first, say, renewal endorsements, so wins appear within a single quarter.
Keep training short and in context (30-minute “how we do it now” demo beats a half-day classroom).
Broadcast results: “We cut intake time from 3 hours to 45 minutes” resonates far more than abstract AI chatter.
Checkpoint: Pair every tech milestone with a behavior milestone (e.g., 90 % of submissions now arrive through the new form). Celebrate both publicly.
Looking ahead, the speakers agreed on three trends:
Modular stacks beat monoliths. Open APIs and low-code tooling let carriers swap components without ripping out the core.
Data cleanliness becomes non-negotiable. The better your intake, the easier it is to plug emerging AI models into pricing, claims, or fraud workflows.
Talent will follow better tools. Early-career underwriters want to analyze risk, not copy-paste values. Clean data and automation help attract (and keep) that talent.
The takeaway? Underwriting isn’t turning into a “push-button” discipline overnight, but the firms that tame data intake today will out-quote and out-price slower rivals tomorrow.
Manual data entry is the slow leak no one notices until the floorboards warp: lost hours, hidden errors, deals that die on someone else’s desk. The webinar guests showed how careful workflow mapping, low-code tooling, and pragmatic AI give you a patch kit that actually sticks.
Ready to see what streamlined intake could look like for your own operation? Check out Feathery’s adaptive form & document-intelligence platform it’s purpose-built to turn messy PDFs into clean, actionable underwriting data.