Optimizing compliance and intake pipelines for high-stakes litigation
Broughton Partners is a technology-first legal services company that connects people harmed by corporate negligence with top mass tort and personal injury litigating firms across the United States. It runs lead generation, intake qualification and claim confirmation in house, and delivers each firm a fully vetted, compliant case packet.
In this business, quality is the product. Every packet has to stand up to attorney review and outside scrutiny long after it leaves Broughton, and every packet has to arrive on time. Broughton has run its claimant, referral-firm and litigating-firm journeys on Salesforce for more than a decade, so the data was already in one place. The audit work on top of it was still largely manual.
Manual discernment checks created downstream bottlenecks and claimant drop-off
Before any case reached a law firm, the Claimant Services Fulfillment team audited it against more than 20 checkpoints covering records, documents, eligibility and discrepancies. Those checks protect both the claimant and the firm. Repeated by hand on every case, they became the single biggest constraint on growth.
- Late-stage friction: More than 20 binary and discernment checks ran late in the fulfillment cycle. When a problem surfaced there, the case stalled at the worst possible moment.
- Unpredictable volume surges: Case volume spikes sharply at certain points in the year. Hiring for the peaks would have locked in fixed headcount that could not flex down when volume dropped.
- Speed vs. quality: Broughton commits to delivering cases to firms within 72 hours, and refuses to ship a case that hasn't passed quality assurance. Manual audits put those two promises in tension.
- Fragmented evidence: Reviewers jumped between CRM records, survey responses, signed documents and third-party reports, then chased claimants when evidence was incomplete. Every context switch cost time and raised the risk of error.
- Medical verification backlog: In early 2026, about 16.5% of incoming medical data was waiting for review at any given moment, and roughly 2,500 cases were queued for comparison against claims data.
- Data vulnerabilities and drop-off: Delays increased compliance errors and caused avoidable claimant disengagement before packets ever reached litigators.
- Resource strain: Skilled specialists spent most of their day on routine validation instead of the nuanced exception handling only they can do.
“We had to figure out a way to get these cases out the door as quickly as possible while maintaining the high quality of cases that we are known for.”
Elizabeth Gilmore
Director, Claimant Services, Broughton Partners
The goal was clear: more capacity and more consistency, without adding fixed cost and without taking consequential decisions away from people.
A multi-layered autonomous agent framework powered by centralized orchestration
SaasWorx designed and built a framework of 18 embedded, autonomous Agentforce agents running under a central orchestration layer on Broughton's existing Salesforce org. No new system, no rip-and-replace.
The three-lens rule: AI only where it earns its place
Every one of the 20+ checks was scored before a line of code was written. That discipline is what keeps the system predictable and keeps total cost of ownership under control.
Clear-cut, rule-based checks run on Flow and Apex. Agentforce is used only where evidence has to be read and interpreted.
Each check is weighed on its recurring cost, so AI spend goes where it adds value, not everywhere.
The higher the impact of a wrong answer, the more the design leans on deterministic logic and human review.
“We looked at each check through three lenses: whether it needed deterministic logic or AI reasoning, the cost of running it, and how much room there was for error. That helped us use AI where interpretation adds value while keeping clear-cut decisions predictable.”
Pranay Saini
Practice Head – CRM, Data and AI, SaasWorx
What we built
- Four specialized agent domains: Agents are grouped by audit area (identity, fraud, contracts and medical), so each one does a narrow job well and new checks slot in without reworking the rest.
- Multi-modal document verification: Custom prompt templates with multi-modal capability read signed documents and records and compare them against case data.
- Parallel medical verification: When medical records land on a case, six agents assess different evidence at the same time and roll their findings into standardized Salesforce fields plus a single verification result.
- Extensible API integration layer: MuleSoft connects agents to proprietary and government databases through custom and open APIs, and to law-firm delivery channels.
- Cross-object Salesforce validation: Localized validation agents check records across objects directly inside Salesforce, where the case data already lives.
- Event-driven orchestration: Agents fire automatically the moment the information they need arrives, moving checks upstream instead of leaving them to the end of the cycle.
- Agent Console for human-in-the-loop review: Results flow into one console that shows specialists exactly which checks need attention, with the evidence and reasoning behind every flag.
- Built-in governance and audit trail: Agents never reject a claimant. Ambiguous findings always go to a person, and every decision or override is logged with who made it and why, creating a defensible record and a feedback loop for improving the agents.
How AI and deterministic logic split the work
| Layer | Handles | Why |
|---|---|---|
| Flow & Apex | Binary, rule-based checks and final critical actions | Predictable, low-cost, fully repeatable |
| Agentforce agents | Discernment checks, document reading, discrepancy detection, recommendations | Interpretation adds real value |
| MuleSoft | External data sources and delivery channels | Reusable, governed integrations |
| Claimant Services team | Exceptions, ambiguous findings, every consequential decision | Human judgment where it matters most |
Delivery approach
- Discover: Observed the team, reviewed procedures and documented each of the 20+ checks.
- Design: Scored every check through the three lenses and designed the orchestration architecture.
- Launch: Took the first seven agents live to prove value and build trust with the team.
- Scale: Expanded to 18 agents across identity, fraud, contract and medical verification.
- Optimize: Continuously tuning performance and how results are presented to reviewers, using the audit trail.
Accelerated turnaround times and fully automated audit checks
The deployment turned the compliance workflow from a slow, manual bottleneck into a scale-ready intake pipeline, and gave a small team the capacity of a much larger one without adding fixed cost.
- 2x daily audit capacity: The team now audits twice as many cases per day, while quality goes up rather than down.
- 70% more cases verified: The same five specialists verify significantly more cases, focusing their judgment on the files that need it.
- 60% less daily audit work: Repetitive checks moved to agents, freeing the team for nuanced verification and claimant follow-up.
- 16x processing acceleration: Fully automated, orchestrated audit checks cut case processing time by 16x.
- 10% lower claimant drop-off: Faster upstream verification keeps claimants engaged through to delivery.
- Zero-intervention routing: An initial 5% of cases now flow through end to end with no human touch, a baseline that grows as agents mature.
- Catches what manual review missed: In one early case, the duplicate-check agent flagged a match a manual search had not found. The specialist opened the cited file and confirmed the agent was right, a turning point in the team's trust.
- Elastic capacity: Volume surges are absorbed by the agents, not by permanent headcount.
What's next
All 18 agents are live and being refined. Because the architecture is modular, new checks can be added as Broughton's work evolves. Next on the roadmap: extending the agentic model to other departments, raising the straight-through processing rate, and streamlining delivery once a case has cleared every required review.
“Working with SaasWorx has been an absolute delight. Their attention to detail and forward communication style has led to two very successful production deliveries. We are very much looking forward to future partnerships with their team.”
Broughton Partners
“AI agents do not replace human judgment. They elevate our delivery standards, harden our fraud defenses, and focus our team on the cases that need human intervention most.”
Elizabeth Gilmore
Director, Claimant Services, Broughton Partners
Featured on Salesforce.com →