Business verification often stalls when evidence moves out of official registries and into the real world. A submitted business address might look valid on paper, yet turn out to be a commercial maildrop, a virtual suite, or an empty lot. When automated data runs out of answers, verification falls back to an uploaded document and an analyst review queue.

Historically, that meant analysts had to take on manual investigative legwork: opening Google Maps, scanning Street View imagery, zooming in on storefront signage, reviewing uploaded PDFs for tampering, and manually reconciling line items.

Middesk Agents automate that investigative legwork. Instead of returning passive flags, agents conduct focused investigations and deliver evidence-backed findings directly inside the verification workflow.

Today, we are highlighting a bundle of major platform updates: visual inspections for address verification, automated document analysis, custom operating rules in the dashboard, and a complete, auditable decision trail across every run.

Address Verification and Address Type Risk agents now incorporate visual inspection directly into the verification run.

When an address warrants a closer look, agents retrieve and evaluate Google Street View imagery to answer a fundamental question: does a real business operate at this location?

To make that determination, the agent conducts a multi-step visual review:

• Multi-angle inspection:​ Examines imagery across multiple camera headings to assess the full building exterior and surrounding context, avoiding reliance on a single static frame.
• Exterior signage analysis:​ Reads visible exterior signage, matches branding against the applicant business name, and looks for signs of active commercial operation.
• Property classification:​ Distinguishes commercial storefronts, corporate offices, and industrial sites from residential homes, postal maildrops, or vacant land.
• Auditable visual proof:​ Embeds the reviewed imagery, camera angles, detected signage crops, and a composite legitimacy assessment directly into the analyst review panel and audit PDFs.

Instead of an analyst spending 5 to 10 minutes manually inspecting a map, reviewers receive a clear assessment supported by visual proof in seconds.

When official registries lack sufficient information to verify a business, teams turn to uploaded documents. But document review has always been a time-consuming bottleneck: opening PDFs, checking for forgery, reading details by eye, and reconciling numbers against state filings.

Our new Document Verification agent automates that review process:

• Tampering and fraud screening:​ Evaluates uploaded files for digital alteration, inconsistencies, and manipulation before processing information.
• Attribute extraction and reconciliation:​ Reads core business details and reconciles them against verified records, helping turn unverified applications into verified business profiles.
• Broad document support:​ Handles key onboarding documents, including IRS confirmation letters, formation filings, certificates of good standing, leases, utility bills, and bank statements.

Document Verification is currently available to select customers partnering with our team on early rollouts. If document review is a major friction point for your operations, reach out to your Middesk account team to discuss early access.

Automation is only valuable when compliance and risk teams can govern how it runs. We have introduced new controls and documentation that place policy management and integration directly in your hands:

• Self-serve settings in the dashboard:​ Define agent behavior directly in the product dashboard. Choose which agents run automatically, establish confidence thresholds that require human review, configure default behaviors for newly introduced agents, and specify permitted data sources.
• Public developer documentation:​ The full guide for embedding agent findings into custom onboarding workflows and internal applications is now available in the knowledge base.
• Sub-minute turnaround times:​ Dedicated infrastructure and asynchronous evidence rendering have reduced P95 agent turnaround from approximately 2 minutes to roughly 1 minute.

For compliance teams, bank partners, and audit-facing institutions, adopting AI requires total transparency. Inferring what happened after the fact is not enough when defending an onboarding decision to an auditor or regulator.

Agent runs now record a complete, unified audit trail:

• Unbroken chain of custody:​ Tracks every step from the initial agent finding, to the proposed action, to the human or automated decision, the specific actor, the executed action, and the final report change.
• Consistency across surfaces:​ The dashboard review panel, webhook event payloads, and audit package PDFs reflect the exact same attribution history, ensuring internal systems and external audits tell the same story.
• Clear evidence distinction:​ Every finding explicitly separates confirmed primary evidence from a verified absence of evidence, giving reviewers a reliable foundation for escalation or approval.