Skip to content
StrataEdge

AI that shows its work.

We build AI agents that audit and reconcile the documents an operation runs on: insurance statements, commission calculations, loan files, medical records. Each result shows its sources, its calculation, and its confidence score, and anything below the floor goes to a person. We build inside your environment, and a working system takes four to six weeks.

audit-agent.internal
Premium & claims audit — Q2 2026Always-on

Commission calc · MGA 114

Variance

Sources

bordereau_q2_mga114.xlsxpremium $247,450
producer_agreement_114.pdf · §4.2rate 7.5%

Determination

Expected commission$18,559
Reported$30,967
Variance$12,408

Confidence 0.84 — below 0.92 floor → routed to human review

Decision trace recorded

ConfirmRoute to ops

Premium & claims audit agent, specialty insurance · recreated from production, figures representative

Customers

  • BCG
  • Accelerant
  • DigiKey
  • Defense Health Management Systems

Technology partners

  • Snowflake
  • AWS
  • Google Cloud
  • Anthropic

01 / Measured work

The numbers, as delivered.

StrataEdge
Quarterly audit time
−75% verifiedSpecialty insurance group
StrataEdge
Automated accuracy
99% → ~100% verifiedWith human validation
StrataEdge
Annual run-rate savings
$75M verifiedHealthcare org redesign
StrataEdge
Cloud spend
−$50M / yr verifiedMulti-cloud optimization
Allan, as CTO of LoanFlow
Loan closing time
30 days → 24 hrs verifiedAI mortgage platform
Allan, as CTO of LoanFlow
Back-office cost
−75% verifiedAI mortgage platform
Exhibit B / Question-level reviewSample data
A production audit workspace excerpt showing question-level review and completed verification states for policy, transaction, class-of-business, premium, and commission fields

02 / What we build

Three ways into the work.

01

Audit and reconciliation agents

Agents that check documents against the numbers they are supposed to support, reconcile records across systems, and assemble certified records from fragmented sources. They run continuously. Exceptions go to a person, and every result cites its source documents.

  1. 01Ingest
  2. 02Reconcile
  3. 03Verify
  4. 04Communicate

02

AI-ready data ecosystems

We build governed data products with automated lineage, and prepare document collections so models can query them. One government health program is replacing 160+ warehouses with this design.

03

AI operating model and org design

We work out with your leadership who funds the system, who owns it, and when to shut it down.

03 / How it works

Four to six weeks to a working system.

  1. 01

    Discovery

    We document the workflow with your experts and agree on how to measure the result.

  2. 02

    Architecture and build

    We design the data contracts, the controls, the agent, and the review screens, then build them.

  3. 03

    Test and validate

    We run your historical work through the system and tune what gets sent to a person.

  4. 04

    Production decision

    You review the measured results and the controls, name an owner, and decide whether to put it in production.

Compared with the alternatives

A large consultancy

The two principals design and build it with a small senior team. You work with the same people from the first meeting to handoff.

A vertical software product

We build inside your cloud account, on your documents. You own the system and its audit trail.

An internal build

You get a working system in four to six weeks. Your operations team then runs it with runbooks and named owners.

04 / Principals

The people in the pitch stay in the decisions.

Linda Apsley, Managing Partner & CEO

Linda Apsley

Managing Partner & CEO

Works with the executive team on the business case and on how the organization will run the system after launch.

CTO, GEICO$1.2B remit · $150M annual savings
Head Architect, CitiArchitecture across 3,000 engineers
CareerAmazon · Microsoft · Time · Capital One
Allan Carroll, Managing Partner & CTO

Allan Carroll

Managing Partner & CTO

Leads architecture, controls, and engineering.

CTO, LoanFlow30 days → 24 hours · 5.6× productivity
CTO, FamilyLink100M users · 40M monthly active
CareerSingleStore · Microsoft Research

When teams bring us in

Reconciliations are done by hand and leak money.

In specialty insurance, manual commission and bordereaux work leaks 3–8% and takes 3–10 days a month.

A pilot worked, and the audit committee said no.

Human review, controls, ownership, and reporting were never defined. We define them before we build.

The strategy is approved. The system still has to exist.

We build alongside strategy firms, inside their client engagements.

Bring the workflow, the constraint, and the business case.

We’ll tell you whether AI fits, and what we would build in the first four to six weeks. It is a 30-minute working call with a managing partner, no deck.