Two underwriters open near-identical submissions on the same morning. One quotes standard terms. The other adds a surcharge, and neither can point to a guideline that explains the gap.
That gap rarely traces back to talent. It traces back to what each underwriter had in front of them at the moment of decision: different data, different references, different time pressure, and different unwritten rules.
This article breaks underwriting variance into four sources, shows how to test your own team for each one, and explains where AI assistance fits without taking the decision out of an underwriter's hands.
Is Every Difference Between Underwriters a Problem?
No. Underwriters are paid for judgment, and two experienced people can reasonably price a complex risk differently. The problem is unexplained variance, a difference that nobody can trace back to the risk itself.
Judgment Versus Variance
A defensible difference has a reason tied to the risk. One underwriter weighs a loss history more heavily, or makes a deliberate appetite call on a class.
Unexplained variance has no such reason. A simple test separates the two: ask each underwriter to name, in one sentence, the feature of the risk that drove the decision. If the answer points to the risk, that is judgment. If the answer points to what they happened to see, that is variance.
What Unexplained Variance Costs
A quote that is too high loses the submission. A quote that is too low writes the risk for less than it is worth.
Neither error shows up on a dashboard until loss experience arrives. That delay is why variance survives for long stretches inside teams that believe their standards are applied evenly.
What Causes Inconsistent Underwriting Decisions?
Inconsistent underwriting decisions come from four sources: different input data, different references, different time pressure, and judgment that lives in people instead of in the system.
The strongest recent evidence comes from the Underwriting Edge 2026 survey of 350 senior commercial P&C underwriters in the US and UK, fielded in June 2026 and reported by Insurance Business. Asked what damages decision quality, respondents named inconsistent submission data (44 percent), pressure to bind quickly (38 percent), and no context on similar prior risks (35 percent).
The survey was published by an underwriting software vendor, a point the trade press flagged, so treat the figures as one data point rather than an industry benchmark. They line up with a pattern most underwriting leaders recognize from file reviews. The table below turns that pattern into a working diagnostic, which we call the Four-Source Variance Check.
Source | What it looks like | Where to look |
|---|---|---|
Input variance | Two underwriters work from different versions of the same facts | The data each file held at decision time |
Reference variance | Neither saw how similar risks were written or how they performed | Whether comparable prior risks were consulted |
Time variance | The same risk gets a different depth of review depending on queue pressure | Time on file and time to bind across desks |
Judgment variance | The decision follows an unwritten rule held by one senior person | Whether the stated reason exists in any guideline |
Input Variance
Submission data reaches each desk in a different shape. One underwriter sees clean loss runs. Another sees a spreadsheet re-keyed from a PDF.
Policy, claims, and billing history also sit in separate systems, so what an underwriter sees depends on which screens they open. Our piece on data silos in P&C insurance describes how an underwriter ends up pricing a renewal without seeing the claims that happened mid-term.
Reference Variance
Without a shared view of prior comparable risks, each underwriter rebuilds a benchmark from memory. Two people's memories of "similar" rarely overlap.
The result is two reasonable decisions built on two different yardsticks. Neither underwriter is wrong by their own reference, which is exactly why the gap is hard to spot.
Time Variance
When a broker needs an answer by end of day, the depth of review shrinks to what fits. The same risk gets a full file review on a quiet Tuesday and a skim on deadline day.
Nobody records this difference. It lives in the gap between how long a file took and how long it needed.
Judgment Variance
Senior underwriters carry rules that no guideline records: which classes to avoid at which limits, which broker submissions need a second look, which exposures behave differently than the rating plan assumes.
Those rules shape decisions on their desk and nowhere else. A colleague one floor away prices the same risk without them.
Why Don't Written Guidelines Close the Gap?
Written guidelines state the rule for the standard case. Risks arrive with exceptions the rule never listed, and the exception is where two underwriters diverge.
Guidelines Cover the Standard Case
A guideline can say what to do with a frame-construction property inside a protection class. It cannot list every combination of construction, occupancy, loss history, and limit that lands on a desk.
In the space between the listed cases, each underwriter fills the gap with their own experience. That is a reasonable thing to do, and it is also the largest source of variance a guideline cannot see.
Referrals Move Variance Instead of Removing It
A referral sends a hard case up the chain, where a different person applies a different read. Authority limits decide who makes the call, not how the call gets made.
If the person receiving the referral sees less than the person who sent it, the referral adds a second point of variance.
Faster Intake Does Not Repair the Inputs
Speeding up intake gives each underwriter more time. It does not give each underwriter the same facts, the same references, or the same unwritten rules.
Time pressure is one of the four sources, so recovered hours help there. The other three need a different fix.
How Do You Find Where Your Own Variance Starts?
Run a matched-pair review. It tests your team against the Four-Source Variance Check using decisions you already made, so it needs no new tooling.
Run the Review
Pull matched pairs: risks with a similar class, limits, and profile where two underwriters reached different terms.
Rebuild what each underwriter saw: the data, the notes, the loss runs, and any prior risks they consulted.
Ask each underwriter for the one-sentence reason behind the call.
