RPA + IDP at Scale for a US Background Verification Firm — 3,000 Background Checks Per Day Across 40 Countries
A unified RPA + IDP stack handling identity, employment, education, criminal and credit checks across forty country jurisdictions — straight-through, with full audit chain.
The challenge
The client — a mid-sized US-based background verification firm serving Fortune 500 employers across employment, financial-services and gig-economy hiring — was running a delivery operation that had grown organically over 15 years. They handled background checks across 40 country jurisdictions, with each jurisdiction requiring different combinations of identity verification, employment history, education verification, criminal records search, civil litigation search, credit history (where permitted), professional licence verification and sanctions screening.
The delivery model was deeply manual. A team of approximately 280 verifiers in two delivery centres (one in the US, one in the Philippines) handled the manual research work — logging into employer HR portals to confirm employment, calling universities to confirm degrees, querying public-records databases for criminal history, and assembling the final report.
Average turnaround time was 7 calendar days, which had become uncompetitive — newer entrants were quoting 48-hour turnaround for comparable scope. Cost per check was rising as labour costs increased in both delivery centres. Worse, the manual model was error-prone: the firm's QA team was identifying material errors on roughly 4% of completed reports, which was triggering re-work, client complaints and occasional regulatory issues.
The firm's COO had a clear brief: cut turnaround time by at least half, reduce error rate to below 1%, and absorb the next three years of volume growth without proportional headcount growth.
The approach
We deployed a combination accelerator stack rather than a single product: RPA bots (built on UiPath, the firm's existing platform) for the high-volume, structured-portal-based checks; DocuMage for the document-heavy components (offer letters, degree certificates, ID documents); Multi-Agent Orchestrator (Mo) as the workflow coordinator; and Workflow Automator for the cross-system orchestration.
Phase one was process discovery and segmentation. We worked with the firm's operations team to map all 40 country jurisdictions into a structured catalogue of approximately 220 distinct check types. Each check type was scored on automation feasibility: high-volume / structured-data / known-portal checks (e.g. employment verification through a major US employer HR system) were Band A; medium-volume / semi-structured checks were Band B; long-tail / unstructured checks (e.g. niche-jurisdiction criminal records) were Band C.
Band A — approximately 60% of total volume — went through full RPA automation. We built bots for the top 40 employer HR portals, the major US criminal-records databases, the standard sanctions lists, and the most-used university verification portals.
Band B — approximately 25% of volume — went through a hybrid model: RPA for the portal interaction, DocuMage for the document parsing where the portal returned PDFs, and lightweight human review for the final report assembly. This was the band where the orchestration agent did the heaviest lifting — coordinating multiple parallel checks for a single applicant and assembling the unified report.
Band C — approximately 15% of volume — remained primarily human-led, but with the orchestration agent providing the case workspace, surfacing the relevant historical research the firm's team had done for similar cases, and pre-populating the report skeleton.
The pre-built building blocks
Rather than commission a ground-up build, the engagement leaned on MindMap's pre-built accelerator library — production-tested components that compress what would otherwise be a six-to-nine-month build into weeks.
Multi-Agent Orchestrator
Cross-check workflow orchestration and parallel execution
DocuMage
Document intelligence for credentials and IDs
Workflow Automator
Salesforce integration and case-state management
Escalation System
Exception routing and auto-escalation for blocked checks
The architecture
The platform runs in AWS, with the orchestration and intelligence layers in us-east-1 and country-specific RPA bot fleets distributed across regions to comply with local data-handling requirements. The UiPath estate uses 240 attended and unattended bots, managed through UiPath Orchestrator and monitored 24/7 by MindMap's NOC team.
The orchestration layer uses Multi-Agent Orchestrator built on LangGraph, which treats each check (employment, education, criminal, etc.) as a sub-agent with its own tool stack. For a typical applicant requiring six different checks across three jurisdictions, the orchestrator parallelises the work across six sub-agents, monitors progress, retries failed checks with configurable back-off, and assembles the final report only when all components have returned.
