SuperDial Has Made 10 Million AI Calls to Insurers and Wants Calls to Become Unnecessary
SuperDial has brought its count of AI calls to insurance companies to 10 million, and now it wants to eliminate them. Clinics send bots to find out payment terms, and insurers respond with automation, so the result is a conversation between two machines. According to the company, a call to an insurer's automated line takes about 4 minutes, versus 17 or more with a live operator. The agent's value should be measured by the number of requests resolved, not by the number of calls.
AI-processed from TNW; edited by Hamidun News
In American medicine, something strange is happening: a clinic sends an AI agent to find out from an insurer whether a procedure is covered, and on the other end of the line, another automated system answers. Two machines talk by voice to pass along data that already sits in someone's database. SuperDial, a startup that builds such agents, has announced 10 million AI calls to insurers and drawn an unexpected conclusion from them: the next step is to make the calls unnecessary.
It is important to note up front that the source material is a byline piece published by the company itself, not an independent investigation. The figures are SuperDial's own claims, and they cannot be verified from outside. Still, the case is instructive for anyone building agents into enterprise processes.
What Problem Does It Address?
Administration in the American healthcare system is a huge cost. Back in 2021, McKinsey estimated that streamlining this routine work could save up to $265 billion a year. Clinics and revenue cycle management (RCM) companies constantly have to check with insurers whether a policy is active, what it covers, whether prior authorization has been approved, and where a claim stands.
Previously, people did this, sitting on the phone for hours. Now agents make the calls instead. But insurers are also deploying voice automation that answers such requests. According to SuperDial, each side is getting better at keeping the exchange going, and neither is getting better at ending it.
What SuperDial Proposes
The company calls its approach Agent-to-Agent RCM. The idea is that success is measured not by the number of calls placed but by the number of requests resolved. The client sets the task, and SuperDial chooses the channel itself:
- If the answer can be obtained through an electronic transaction (EDI), it uses that.
- If the data is on the insurer's portal or arrives by fax, the agent retrieves it there.
- If a call cannot be avoided, the agent takes along everything already known and asks only for what is missing.
This resembles a good secretary: they check the website and email first, and only then dial the number, and they don't ask again what has already been established during the conversation.
Among the stated figures: in successful requests, where the needed information was obtained, a call to an insurer's voice bot takes an average of about four minutes per attempt. A conversation with a live representative takes seventeen minutes or more. The platform works through phone, portals, EDI, APIs, fax, and documents, and it returns structured results to the client's systems and work queues. The company says it complies with HIPAA and has passed a SOC 2 Type II audit.
What This Means for Practitioners
First, the metric. When agents are judged by the number of calls, tickets, or processed emails, the system rewards activity rather than outcomes. SuperDial proposes counting how much work the agent removed. This is a useful framework for any product with agents: if a question can be answered without contacting anyone, the best agent is the one that prevented the contact.
Second, multichannel operation. Voice is not the goal here but a fallback. A reliable agent in a business process should be able to switch between APIs, portals, and documents and keep context between steps, so it doesn't ask the same question twice.
Third, the long-term bet. SuperDial wants to build an AI network for exchanging data between clinics and insurers, in which agents pass verified information directly instead of extracting it through scattered systems. The company names a billion requests handled through such agent interaction as its goal.
Limitations
For now, this is more of a claim than finished infrastructure. Direct agent-to-agent exchange requires insurers to open their channels and agree on shared formats, and they have no obvious incentive to abandon current processes. The four- and seventeen-minute figures apply only to successful requests and do not show how many attempts end in failure. The company also says nothing about the cost of a single request or the savings for its customers.
Still, the main lesson is clear: automating a bad process makes it faster, but not better. If agents on both sides simply accelerate pointless calls, nothing is gained. The gain comes when the process itself shrinks.
- Nov 21, 2026VibeCBS collects project nicknames in advance: the agent finds 2.8 times more documents
- Nov 21, 2026FDE job postings grew 4.6x in a year: what a forward deployed engineer actually does at the client's site
- Nov 21, 2026Sakana AI Opens Applied Research Lab for Market and Economic Modeling
- Nov 21, 2026Meta Is Reportedly Preparing the $300 Muse Charm AI Tamagotchi With a Year of Verizon 5G
Key Questions Answered
What Problem Does It Address?
Administration in the American healthcare system is a huge cost. Back in 2021, McKinsey estimated that streamlining this routine work could save up to $265 billion a year. Clinics and revenue cycle management (RCM) companies constantly have to check with insurers whether a policy is active, what it covers, whether prior authorization has been approved, and where a claim stands. Previously, people did this, sitting on the phone for hours. Now agents make the calls instead.
What SuperDial Proposes?
The company calls its approach Agent-to-Agent RCM. The idea is that success is measured not by the number of calls placed but by the number of requests resolved. The client sets the task, and SuperDial chooses the channel itself: If the answer can be obtained through an electronic transaction (EDI), it uses that. If the data is on the insurer's portal or arrives by fax, the agent retrieves it there. If a call cannot be avoided, the agent takes along everything already known and asks only for what is missing.
What This Means for Practitioners?
First, the metric. When agents are judged by the number of calls, tickets, or processed emails, the system rewards activity rather than outcomes. SuperDial proposes counting how much work the agent removed. This is a useful framework for any product with agents: if a question can be answered without contacting anyone, the best agent is the one that prevented the contact. Second, multichannel operation. Voice is not the goal here but a fallback.
Ship AI Faster with Founders, Engineers & Weekly Architecture Reviews
Inside Vibe Club (hamidun.com/club) — weekly live masterminds with Zhemal Khamidun, Tech Radar, full LMS access, and 370+ production Claude Code skills.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.