Commercial Insurance

Redesigning Commercial Underwriting for AI

How AI can increase underwriting capacity without removing underwriting judgment.

Start with the decision, not the technology.

Illustrative analysis: This teardown is based on common industry operating patterns and is not a description of work performed for a specific client. Actual workflows, economics, regulatory requirements, and technical constraints vary by organization.

The operating problem

Experts are being used for work that does not require expertise

Commercial underwriting concentrates some of the most expensive judgment in an insurance organization. Yet in most operations, the underwriter's day is dominated by work that does not require that judgment: opening attachments, searching internal systems, retrieving policy history, chasing brokers for missing documents, and reassembling context after every interruption.

The problem is rarely a lack of capability. Submission data exists. Claims history exists. Guidelines are documented. The workflow was simply designed before any system could assemble that context automatically — so the most expensive person in the process became the integration layer.

The current workflow

Submission arrives
Check completeness
Open attachments
Search internal systems
Retrieve policy history
Review claims data
Compare against guidelines
Identify missing information
Contact broker
Wait
Reassemble context
Assess risk
Decide / escalate
Document reasoning
Quote / decline / refer

The underwriter is being used as

Data retriever
Document reader
Workflow coordinator
Policy lookup engine
Analyst
Decision maker

Only part of that work requires scarce underwriting judgment.

The transformation

From manual assembly to a decision-ready brief

Before the mechanics, the destination. This is what changes for the person making the decision.

Current state

The underwriter rebuilds the risk picture before they can underwrite it.

Open submission and attachments
Search policy and claims systems
Retrieve historical context
Identify missing information
Compare against guidelines
Determine what requires escalation

Decision-ready state

The system assembles the context. The expert exercises judgment.

ACME MANUFACTURING

Risk profile: Moderate
Key issue: Increasing loss severity
Missing information: 2026 facility inspection
Recommended path: Senior review before quoting

Decision-ready brief → expert judgment

The goal is not to give the operator more AI to interact with. It is to redesign the workflow so the relevant judgment arrives with the right context already assembled.

Start with the decision

Work backward from the consequential decision

Rather than asking where AI could be inserted into the existing workflow, work backward from the consequential decision. Everything else in the workflow — information assembly, analysis, routing — exists to put an underwriter in a position to make this decision well.

The decision
Should we accept this risk, under what conditions, at what price, and does anything require escalation?

Information required

The decision sits downstream of fragmented information

Risk profile

Loss history

Policy guidelines

Financial strength

Exposure information

Prior customer relationship

Broker context

Third-party risk signals

Underwriting decision

The decision is downstream of many fragmented information sources. Today, the underwriter assembles them by hand, one submission at a time.

The enterprise AI stack

Where the workflow sits in the technology stack

Organizations have invested heavily in the layers below. Business value is only created when those capabilities change how consequential work actually gets done.

Accelerated computing

NVIDIA GPUs / networking / compute foundation

Cloud infrastructure

AWS / Azure / Google Cloud

Data movement & transformation

Fivetran / Airbyte / dbt / Kafka

Enterprise data platform

Databricks / Snowflake / similar systems

Analytics, models & AI

Databricks Mosaic AI / OpenAI / Azure OpenAI / Vertex / SageMaker

Business workflow

The actual operator and decision process

Business outcome

Capacity, revenue, speed, quality, cost, risk, customer experience

Platforms create capability. Workflows determine whether that capability becomes business value.

Illustrative architectural categories — not a recommendation that every organization use these technologies. Platforms such as Databricks can unify governed data, analytics, and AI; the design remains vendor-neutral.

What AI can actually do

The design problem is where judgment belongs

This is not about removing humans from the workflow. It is about separating the cognitive activities a system can perform reliably from the ones that require accountable human judgment.

AI can assist with

Extract information from submissions
Identify missing documents
Classify submissions
Enrich accounts with external information
Summarize loss history
Retrieve relevant guidelines
Compare circumstances against policy rules
Surface anomalies
Calculate predefined risk indicators
Retrieve precedent
Draft an underwriting brief
Recommend a next action
Trigger deterministic routing

Human judgment remains responsible for

Resolve ambiguous or borderline risk
Interpret unusual circumstances
Price complex exposures
Override system recommendations
Evaluate meaningful exceptions
Own the accept / decline / refer decision

Human vs. machine

A clear division of labor

System should

Human should

Retrieve information

Resolve ambiguous risk

Extract structured data

Interpret unusual circumstances

Check completeness

Exercise underwriting judgment

Apply deterministic rules

Override recommendations

Summarize context

Evaluate meaningful exceptions

Calculate defined indicators

Own consequential decisions

Route standard cases

Handle escalation

The goal is not maximum automation. It is maximum leverage of human judgment.

