Heartland Mutual Insurance
What the funded pilot consists of
The six-week funded slice, not a workshop agenda. Heartland Mutual Insurance's team would stand this up in their own account.
The brief goes to Heartland Mutual Insurance's build team; they stand the environment up after security review.
Inherited from the session
- Use case
- AI-assisted claims intake extraction
- Owner
- Alex Chen
- Constraint
- Audit trail required on any automated decision; assisted extraction is acceptable. (Robert Osei, Compliance Officer)
- Next step
- 6-week pilot on 500 anonymised claims
- Spec source
- Reused from prior funded document-pattern pilots
Readiness: data owner identified. Security review needed — allow 5 days.
Reusable pilot setup
Generated from the document-heavy intake pattern and the constraints agreed in the session. The same five requirements apply to every pilot on this pattern.
Document store
Holds the 500 anonymised claim packets.
Identity
Names who can confirm and review extractions.
Logging
Audit trail for every assisted decision.
Review queue
Low-confidence fields go to a human.
Extraction worker
Runs extraction against the agreed sample.
Data: 500 anonymised claims with handwritten notes redacted from production identifiers.
Provisioning happens in the customer's own cloud account, not here.
View technical enable list
# Enable list — customer cloud account # This is a briefing, not a provisioner. services: - document-store # hold 500 anonymised claim packets - identity # named owners and review roles - logging # audit trail for assisted decisions - review-queue # human review for low-confidence fields - extraction-worker # run extraction against the sample set
Starter kit · Document-heavy intake
What the pattern already covers, and where it will still fail.
Known gaps
- Handwritten adjuster notes in the margin
- Multi-page claims with inconsistent field order
- Low-confidence routing not implemented