How Abridge moves at the speed of trust
For an AI company handling patient data, speed and efficiency can’t come at the expense of security and trust. Serval helps them accomplish both.
“About fifteen minutes into the first call, I was like, okay, wait, this is fundamentally different from how ServiceNow is approaching this.”
“About fifteen minutes into the first call, I was like, okay, wait, this is fundamentally different from how ServiceNow is approaching this.”
“About fifteen minutes into the first call, I was like, okay, wait, this is fundamentally different from how ServiceNow is approaching this.”
John Henry Boes
John Henry Boes
John Henry Boes
Head of IT & Automation, Abridge
Senior Manager of Enterprise
Security, Together AI
Head of IT & Automation, Abridge
How Serval Helped
How Serval Helped
How Serval Helped
workflows built
workflows built
workflows built
teams set up on Serval
teams set up on Serval
teams set up on Serval
weeks
weeks
weeks
to clear a 2 month ticket backlog
to clear a 2 month ticket backlog
to clear a 2 month ticket backlog
Company Name
Company Name
Abridge
Abridge
Industry
Industry
Healthcare Technology
Healthcare Technology
Team
Team
John Henry Boes, IT Engineering
John Henry Boes, IT Engineering
Gabe Chan, Staff Automations Engineer
Gabe Chan, Staff Automations Engineer
Zach Powers, CISO
Alex Rodriguez, Legal Operations Manager
Dan Gloede, VP of Finance
Dan Gloede, VP of Finance
Meet Abridge
Meet Abridge
Abridge is a leading AI-native healthcare company on a mission to power deeper understanding in healthcare through purpose-built AI. This year, Abridge is projected to support more than 100 million patient-clinician conversations across 250 of the largest and most complex health systems in the U.S.
Abridge is building in one of the most regulated corners of software. Its product touches patient medical data, which puts it under an exacting security and compliance standard with no margin for error.
Behind a deeply technical workforce whose product gives clinicians their time back, a small team of four IT engineers and five operators needed a secure tool that could do the same for them: free them up for the harder problems Abridge actually needed them on.
Abridge is a leading AI-native healthcare company on a mission to power deeper understanding in healthcare through purpose-built AI. This year, Abridge is projected to support more than 100 million patient-clinician conversations across 250 of the largest and most complex health systems in the U.S.
Abridge is building in one of the most regulated corners of software. Its product touches patient medical data, which puts it under an exacting security and compliance standard with no margin for error.
Behind a deeply technical workforce whose product gives clinicians their time back, a small team of four IT engineers and five operators needed a secure tool that could do the same for them: free them up for the harder problems Abridge actually needed them on.
Abridge is a leading AI-native healthcare company on a mission to power deeper understanding in healthcare through purpose-built AI. This year, Abridge is projected to support more than 100 million patient-clinician conversations across 250 of the largest and most complex health systems in the U.S.
Abridge is building in one of the most regulated corners of software. Its product touches patient medical data, which puts it under an exacting security and compliance standard with no margin for error.
Behind a deeply technical workforce whose product gives clinicians their time back, a small team of four IT engineers and five operators needed a secure tool that could do the same for them: free them up for the harder problems Abridge actually needed them on.
The Problem
The Problem
Legacy tools could route tickets, but not resolve them.
Legacy tools could route tickets, but not resolve them.
When John Henry started at Abridge, IT ran on a legacy help desk tool that in practice was email routed into a shared bucket. Two IT engineers were supporting more than 500 employees, and the backlog had tickets two months old. Employees had stopped expecting IT to answer at all.
When John Henry started at Abridge, IT ran on a legacy help desk tool that in practice was email routed into a shared bucket. Two IT engineers were supporting more than 500 employees, and the backlog had tickets two months old. Employees had stopped expecting IT to answer at all.
When John Henry started at Abridge, IT ran on a legacy help desk tool that in practice was email routed into a shared bucket. Two IT engineers were supporting more than 500 employees, and the backlog had tickets two months old. Employees had stopped expecting IT to answer at all.
The requests themselves were the harder problem. These were not password resets and reports of flaky Wi-Fi. Employees were asking for things like custom secure file transfer setups to drop protected health information to client sites. As John Henry put it, these were engineering questions arriving in an IT queue.
The requests themselves were the harder problem. These were not password resets and reports of flaky Wi-Fi. Employees were asking for things like custom secure file transfer setups to drop protected health information to client sites. As John Henry put it, these were engineering questions arriving in an IT queue.
The requests themselves were the harder problem. These were not password resets and reports of flaky Wi-Fi. Employees were asking for things like custom secure file transfer setups to drop protected health information to client sites. As John Henry put it, these were engineering questions arriving in an IT queue.
