Insurance regulation
·
In production
Score every fraud referral
the moment it arrives.
Every referral gets a fraud likelihood at intake, trained on what your own past referrals led to. Investigators open the riskiest first instead of working in arrival order.
The issue
Referrals arrive faster
than anyone can read them.
Every referral gets a full manual review
Low-risk and high-risk referrals take the same time to open, so the queue moves in arrival order.
Linked activity is hard to see
Repeat claimants, shared shops and organized rings sit in separate referrals.
No way to measure the work
Decisions vary by reviewer, and leaders can’t see what turns into a case.
What we found
The answer was already in your history.
No one could use it at intake.
Past referrals already showed which ones led to arrests, warrants and prosecutions. Only a small share ever became cases, and nothing at intake could tell those apart from the rest. BEACON learns from that history and gives every new referral one of three levels.
High
Shares the most patterns with past referrals that ended in an arrest, warrant or prosecution. Goes to the top of the queue.
Medium
Shares some of those patterns. Worked once the High referrals are covered.
Low
Looks like past referrals that didn’t lead anywhere. Moved down the queue, but kept on record and searchable.
1 in 5
of High referrals were sent on for investigation
1 in 100
of Low referrals were sent on for investigation
Figures show the share of referrals at each level that were sent on for investigation in a live state deployment, March to May 2026.
What we built
From intake to disposition,
in one place.
Every referral arrives digital and complete.
Referrals from every source flow into one queue. BEACON reads each one and fills in the record, so nobody re-keys a form.
- Online forms, carrier reports and national referral feeds in one queue
- 85% of fields filled in automatically
- Follow-up emails to the reporter sent and saved in the record
A fraud likelihood, with the reasons attached.
Each referral is scored High, Medium or Low by a model trained on your own past referrals and what they led to. Investigators see why, not just a number.
- Trained on historical outcomes: arrests, warrants, prosecutions
- Links to earlier referrals, people and businesses shown inline
- Scored with 85 to 90% accuracy against past outcomes
The queue runs by risk, not arrival order.
High-likelihood referrals go straight to an investigator. Supervisors set first-action deadlines, and nothing sits forgotten.
- Routing rules set by your division, changed by your division
- Dormant cases flagged at 14 days, with the supervisor copied
- Workload and results by investigator, team and period
The case file builds itself as you work.
The score and its reasons carry into the case. Logged activity becomes the narrative, and every outcome goes back into the model.
- Case narrative drafted from logged activity
- About 50,000 past referral records searchable
- Closed cases make the next score sharper
Results
What changed
after go-live.
2×
High-likelihood referrals are twice as likely to be escalated, investigated and prosecuted.
~50%
Fewer referrals needing a full manual review.
18→9 mo
Referral backlog, cut in half at go-live and still falling.
30+ min
Saved every time an investigator opens a case.
Results from a live state deployment of the fraud intake and investigation pack, March to May 2026. Backlog compares queue position at implementation with today.
Engines used
Three of BEACON’s five engines,
configured for fraud.
FIND
Reads every referral
Pulls names, dates, policies and narratives out of forms and attachments, and makes 50,000 past referrals searchable.
ADAPT
Scores and routes
Learns from your past outcomes to score each referral, then applies your division’s routing rules and deadlines.
RENDER
Shows the work
Queues, case narratives and performance reports by investigator, team and period, ready for leadership.
More in Insurance regulation
Other questions
BEACON answers.
Bring your referral queue.
We’ll score it with you.
We’ll load a sample of your past referrals and walk through the first scores with your team in a working session.