Tag each gap to one of the four sources.
Count the tags, then fix the largest source first.
Read the Results
If most gaps tag as input variance, rewriting guidelines will not help. The underwriters followed the guideline. They saw different facts.
If most gaps tag as judgment variance, the fix is capture: recording the reason behind each exception so the next underwriter can see it. A mixed result is normal, and it tells you the order to work in.
What Happens When Senior Judgment Leaves?
Senior judgment leaves with the person unless the organization captures it. In the same survey, 44 percent of senior commercial underwriters named losing senior judgment without passing it on as a top-three worry.
A combined 40 percent across the US and UK said their organization captures expertise poorly or not at all. Only 8 percent said their organization is investing in coaching and knowledge transfer over the next 12 to 18 months, which ranked last of ten areas.
Judgment Variance Compounds
Every retirement removes a set of unwritten rules and leaves a desk to rebuild them. Junior underwriters then fill the gap with their own reading of the guideline, and the spread between desks widens.
Our article on what happens to institutional knowledge when experienced adjusters retire describes the same dynamic in claims, where underwriting judgment is named as part of what leaves.
Capture Starts With Decisions, Not a Manual
A manual written after the fact records what people believe they do. A record of each exception, with the reason attached, records what they did.
Start with the exceptions found in your matched-pair review. Each one is a candidate for a documented rule, a guideline update, or a reference example for newer underwriters. The same pattern shows up where similar claims settle at different amounts, and it has the same cure.
Where Should AI Assist in Underwriting Decisions?
AI should assemble and surface information at the point of decision, while the underwriter keeps the call. Underwriters in the survey described the same split: AI takes on data, triage, and drafting, and the underwriter keeps complex risks and the final decision.
Context at the Point of Decision
When forced to choose, 73 percent preferred a live signal on the state of the book over a report that refreshes on its own. Underwriters asked for context over speed.
Map that preference to the Four-Source Variance Check. Consistent inputs close the input gap. A view of comparable prior risks closes the reference gap. Recorded reasons close the judgment gap.
Why a Person Stays in the Loop
An underwriting decision has to be defensible to a broker, a reinsurer, and a regulator. A suggestion from a tool is not a decision, and an underwriter who reviews and approves it owns the result.
Human review also keeps the tool honest. When an underwriter overrides a suggestion, that override is the exact record of judgment that most teams are missing.
How InsOps Helps
InsOps builds an insurance-trained AI that assists underwriting teams with the data behind each decision. LiLa, our insurance-trained LLM, runs inside your own environment, so PII and PHI never leave controlled infrastructure. A person reviews and validates every mapping before it is deployed.
Our Integration Gateway connects to Guidewire PolicyCenter and UnderwritingCenter, so policy and underwriting data from your source systems arrives in consistent Guidewire structures without custom engineering for each source. That addresses one part of input variance: the structure the data arrives in.
Surfacing comparable prior risks and company guidelines at the point of an underwriting decision is a capability InsOps is building toward inside LiLa. It is not a shipped feature today, and it would never replace an underwriter's judgment. The underwriter reviews every suggestion and makes the final call.
If you are evaluating how to make underwriting decisions more consistent without taking judgment away from your underwriters, contact us to talk through what this could look like for your operation.
Frequently Asked Questions
What is inconsistent underwriting?
Inconsistent underwriting is a difference in decisions on similar risks that no one can trace to the risk itself. Two underwriters see comparable facts and reach different terms, and neither can name the feature of the risk that explains the gap.
Why does inconsistent underwriting matter?
It prices similar risks differently inside one book, so some risks are quoted too high and lost while others are written too low. The cost stays hidden until loss experience arrives, which makes the problem easy to miss.
How do you reduce inconsistency without removing underwriter judgment?
Give every underwriter the same inputs, the same references, and the same recorded reasons, then leave the decision with them. Consistency comes from shared context at the moment of decision, not from taking the decision away.
Why is underwriting inconsistency hard to see?
No single dashboard compares what two underwriters saw when they decided. Results arrive late, and each decision looks reasonable when read alone. Variance only shows up when you line up matched pairs side by side.
What metrics show whether underwriting is becoming more consistent?
Track the share of matched pairs with unexplained gaps, the spread in terms on similar risks, the overturn rate on referrals, and the override rate on any suggestion tool. Falling gaps and stable, explained overrides signal progress.
Can AI help underwriters apply guidelines consistently?
AI-assisted tools can surface policy language, comparable prior risks, and company guidelines at the point of decision, while the underwriter reviews every suggestion and makes the final call. InsOps is building toward this inside LiLa, and it is not a shipped feature today.
How long does it take to improve underwriting consistency?
Expect the first matched-pair review to show where variance starts, not to close it. Consistency improves as each source gets its own fix, so plan for several review cycles instead of one. The common pitfall is rewriting guidelines before you know which source drives the gap.
Does stricter guideline enforcement fix inconsistent decisions?
Not on its own. Enforcing a rule on a case it never listed pushes the exception out of sight instead of explaining it. Pair enforcement with a way to record the reason for each exception.