A cession accounting manager at a mid-market reinsurer needs to hire. The last operator left three weeks ago, the bordereaux are piling up, and the remaining team is running at 120%. She opens the HR system and starts writing a job description.
"Proficiency in DXC's SICS platform. Experience in treaty accounting. Attention to detail."
She stares at the screen. None of those describe the actual job. The real work is six tasks that have never appeared in any posting she has ever written — tasks that no training manual covers, no certification tests for, and every new hire discovers the hard way after six months of mistakes.
Brisc AI is an insurance-native AI platform built specifically for reinsurance cession accounting. DXC's Strategic Information and Cession System runs the reinsurance back office at over 100 reinsurers in 38 countries. SICS is a powerful system of record. But six operational tasks sitting between each cedant's raw documents and SICS's structured entries are performed entirely by humans — and each one depends on accumulated knowledge that human memory cannot sustain reliably at scale.
Here are the six tasks. Each one is knowledge-dependent, not speed-dependent. Each one degrades with turnover. And each one is a better fit for a system that remembers everything it has ever processed than for a person who will eventually forget.
Every cedant sends bordereaux in a different layout. Cedant A sends a multi-tab Excel workbook with premium in Tab 3, Column F. Cedant B sends a PDF with tables that change column order every quarter. Cedant C sends everything in the body of an email. The operator's first task — before any SICS entry is even considered — is recognizing which format arrived and where the fields are in that specific document.
This task looks trivial from a distance. A reinsurer processing 3,000 bordereaux a year across 400 contracts encounters hundreds of distinct document layouts. A senior operator recognizes Cedant A's format on sight. A new hire spends 15 minutes figuring out what they are looking at — and misses a nested tab 30% of the time until they have seen it three or four cycles.
Format detection is a pattern-recognition task that compounds with exposure and degrades with turnover. It does not require judgment. It requires memory.
The cedant's bordereaux says "commission." SICS expects three separate entry codes: ceding commission, override commission, and profit commission. The operator's job is to read the cedant's single "commission" column and split it into the right buckets — every month, every cedant, without error.
This is the task that separates a clean close from a forensic one. An operator managing 50 cedants carries 50 different commission-splitting rules in their head — which contracts have sliding scales, which have profit-share true-ups, which bundle brokerage into the commission line. Brisc maintains 97%+ accuracy on bordereaux reconciliation because this classification logic persists permanently in the system's per-cedant dictionary, not in a person's working memory that resets with every resignation.
Not every discrepancy is a real problem. Some are timing differences — the cedant booked in March, the broker confirmed in April. Some are FX rounding artifacts. Some are format changes the cedant made without telling anyone. And some are genuine mispostings that, if left uncaught, compound into next quarter's reconciliation failure.
A senior operator can look at a $47,000 discrepancy and know within minutes whether it is a timing gap or a classification error, because she has traced similar discrepancies through months of broker correspondence before. A new hire treats every discrepancy the same way: escalate, ask a colleague, wait for an answer. The exception queue backs up. The real mismatches get lost in the noise.
McKinsey and Accenture estimate that 30-40% of underwriter and operations time in insurance goes to administrative tasks like these. The more dangerous cost is not the time — it is the accuracy that fades as the queue grows and the experienced operators burn out.
Case reserves are not a transaction — they are a snapshot that changes every reporting period. SICS records the delta from the previous month, not the absolute number. The operator's job: pull this month's reserve from the cedant's bordereaux, find last month's posted entry, calculate the change, and post the difference.
The work multiplies when the cedant restates a prior period — and many do, especially in long-tail casualty lines where loss development extends 5 to 10 years. A senior operator who has handled Cedant F's bordereaux for three years knows that Q3 restatements are routine and pre-flags them. A new hire books the initial figures as final and discovers the error three months later during the audit.
Cross-period reconciliation is a memory task disguised as a math task. The arithmetic is trivial. Knowing when to re-run the arithmetic is the hard part — and that knowledge walks out the door every time someone leaves. The insurance back-office experiences 20-40% annual turnover, making this the most fragile step in the entire workflow.
Regulators, auditors, and internal compliance teams need to know why a particular classification was made. Was Cedant D's premium adjustment posted against Q2 or Q3? Why was that brokerage line separated from commission? What source document drove the entry?
In most reinsurance operations today, the audit trail is reconstructed after the work is done — not captured as the work happens. The operator makes the classification, posts the SICS entry, and moves on. Three months later, when the auditor asks, someone has to rebuild the logic from memory, from email threads, or from the document itself.
The Brisc Reconciliation Analyst captures the audit trail as a byproduct of the classification process. Every entry ships with the source document, the extracted fields, the classification logic, and a confidence score. Auditors, regulators, and quarterly reviews can reconstruct any decision at any time — because the decision was documented when it was made, not when someone asked about it.