DocuMage handles the document intelligence layer — degree certificates (with university-specific template recognition for the top 800 institutions globally), offer letters, payslips for income verification, driving licences and ID documents. The document layer includes a forgery-detection model trained on roughly 40,000 known-forged documents from the firm's own historical records.
Integration with the firm's existing case-management system (a customised Salesforce instance) is handled via the Salesforce REST APIs, with bi-directional sync of case state, document artefacts and final reports. The firm's clients receive completed reports via their existing API or portal integration; we did not change the customer-facing surface.
The full processing chain — every API call to an external portal, every document extracted, every model decision — is logged with seven-year retention to support the firm's regulatory obligations under FCRA in the US and equivalent regulations in its other markets.
The numbers behind the story
Total daily throughput has risen from approximately 950 background checks per day to 3,000 per day, with the same human team size. Average end-to-end turnaround time has dropped from 7 calendar days to 2.1 calendar days. 74% of checks now complete fully straight-through, with the remaining 26% requiring some human intervention — most of which is light verification rather than full manual workup.
Error rate on completed reports has dropped from 4% to under 0.7% on the firm's QA sampling. The automated pipeline has actually identified categories of error that the manual process was making consistently (such as date-format mismatches between US and EU sources) and the validation rules now catch these systematically.
Commercial outcomes: the firm has been able to take on three new Fortune 500 clients in the past year, including one whose RFP specifically required 48-hour SLA which the firm could not previously credibly commit to. Average revenue per check has risen as the firm has moved up-market into more complex, higher-value check packages.
The 280-person delivery team has not been reduced. The team has been redeployed: from manual verification to high-judgement work (complex case investigation, jurisdiction-specific expertise, client relationship management) and to a new internal product team that is now building automation extensions for additional jurisdictions in-house, with MindMap's accelerators as the platform foundation.
“We had tried point-RPA before and it did not stick — automating individual portals does not solve the operations problem. MindMap treated it as an orchestration problem and the difference was night and day. Three thousand checks a day, 48-hour SLA, and our delivery team is doing the genuinely hard cases instead of the repetitive ones.”— Chief Operating Officer· US Background Verification Firm
Why MindMap was chosen
The firm had previously attempted to automate aspects of the workflow with point RPA vendors and an in-house build. Both had failed to scale because they had treated automation as a per-check problem rather than as a per-orchestration problem — they automated individual portal interactions but did not solve the orchestration, document and exception challenges that made the operation hard.
MindMap was selected for the combination capability — RPA, IDP, agentic orchestration and case-management integration as a single delivery rather than as four separate vendor relationships. Our pre-built accelerators meant the firm was not commissioning a ground-up build; we had comparable orchestration patterns in adjacent BPM contexts that we could demonstrate in production.
The 24/7 NOC model was also a significant factor. Background-check delivery is time-critical — when a major employer client submits a batch of candidate checks Friday afternoon, the work needs to happen overnight. MindMap's NOC monitors the bot fleet around the clock and resolves most exceptions before the firm's day team logs on.
Commercially, our pricing model was outcome-linked: a fixed-fee implementation plus a per-check operational fee that the firm could pass through transparently to its clients. The competing vendors were proposing capital licence models that the firm could not have absorbed.
Related deployments
RPA-to-Agentic Migration
A migration from 38,000 brittle RPA bots to 240 agentic workflows — handling more volume at a fraction of the maintenance burden.
Big-4 AI Delivery Practice
MindMap's platform underpins the Big-4 firm's AI delivery practice — 2,400 consultants delivering AI engagements to the firm's enterprise customer base.
FP&A Reporting Automation
FP&A Insights replaced a network of Excel-driven analyst workbooks with a connected master data table and self-service reporting — taking MIS from 10 days post-close to 1 day, with reach growing from 200 viewers to 20,000.
Want an outcome like this?
Start with a 2-week AI Readiness Sprint. We deliver a prioritised use-case backlog and business case grounded in what's actually buildable with our accelerator library.