01. The old help desk tool was little more than email routed into a shared bucket: no automations, no workflows.
01. Automated workflows were already in place, but they were time-consuming to build, hard to maintain, and impossible to hand off to anyone without technical expertise
01. The old help desk tool was little more than email routed into a shared bucket: no automations, no workflows.
02. Two IT engineers supported 500+ employees.
02. New ideas were always there. The bandwidth wasn't.
02. Two IT engineers supported 500+ employees.
03. Tickets ran two months old, and employees had stopped expecting a response.
02. New ideas were always there. The bandwidth wasn't.
03. Tickets ran two months old, and employees had stopped expecting a response.
04. Operating in a highly regulated industry with no margin for error.
02. New ideas were always there. The bandwidth wasn't.
04. Operating in a highly regulated industry with no margin for error.
They needed a tool to automate the everyday work of running IT. Whatever they adopted had to meet the security standards that Abridge and its customers demanded.
They needed a tool to automate the everyday work of running IT. Whatever they adopted had to meet the security standards that Abridge and its customers demanded.
They needed a tool to automate the everyday work of running IT. Whatever they adopted had to meet the security standards that Abridge and its customers demanded.
The Solution
The Solution
Combining the magic of AI with deterministic workflows
Combining the magic of AI with deterministic workflows
John Henry took his first call with Serval with low expectations. He had used the full gamut of IT tooling, and expected more of the same. But it only took fifteen minutes for him to recognize Serval’s potential to solve the ticket backlog and unblock the IT team. The Serval team proved him right when they shipped a custom integration that would have required months of work in a legacy system, in a day and a half.
John Henry took his first call with Serval with low expectations. He had used the full gamut of IT tooling, and expected more of the same. But it only took fifteen minutes for him to recognize Serval’s potential to solve the ticket backlog and unblock the IT team. The Serval team proved him right when they shipped a custom integration that would have required months of work in a legacy system, in a day and a half.
John Henry took his first call with Serval with low expectations. He had used the full gamut of IT tooling, and expected more of the same. But it only took fifteen minutes for him to recognize Serval’s potential to solve the ticket backlog and unblock the IT team. The Serval team proved him right when they shipped a custom integration that would have required months of work in a legacy system, in a day and a half.
The harder sell was the brand-new CISO, who was appropriately cautious about letting an AI tool near sensitive systems. What convinced him was Serval’s two-agent architecture. Catalyst, the automation agent, builds code-based workflows that run deterministically and have version and run history. Everything is drafted and sent to an admin for review before it’s published.
The harder sell was the brand-new CISO, who was appropriately cautious about letting an AI tool near sensitive systems. What convinced him was Serval’s two-agent architecture. Catalyst, the automation agent, builds code-based workflows that run deterministically and have version and run history. Everything is drafted and sent to an admin for review before it’s published.
The harder sell was the brand-new CISO, who was appropriately cautious about letting an AI tool near sensitive systems. What convinced him was Serval’s two-agent architecture. Catalyst, the automation agent, builds code-based workflows that run deterministically and have version and run history. Everything is drafted and sent to an admin for review before it’s published.
This meant that the help desk agent could use flexible AI to interact with employees and reason over their requests, but any actions it takes against Abridge business systems would be deterministic, auditable, and not subject to hallucination.
This meant that the help desk agent could use flexible AI to interact with employees and reason over their requests, but any actions it takes against Abridge business systems would be deterministic, auditable, and not subject to hallucination.
This meant that the help desk agent could use flexible AI to interact with employees and reason over their requests, but any actions it takes against Abridge business systems would be deterministic, auditable, and not subject to hallucination.
“A less informed person might ask why we’re trusting a platform not to hallucinate when it runs a workflow touching very sensitive environments. It’s not until you understand what the platform is doing. You create the workflow with an LLM, and once the workflow is created, it’s deterministic. It’s a repeatable, executable line of code.”
“A less informed person might ask why we’re trusting a platform not to hallucinate when it runs a workflow touching very sensitive environments. It’s not until you understand what the platform is doing. You create the workflow with an LLM, and once the workflow is created, it’s deterministic. It’s a repeatable, executable line of code.”
“A less informed person might ask why we’re trusting a platform not to hallucinate when it runs a workflow touching very sensitive environments. It’s not until you understand what the platform is doing. You create the workflow with an LLM, and once the workflow is created, it’s deterministic. It’s a repeatable, executable line of code.”
John Henry Boes
John Henry Boes
IT Engineering, Abridge
IT Engineering, Abridge
The Results
The Results
From IT to the whole company
From IT to the whole company
Time to value was immediate on the basics. Roughly 50% of standard access requests were handled by out-of-the-box integrations that took seconds to stand up. 2 weeks after adopting Serval, the IT team cleared their 2 month backlog.
Time to value was immediate on the basics. Roughly 50% of standard access requests were handled by out-of-the-box integrations that took seconds to stand up. 2 weeks after adopting Serval, the IT team cleared their 2 month backlog.
Time to value was immediate on the basics. Roughly 50% of standard access requests were handled by out-of-the-box integrations that took seconds to stand up. 2 weeks after adopting Serval, the IT team cleared their 2 month backlog.