Cedant report formats change. Broker remittance structures evolve. Commission definitions get renegotiated mid-treaty. These shifts do not arrive with a notification — they show up as a column that moved, a label that changed, or a subtotal that stopped reconciling.
A senior operator notices these changes because she has enough pattern history to spot when something is different. She remembers that Cedant E's brokerage split was always 2.5% until the contract renegotiation last September, and she flags the new rate when it first appears in the January bordereaux. A new hire does not know what the old rate was, so the change slips through undetected until it causes a downstream mismatch.
At a reinsurer with 50 cedants and 300+ entry decisions a month, pattern monitoring is a continuous background task. One named Brisc customer, Helix Underwriting Partners, reports an 80% reduction in manual labour after deploying Brisc — a figure that reflects how much of this pattern-matching and monitoring work can be retained permanently rather than rebuilt with every hire.
These six tasks are the real job. They are knowledge tasks, not data-entry tasks. They depend on accumulated memory of each cedant's specific quirks, formats, and history — memory that fades with fatigue, disappears with turnover, and cannot be transferred in a two-week handover. A new hire takes 90 to 180 days to reach productive independence, and each new cedant resets the clock.
The question for a reinsurer is not whether these tasks can be done faster. It is whether they should be done by a system that remembers, or by a person who will eventually forget.
| Task | Today | With the Reconciliation Analyst |
|---|---|---|
| Format detection | Human pattern memory | Permanent per-cedant format dictionary |
| Classification | 30-50 cedant rule sets in one head | 97%+ accuracy; exceptions routed to human review |
| Exception triage | Trace history in email and docs | Auto-classified by type; human reviews flagged items |
| Cross-period reconciliation | Remembers restatement patterns | Delta logic per cedant, per period; restatement-aware |
| Audit trail | Reconstructed after the fact | Captured in real time with source, logic, and confidence |
| Pattern monitoring | Background human awareness | Drift detection across all cedant-month combinations |
Deployments of the Brisc Reconciliation Analyst typically go live in 2-6 weeks, depending on cedant complexity and the initial dictionary coverage. The dictionary starts learning from day one and improves with every cedant cycle reviewed.
Q: Does Brisc replace SICS? No. Brisc is the layer between cedant documents and SICS — it reads unstructured bordereaux and outputs structured SICS entries. SICS remains the system of record for treaty management and cession accounting. Your existing workflow stays in place.
Q: How does Brisc learn each cedant's format and classification rules? During onboarding (typically 2-6 weeks), Brisc builds a per-cedant dictionary that maps document layouts, field names, and classification patterns to your SICS entry-code structure. The dictionary improves with every reporting cycle as new edge cases are reviewed and absorbed. This accumulated knowledge persists permanently — it does not leave with any individual operator.
Q: What happens when a cedant changes their bordereaux format without notice? Brisc's format-detection layer flags deviations from the learned pattern and routes them to human review. The first occurrence of a new format is a human decision; every subsequent occurrence is handled automatically. Format drift is one of the primary patterns the system monitors continuously.
Q: What about ACORD ADEPT — doesn't that solve the format problem? ACORD ADEPT solves the structured-data exchange problem between parties that agree to use a common digital format. ADEPT transactions grew 73% from 2024 to 2025, and the growth is real. But ADEPT works when both sides adopt. Roughly 80% of cedants today still send unstructured documents — PDFs, Excel files, emails. Brisc handles the unstructured 80%. They are complementary, not competing.
Q: How do we know Brisc's classifications are correct? Every classification ships with a confidence score and a full source-document reference. High-confidence entries process automatically; low-confidence entries route to your operator for review. The audit trail captures the source document, the extracted fields, the classification logic, and any human review applied — confidence-scored, source-traceable, fully reproducible by design.
Q: Is this only for reinsurers running SICS? No. The architecture is platform-agnostic. Brisc works with SICS, Sapiens ReinsuranceMaster, Verisk Sequel, and in-house systems. The per-cedant dictionary changes per platform; the classification and monitoring logic does not. SICS is the focus of this post because it is the largest deployed reinsurance admin platform.
Q: Can Brisc handle cedants we have never processed before? Yes. Brisc is trained on bordereaux as a category, not on individual cedants. New cedants generally process on the first pass; unusual formats get flagged for review and added to the dictionary immediately. Each new cedant compounds the system's knowledge — the next person to join your team inherits the full accumulated dictionary from day one.
Q: What does the operator actually do once Brisc handles these six tasks? Exception handling, cedant relationship management, audit support, and the judgment calls that require human expertise rather than human memory. The operator shifts from being a translator to being a reviewer and decision-maker — the work they were actually hired to do.
If your cession-accounting team's real job is these six tasks, the question is whether those tasks should depend on who happens to be at the desk this quarter — or on a system that retains every cedant's patterns permanently.
The institutional knowledge should belong to the reinsurer, not to whoever is reading the bordereaux today.