How Affix started analyzing 10x more contracts with the same team and still raised the quality bar.

Context

Affix is a benefits administrator: it operates in the middle of the corporate health plan market. The core of the operation's work is contract analysis: long, dense documents, with multiple clauses, attachments, regional particularities. Each of these documents goes through human review before becoming a closed contract with the client.

The scale of the work was a classic problem: each contract consumed between 2 and 3 hours of analysis by a qualified analyst. Multiply that by dozens of contracts a month, in a growing operation, and you have an entire team doing document analysis: expensive people, with specific training, doing work that mixes bureaucratic parts (checking a field, cross-referencing information) with clinical-legal judgment calls (deciding whether a clause is adequate, whether there's risk, whether it needs review).

Affix knew it needed to scale this operation. Hiring more analysts was the traditional way out. And it was expensive, slow (training a new analyst takes months), and still didn't attack the real bottleneck, which was the bureaucratic part eating into the time meant for judgment.

The challenge

Just "applying AI" to the contract wasn't enough. A health plan contract document has a complexity that a generic OCR system doesn't cover: specific technical terms, clauses that reference each other, attachments with tables, suspensive conditions, regional particularities. Getting a clause wrong is costly: it can trigger litigation years later.

The AI stepping in there needed to have the level of care of a senior analyst. And it needed to be auditable, because if the system approved a contract with a problem, someone needs to know why.

The solution

NexUnio built a POC with 14 specialized agents, each focused on a specific type of clause or validation. The architecture is modular: the document comes in, gets classified by type, is distributed to the relevant agents, validated in parallel, and consolidated into a final analysis report.

Each agent has a clear responsibility: one validates coverage, another validates waiting periods, another checks reimbursement, another audits the accredited network, and so on, up to the 14th.

When all the agents finish, the system generates a consolidated report with points of attention, a confidence score per clause, and a recommendation. The human analyst receives this report and focuses their time where it matters: on the clauses the system flagged as doubtful.

The side-by-side comparison interface between two contracts was the icing on the cake: when Affix needs to evaluate a contractual change, it sees everything highlighted in seconds.

The results

• Contract analysis went from 2-3 hours to 20 minutes. A reduction of roughly 90% in time spent per document.
• 94% accuracy verified in the pilot. The AI gets it right on the first pass in the vast majority of cases. The remaining 6% are exactly the cases where a human needs to take a look, and the system flags that explicitly.
• Processing capacity increased by an order of magnitude. The operation that used to process dozens of contracts a month can now process hundreds, with the same team.

What this unlocks

Scale freedom here means growing without inflating qualified headcount. Affix's analyst is expensive because they're rare: people who combine healthcare knowledge with contract knowledge. You can't hire ten of them at once. The AI took the repetitive part of that analyst's work and turned it into curation. Now they do what a machine doesn't: edge-case judgment, atypical cases, exception decisions.

That's the difference between "automating to cut costs" and "automating to raise the bar". Affix didn't cut anyone: it freed up the qualified team to do qualified work.