← Compliance home · Case studies
Your compliance stack captures far more than it reads. The receipt image behind a flagged dinner. The document attached to a consulting payment. Stored, but never opened.
NeoCortex opens them - and applies human-like judgment at machine scale, so your team can verify every flagged item instead of sampling a few, and stand behind each decision when a regulator asks.
Five examples belowWe read the part nobody has time to read - the unstructured images and documents where the real risk hides. Here is what that looks like in practice.
The gap
Expense systems store the receipt image, the menu, the attendance artifact. Almost no one reviews them at scale. The risk does not sit in the expense field - it sits in the picture behind it.
What we did
We built a bot to pull receipts automatically from the expense system, then had AI read each receipt image against the client's own expense policy.
What it found
The AI matched policy at over 90% accuracy. It flagged hard alcohol and brand-name liquor charged against a breakfast meal. It cross-referenced venues against outside benchmarks, including Michelin-star restaurants. The same approach extends to meal splitting and other patterns that only show up once you read the receipt.
Why it matters
Every flagged item can be checked, not sampled. And each decision carries the reason it was made - so it holds up when a regulator asks.
The gap
Approval teams cannot read every receipt against every policy. So they sample, or they rubber-stamp. Either way, most transactions get a fraction of the scrutiny the policy demands.
What we did
We built an AI agent that reads each receipt alongside the client's expense policies, approval matrices, and delegation-of-authority rules. It recommends approve, reject, or defer - with the reasoning written out. A second, different AI model gives an independent opinion before anything moves. Cases that need a person are routed to the right manager with the recommendation and reasoning attached.
What it found
Continuous coverage instead of spot checks. One reviewer can cover ground that used to take a team - without cutting the team.
Why it matters
Every decision arrives with its justification already on the record. Defensible by default.
The gap
Structured spend totals tell you how much went out. They miss the relationships - which recipients hold influence, and which ones sit at the center of several manufacturers at once.
What we did
We built a directed network graph from public Open Payments data covering the top 20 US pharma manufacturers across 2015-2019. We ranked recipient influence using PageRank, surfaced top recipients, and mapped recipients shared across competing manufacturers.
What it found
A clear view of high-influence HCPs and shared-recipient concentration - the patterns that a spend report alone will never reveal.
Why it matters
You can spot transparency and relationship risk early, using evidence that is already public, before it becomes a question you have to answer.
The gap
Third-party risk is hard to score. The data is sparse, the rules are non-linear and contextual, and the population is large. For one Fortune 100 company that meant roughly 6,000 vendors and no simple way to rank them.
What we did
We combined category, country, and industry signals with available domain rules to produce a single risk profile for each vendor - built to adapt as events and conditions change.
What it found
One defensible risk number per vendor, across thousands of them, instead of gut feel or a static checklist.
Why it matters
Diligence goes where the risk actually concentrates. The scoring is repeatable, and it holds up under review.
The gap
Supplier-concentration risk is buried in transaction history. When one CMO or raw-material supplier sits behind a large share of revenue, nothing in the day-to-day data makes that obvious.
What we did
For every finished-goods lot shipped on commercial orders in 2022, we traced each lot back to the CMO and raw-material suppliers behind it, and forward to destination country and revenue - quantifying the revenue tied to each supplier.
What it found
A direct line from a single supplier to the revenue that depends on it.
Why it matters
You can quantify revenue exposure from supplier dependence and act on it before a disruption forces your hand.
We complement the compliance stack you already run. We just read the part nobody has time to read.
See what's inside