300+
13
50%+
workflows built
proactive agents running
automation rate
300+
13
50%+
workflows built
proactive agents running
automation rate
Today, they have over 300 workflows built and 13 proactive agents running. They’ve achieved a 50% automation rate, which will continue to increase as they roll out access request workflows across all applications.
Today, they have over 300 workflows built and 13 proactive agents running. They’ve achieved a 50% automation rate, which will continue to increase as they roll out access request workflows across all applications.
Today, they have over 300 workflows built and 13 proactive agents running. They’ve achieved a 50% automation rate, which will continue to increase as they roll out access request workflows across all applications.
With the ticket queue under control, John Henry decided to centralize all employee support on Serval. If every team’s requests lived in one place, any of them could become a governed automation, and IT could keep security and compliance review in the loop as usage spreads. With Catalyst, which lowered the barrier to building automations, Serval was easily adopted by 10 Abridge teams including the Finance, Legal, and People teams.
With the ticket queue under control, John Henry decided to centralize all employee support on Serval. If every team’s requests lived in one place, any of them could become a governed automation, and IT could keep security and compliance review in the loop as usage spreads. With Catalyst, which lowered the barrier to building automations, Serval was easily adopted by 10 Abridge teams including the Finance, Legal, and People teams.
With the ticket queue under control, John Henry decided to centralize all employee support on Serval. If every team’s requests lived in one place, any of them could become a governed automation, and IT could keep security and compliance review in the loop as usage spreads. With Catalyst, which lowered the barrier to building automations, Serval was easily adopted by 10 Abridge teams including the Finance, Legal, and People teams.
Finance
A travel request triggers Serval to reach into Ramp, spin up a virtual card, and set the daily limit from the employee’s role on the org chart. The team is exploring Envoy office access off the same trigger.
A travel request triggers Serval to reach into Ramp, spin up a virtual card, and set the daily limit from the employee’s role on the org chart. The team is exploring Envoy office access off the same trigger.
A travel request triggers Serval to reach into Ramp, spin up a virtual card, and set the daily limit from the employee’s role on the org chart. The team is exploring Envoy office access off the same trigger.
Legal
Alex Rodriguez, the legal operations manager, built Serval workflows to answer questions about BAAs, NDAs, and data-sharing.
Alex Rodriguez, the legal operations manager, built Serval workflows to answer questions about BAAs, NDAs, and data-sharing.
Alex Rodriguez, the legal operations manager, built Serval workflows to answer questions about BAAs, NDAs, and data-sharing.
Serval is now the central tool allowing for faster, more efficient internal support across Abridge. When employees get the answers they need instantly, they can get back to the work they were hired to do.
Serval is now the central tool allowing for faster, more efficient internal support across Abridge. When employees get the answers they need instantly, they can get back to the work they were hired to do.
Serval is now the central tool allowing for faster, more efficient internal support across Abridge. When employees get the answers they need instantly, they can get back to the work they were hired to do.
“I would attribute Serval as to why we made it. Not only made it out of our two month ticket backlog, but why we were able to build so quickly in engineering function and really kind of push the bleeding edge on the type of scale.”
“I would attribute Serval as to why we made it. Not only made it out of our two month ticket backlog, but why we were able to build so quickly in engineering function and really kind of push the bleeding edge on the type of scale.”
“I would attribute Serval as to why we made it. Not only made it out of our two month ticket backlog, but why we were able to build so quickly in engineering function and really kind of push the bleeding edge on the type of scale.”
John Henry Boes
John Henry Boes
IT Engineering, Abridge
IT Engineering, Abridge
What’s Next
What’s Next
Automating access requests
Automating access requests
Abridge is currently migrating their IdP to Okta. Given the list of apps and their Okta groups, Catalyst is rebuilding all the birthright access workflows and access requests across 74 apps, saving the team about 15 minutes per app.
Abridge is currently migrating their IdP to Okta. Given the list of apps and their Okta groups, Catalyst is rebuilding all the birthright access workflows and access requests across 74 apps, saving the team about 15 minutes per app.
Abridge is currently migrating their IdP to Okta. Given the list of apps and their Okta groups, Catalyst is rebuilding all the birthright access workflows and access requests across 74 apps, saving the team about 15 minutes per app.
As Abridge continues its rapid growth, it is expanding its footprint into the public sector, where the security and compliance bar is higher still. Serval scales with the company, a key tool for building and running automations while maintaining security and governance standards.
As Abridge continues its rapid growth, it is expanding its footprint into the public sector, where the security and compliance bar is higher still. Serval scales with the company, a key tool for building and running automations while maintaining security and governance standards.
As Abridge continues its rapid growth, it is expanding its footprint into the public sector, where the security and compliance bar is higher still. Serval scales with the company, a key tool for building and running automations while maintaining security and governance standards.
Takeaway
Takeaway
Serval enabled Abridge to move at AI speed without compromising security and governance.