Insurance execs keep hearing the same message: Generative AI (GenAI) will transform the business.
After working with carriers and managing general agencies (MGAs) across the market, I have learned that artificial intelligence (AI) initiatives do not fail because the models are weak or the vision is wrong. The failure is much earlier, when the very first record is created at intake.
That first record is the moment data about a person, business, or risk enters the ecosystem. The quality of that record drives how quickly a workflow moves, how many exceptions appear, how often teams request follow-up information, and how often transactions stall over missing details. It also shapes model performance across predictive scoring, routing, appetite alignment, product recommendation, and quote optimization. Clean records make a lot of AI "magic" look surprisingly achievable.
AI Is here
It has taken a minute, but the industry has finally moved past the hypothetical stage for AI. In a June 2024 Deloitte survey of 200 U.S. insurance executives, 76 percent said GenAI capabilities had been implemented in one or more business functions, and property and casualty (P&C) respondents, in particular, reported 70 percent. AI now lives in real insurance workflows, and production AI does more than create value; it magnifies the workflow quality already in place.
Clean, structured, and consistent intake records give predictive scoring, routing, appetite alignment, product recommendation, and quote optimization solid operating ground. Patchy or ambiguous records force those same capabilities to work harder for a weaker result. Deloitte's research on scaling GenAI highlights weak data foundations, legacy IT, and inadequate business-tech collaboration as core reasons GenAI implementations fall short of expectations.
When leaders feel disappointed with AI ROI, causation suggests models are working, but the inputs are not reliable. Then, lack of context leads to "hallucinations" or assumptions by the AI solution in question. The results in the workflow continuously tripping over missing data. That is, fundamentally, an intake problem.
Intake quality is a strategic lever
The first record created at intake quietly determines how quickly a workflow moves from submission to quote to bind, how many exceptions appear, how often teams go back to the agent or customer for clarifications, and how well predictive models perform.
Intake gaps and inconsistent fields create costs the organization absorbs later. Extra back-and-forth slows the workflow. Re-rating and re-underwriting consume capacity. Preventable quote fallout erodes confidence. AI initiatives under-perform because the data was never stabilized upstream.
Industry conversation often frames "digital transformation" and "AI-first operating models" as technology projects, but the most successful initiatives treat intake quality as core infrastructure. That is where pre-fill earns its place.
Pre-fill as the intake foundation
Pre-fill sometimes gets described as a convenience feature that results in fewer keystrokes and a better user experience. In an AI-enabled property and casualty (P&C) operation, pre-fill functions as a foundational control that normalizes and standardizes how opportunities enter the system.
Pre-fill pulls from trusted data sources to populate and standardize key fields the moment a record gets created. It reduces manual entry, enforces consistent formats, and improves completeness before a human or a model has to act. Teams need consistent taxonomies for entities, exposures, industries, and products, plus enrichment and identity resolution that connect each record to loss history, firmographics, and geospatial context.
Where pre-fill meets modern distribution
Business now enters the ecosystem through agency and managing general agency (MGA) portals, comparative raters, embedded distribution paths, and online funnels that create partial records from quote-starts and lead forms.
Each channel has its own quirks, standards, and norms (Translation? Different ways of formatting data.) Without pre-fill and a consistent intake approach, every new path increases the probability of mismatches and downstream friction, and the impact compounds exponentially over time. Cleaner intake records reduce touch count. Lower touch count compresses cycle time. Faster cycle time strengthens responsiveness across distribution relationships.
Insurance companies and MGAs making intake a design priority consistently see more meaningful gains in quote delivery speed, accuracy, and close ratios. The carriers struggling with AI ROI share one thing in common: They treated intake as a data-entry problem when it was always a decisioning one.
Small data problems, big friction
The most painful data quality issues rarely show up as dramatic failures. The impact of repeated clarifications, small corrections, and exceptions that drag people into manual work, is easier to overlook or miss as it happens incrementally.
Names are a simple example. Many systems still struggle with punctuation in identity fields. Pew Research found that five percent of women in opposite-sex U.S. marriages hyphenated their last name upon marriage. That looks like a small percentage until a quoting flow handles volume. A system that mishandles a hyphen creates downstream matching issues across verification, billing, and servicing. Similar small issues appear with address normalization, business name variants, and classification codes. None of them make headlines, but all of them add friction the organization pays for later.
A practical plan for faster AI ROI
Teams expect productivity gains from GenAI and predictive tools. According to McKinsey's November 2024 insurance analysis, users effectively leveraging GenAI can boost productivity by over 20 percent. Those gains show up when teams operationalize the capability inside live workflows.
Leadership teams should treat pre-fill as intake infrastructure for AI-enabled operations. A practical plan starts with five steps:
- Assign intake ownership. Give a senior leader accountability for intake quality across underwriting, distribution, and operations.
- Define measurable outcomes. Set targets tied to touch count, time-to-quote, time-to-bind, exception rate, and quote fallout driven by missing or inconsistent fields.
- Map intake sources and failure points. Identify where records degrade across portals, platforms, MGAs, and embedded paths, then connect the degradation to cost and delay.
- Deploy pre-fill in the highest-volume pathways first. Expand channel by channel as metrics move.
- Connect pre-fill to decisioning use cases. Tie input quality to product recommendation and quote optimization so teams can measure downstream impact.
As GenAI investments accelerate, smoother workflows will reward teams that start with clean records. Pre-fill creates the necessary foundation at intake, where small improvements compound quickly. That is why I say most AI ROI dies at intake, and why much greater benefits become achievable from the very beginning.
Jennifer Linton is the founder and CEO of Fenris. She can be reached for further comment or information at jen.linton@fenrisd.com.
Opinions expressed here are the author's own.
(Featured image credit: Emi/Adobe Stock)
© Arc, All Rights Reserved. Request academic re-use from www.copyright.com. All other uses, submit a request to TMSalesOperations@arc-network.com. For more information visit Asset & Logo Licensing.