Redesigned operating model

How the workflow should operate

Information assembly, analysis, and routing move to the system. Judgment stays with the people accountable for the decision.

Broker submission

Automatic intake

Completeness check
Document extraction
Enrichment
Classification

Unified risk context

Submission information
Policy history
Claims history
External signals
Underwriting rules

AI analysis

Risk indicators
Missing information
Guideline comparison
Relevant precedent
Recommended next action

Decision router

Standard case

Underwriter review

Exception

Senior underwriting review

Incomplete

Automated broker information request

Human judgment

Quote / Modify / Refer / Decline

Outcome data feeds back into analytics and workflow improvement.

What the operator actually sees

A decision-ready brief, not a chatbot

The difference between giving an underwriter an AI chatbot and redesigning their work: the system delivers a decision-ready brief, not a search box.

Underwriting Brief

Illustrative

Applicant

Acme Manufacturing

Overall profile

Moderate risk

Positive indicators

Strong financial position
Four-year customer history
Improving claims frequency

Risk indicators

Severity increasing
New facility has no recent inspection
Requested limit above peer median

Missing information

2026 facility inspection

Recommended action

Refer for senior review before quoting

Loss severity plus an uninspected facility exceeds the standard authority threshold.

Review submissionRequest informationRefer

Illustrative economics

How we would model the economics

A workflow redesign should ultimately be evaluated in operating terms. In a real engagement, Rivington would model the economics using the organization's actual volume, labor costs, cycle times, exception rates, quality requirements, and implementation constraints.

Illustrative scenario

Inputs — illustrative assumptions

Submissions per week: 80
Current preparation time: 60 minutes per submission
Actual expert judgment: 40 minutes per submission
Total underwriting capacity required: ~133 hours / week

If redesigned information assembly reduces preparation time per submission:

60 minutes 15 minutes

Calculated implication

Potential expert capacity released

~60 hours / week

Before

Underwriter spends significant time assembling context before any judgment is applied.

After

System assembles context; the underwriter spends proportionally more time exercising judgment.

Illustrative model only. Actual economics depend on workflow volume, labor costs, automation feasibility, risk requirements, and implementation constraints.

In a real diagnostic, these assumptions would be replaced with observed workflow data. The purpose of the model is to identify which operating variables determine whether redesigning the workflow is economically meaningful.

Operating variables a real engagement would validate

Submission volume
Preparation time
Underwriter compensation
Authority levels
Exception frequency
Quote conversion
Loss performance

Outcome measures the redesign targets

Underwriting capacity
Submission response time
Quote turnaround
Consistency of risk decisions
Escalation quality

Rivington workflow architecture

The methodology behind every teardown

Nine questions, asked in order, starting from the outcome and ending at measurement.

Outcome

What economic or operating result matters?

Decision

What consequential decision creates that outcome?

Information

What context does the decision-maker require?

Judgment

Where is actual expertise necessary?

AI

Where can probabilistic reasoning improve the work?

Automation

What work is deterministic and repeatable?

Exceptions

Where should humans take over?

Routing

Who owns the next action and when?

Measurement

How will we know whether the redesigned workflow improved?

What a real engagement would examine

An illustrative teardown is not a recommendation

An actual Rivington engagement would validate the workflow against:

Operator interviews
Workflow volume
Actual system architecture
Data availability and quality
Exception frequency
Decision rights
Regulatory constraints
AI reliability requirements
Current labor economics
Integration feasibility
Change-management requirements
Business KPIs

The objective is not to prescribe technology from the outside. It is to determine how the workflow should operate given the organization's actual constraints.

Have a workflow like this?

Bring us one consequential workflow.

Rivington will determine where data, AI, automation, and human judgment should sit — and what needs to change for the workflow to produce a better operating outcome.

One workflow. Three weeks. $9,500 fixed fee.

Explore the AI Workflow Diagnostic Talk about a workflow

More teardowns

Insurance Operations