Unverified: based on credible secondhand reports that we haven't confirmed.
Boulder Care uses Amperos AI agents for first-touch denials and AR follow-up, calling payers, working portals and filing appeals with write-back to its PM system, and reports $2.9M AR recovered and cost to collect down 56%
Boulder Care, a substance use disorder telehealth provider, deployed Amperos across first-touch denials and AR follow-up. The agents call payers, navigate payer portals, pull EOBs, file appeals and resubmit corrected claims, writing every action back to Boulder's practice management system in real time, while Boulder's billers keep the oldest and most complex claims. Amperos reports $2.9M in outstanding AR recovered, a 56% reduction in cost to collect and 5.3x total ROI 'to date' (reported by Amperos case study, SVP of Finance Mitzi Yue quoted). The page is undated and gives no measurement window or baseline.
Rapid growth was driving up claims volume, and Boulder wanted to keep scaling collections without hiring more billers; voice-only tools would have left portal work, appeals and resubmissions to humans (reported by Amperos)
Springline's Kansas City client accounting services (CAS) team uses Basis to take the first pass on payroll journal entries, investment and bank statements, credit-card activity, undeposited funds, sales tax and analytical review across clients; staff review the output, resolve exceptions and post. Basis reports payroll journal-entry prep fell from 4 hours to 30 minutes, quarterly trust statement processing on one engagement from about 3 hours to 1, and 35% more clients per junior accountant, with full adoption within 4 months (reported by Basis, quoting Springline's Accounting Department Chair; vendor-published)
America's Auto Auction's AP team processes roughly 32,000 invoices a month from 12,000+ vendors across 50+ auctions; Stampli's Billy AI pre-selects vendor, invoice date, description and coding so invoices arrive coded in the queue and specialists key them in under a minute, with native Dynamics 365 integration saving at least 10 hours a week and volume growth absorbed without added headcount (reported by Stampli)
Twistag built an autonomous AP agent for manufacturer Aralab that uses Claude to read varied supplier invoice PDFs and deterministic code to recalculate lines/tax against POs, processing 2,000+ invoices a month in about six weeks with one engineer and redirecting three finance FTEs from data entry (reported by Twistag)
Baird Global Investment Banking deployed Rogo, which orchestrates AI models over third-party and proprietary data and runs prompt-library workflows and agentic frameworks inside Excel and PowerPoint for financial research, modeling, deck creation and client intelligence. Rogo reports 95% engagement across the ~600-user division within a few months, ~10K workflows a week and 250K+ to date, with users saying it gets them '85% of the way there' in seconds and the top 10 power users saving 5+ hours a day on average (reported by Rogo, quoting Baird's COO and bankers; vendor-published)
Growth by acquisition left Bishop Lifting's 45 branches collecting their own way on up to 1,000 invoices a day across 5,000 active accounts. Stuut went live in 6 weeks with one standardized collections process: it automates 91% of outbound collections communications and responds to customer inquiries in under two minutes. Over a 7-month reporting period Stuut reports a 35% reduction in overdue receivables, a 2-day DSO improvement, a $3M working capital benefit and 50% more accounts handled per employee; CFO Jeff Martini says that within 5 weeks of go-live about 60% of the invoice-to-collections headcount cost was redeployed and invoices over 60 days fell by over 40% in the first 90 days (reported by Stuut; vendor-published)
Insurance broker Brown & Brown centralized accounts payable for five divisions with Auditoria AP Helpdesk Agents that watch shared AP inboxes around the clock, verify contacts, detect intent, take in invoices, flag duplicates and send files to Workday; the agents handled over 90% of incoming requests (reported by Auditoria.AI)