Actuarial Agents for Group Life Pricing

AI-assisted workflows to accelerate actuarial group life insurance pricing.

Context
My Role: Lead Designer & Facilitator (12-week discovery to concept sprint & product dev)

Problem

Actuaries are stuck doing low-leverage work—manually formatting broker data, chasing down inputs, and duplicating analysis. Pricing cycles drag, errors creep in, and decisions stall. Existing tools don’t accelerate judgment—they just document process.

Solution

We designed an agentic AI workflow that automates repetitive actuarial tasks—flagging missing data, surfacing similar cases, suggesting assumptions, and pre-filling models—so actuaries can focus on pricing logic, not paperwork.

✅ 60–70% reduction in manual data cleanup and formatting time
✅ 3x faster quote turnaround for complex or high-risk cases
✅ 70% fewer broker follow-ups and coordination loops
Sprint Timeline

Week 1

AI Use Case Discovery
Used synthesized insurance AI trends to rapidly identify scalable opportunities.

Week 2-3

Focus: Group Life Pricing
Identified group life pricing as a manual, repeatable, high-value opportunity for agentic AI.

Week 3-4

Journey Mapping
Mapped the current-state workflow and surfaced key breakdowns across intake, modeling, and handoffs.

Week 5-7

Wireframes & Testing
Built and validated low-fidelity wireframes with actuaries to test early workflows and AI handoffs.

Weeks 8-12

High-Fidelity Prototype
Designed and tested a polished agentic experience with stakeholders—resulting in strong buy-in.

Problem Definition & Current Journey

Problem
Actuarial workflows today are slow, fragmented, and cognitively taxing. Actuaries spend significant time on repetitive tasks like formatting broker submissions, cleaning data, and manually validating inputs — leaving less time for modeling and risk evaluation.

Information is scattered across folders and teams, making it hard to access prior assumptions or case precedent. Meanwhile, junior actuaries lack embedded guidance and must navigate complex decisions without real-time support.
❌ 30% of data prep time is spent on manual cleanup, reconciliation, and formatting
❌ 40% longer model run times due to fragmented tools and redundant processes
❌ Manual intake and inconsistent access lead to duplicated effort and late-stage errors
Current Journey

Broker to Actuary Intake

  • Inconsistent formats → Slows down initial review
  • Gaps hard to spot → High risk of missing info
  • No broker feedback → Rework happens late
  • Manual cleanup → Wastes expert time

Data Prep

  • Manual eligibility checks → Slow decisions
  • Back-and-forth cycles → Pricing delays
  • No reconciliation → Errors persist
  • Task status unknown → Pricing stalls

Modeling

  • No case triage → Urgent work gets delayed
  • Manual risk review → Time wasted on low-value work
  • Siloed precedent → Redundant analysis

Solution Requirements & Jobs-To-Be-Done (JTBD)

Solution Overview
Through Agentic Actuary, we designed an AI-powered assistant to streamline the end-to-end actuarial pricing process. The agent automates low-level tasks, flags missing data, surfaces relevant precedent, and assists in generating model-ready quotes — allowing actuaries to focus on judgment, not formatting or follow-up.
Cleaner starting points
Eliminate delays from unstructured or incomplete broker submissions
Embedded decision support
Guide junior actuaries in applying pricing logic and compliance rules with less oversight
Easier access to precedent
Quickly reference past cases and assumptions without digging through folders
Transparency & traceability
Back decisions with clear logic paths and support audits with confidence
Seamless orchestration
Coordinate intake, prep, and modeling with fewer handoffs and errors
Jobs to Be Done – Group Life Pricing
📥 Ingest and clean broker data to begin pricing with minimal delay
⚖️ Evaluate eligibility and compliance by surfacing missing or conflicting inputs
🗂️ Access relevant precedent to understand prior pricing decisions quickly
📊 Assess and generate risk assumptions tailored to each case
🧠 Run models and produce quotes without duplicating inputs or logic
📝 Document and trace decisions to support compliance, audit, and governance
Future Journey – Powered by Agentic Actuary

Broker data intake

🧾 15 formats → 1 format
Guided templates standardize broker submissions.

Input validation

⚙️ Manual rekeying → Zero entry
Real-time checks against model fields and rules.

Data transfer to models

🔁 Multiple revisions → 0 revisions
Structured data flows directly into pricing systems.

Eligibility review

📋 Issues found late → Flagged instantly
AI surfaces missing documents and rule conflicts.

Data reconciliation

🔍 10+ steps → 2–4 steps
Gaps trigger structured follow-up tasks.

Task tracking

📆 Multi-day lag → 1-day turnaround
Live visibility into broker-actuary status.

Case prioritization

🚦 Flat queue → Prioritized cases
High-risk or complex cases surfaced first.

Reference retrieval

🧠 Hours searching → Instant reference
Relevant precedent and pricing rationale pulled instantly.

How It Works
Across the entire actuarial journey, agents built on Microsoft Copilot Studio, LLMs, Azure Form Recognizer, and rule-based systems work together to automate low-level tasks, resolve inconsistencies, and accelerate model-ready pricing decisions.

Intake & Validation

  • Agent: RPA + Form Recognizer
  • Rule: Schema + Field-level checks
  • Value: 15+ formats → 1 format
  • Outcome: Manual rekeying → Zero entry

Eligibility Resolution

  • Agent: Rules Engine / Logic App
  • Rule: Product + Regulatory
  • Value: Auto-triage and eligibility checks
  • Outcome: Issues found late → Flagged instantly

Triage & Reconciliation

  • Agent: Copilot + ML + Power BI
  • Rule: Timeliness + Risk Score
  • Value: Flat queue → Prioritized cases
  • Outcome: 10+ steps → 2–4 steps

Model Input & Reference

  • Agent: LLM + Azure Search
  • Rule: Semantic match via embeddings
  • Value: Surface past cases, auto-fill assumptions
  • Outcome: Hours searching → Instant reference

Compliance Check

  • Agent: LLM + Regulatory DB
  • Rule: Versioned policy validation
  • Value: Real-time assumption checks
  • Outcome: Model ready → Audit safe

Key Feature 1: Auto-Validated Intake & Immediate Risk Triage

Key Feature: Auto-Validated Intake & Immediate Risk Triage

Problem

Actuaries often receive incomplete or fraudulent broker submissions, with missing fields only discovered late in the process—slowing turnaround and requiring manual follow-up. They also assess cases in chronological order, not by risk. Without early risk signals, compliance checks, or precedent context, high-value cases are delayed and effort is misallocated.

Solution

An AI-powered case overview validates submissions, flags issues, surfaces precedent, and runs instant compliance checks. Actuaries get a risk snapshot and copilot support, enabling faster triage, fewer follow-ups, and quicker quotes—reducing turnaround time and improving pricing agility.

🔍 Key Experience Features
📈 Outcome & Impact
💡 Instant triage signal with AI-backed precedent and compliance checks
💬 Embedded copilot to explain assumptions, precedent logic, and risk scores
🧠 Pre-filled inputs + quote recommendation so actuaries can accept/decline with context
✅ 60–70% reduction in manual data cleanup and formatting time
✅ 3x faster quote turnaround for complex or high-risk cases
✅ 70% fewer broker follow-ups and coordination loops

Key Feature 2: Semantic Case & Precedent Search

✨ Key Features

🔎 Search prompts surface relevant filters based on recent actuarial queries
🧠 LLM-powered semantic match using case embeddings + metadata
📊 Benchmarks pulled from high-confidence precedent cases
📁 Case summaries include context, match score, and suggested usage

📈 Impact

✅ 4x faster precedent retrieval for risk comparison
✅ 60% reduction in time spent validating rates with past cases
✅ Improved audit readiness via consistent benchmark documentation
Key Feature: Semantic Case & Precent Search

Problem

Actuaries struggle to quickly find relevant past cases to support current pricing. Critical decisions are delayed while they dig through folders or ping colleagues for precedent examples-slowing risk justification and increasing inconsistency.

Solution

An LLM-powered agent enables natural language case search across a centralized precedent repository. Matching cases are returned with key details and Al-generated benchmarking, enabling actuaries to justify pricing with confidence-faster and with less effort.

Key Feature 3: Rapid Config of Assumptions & Risk Factors

✨ Key Features

🧠 AI-generated assumptions based on risk profile + similar precedent
🎚️ Adjustable sliders to fine-tune assumptions with full control
➕ Add custom risk factors inline—no spreadsheet required
⚡ Run model instantly with pre-filled quote inputs
📉 Compare quote vs. market rates in real time with Copilot agent

📈 Impact

✅ 70% reduction in manual model prep and assumption entry
✅ 3x faster quote generation for non-standard cases
✅ Higher consistency in rate-setting logic across teams
Key Feature: Rapid Configuration of Assumptions & Risk Factors

Problem

Actuaries waste time manually adjusting models and entering assumptions. Even with precedent cases, they must duplicate effort, recalibrate fields, and recheck risk logic—leading to delays and inconsistencies.

Solution

AI suggests assumptions and risk adjustments based on similar precedent cases. Actuaries can fine-tune these with sliders, add new inputs, and instantly run the model—turning decisions into quotes without manual prep.