DevRev partners with OpenAI to bring employee and customer self-service to enterprise

An employee needs software access. A customer needs help with an order. Sometimes, a quick piece of information is all they need and the request is resolved in moments. Other times, the response they receive is only the beginning: a request must pass through systems, approvals, and teams before anything changes. DevRev and OpenAI are working to bring frontier intelligence into those real-world workflows, so people can get an answer and a way forward.

DevRev brings Computer’s connected business context, permissions, and workflows. Paired with OpenAI’s models, Computer can help teams move from the first question to an authorized next step – whether that means resolving a technical support request, routing an IT approval, or helping responders coordinate an incident. What matters is that people know what happened and what needs to happen next.

A set of autonomous AI agents are used to carry out this work. These operate across employee and customer self-service, following the organization’s rules for actions and approvals, and bringing in a person when access, judgment, or an exception requires one.

Trust depends on the details

Before Computer can help move a request forward, it needs to understand who is asking, which information applies, and what action is allowed. Those details matter when an employee asks for access. They matter just as much when a customer asks about an entitlement or a sensitive account issue. The same principle applies in incident management: provide context, but when it comes to making important decisions, leave those in the hands of humans.

The test isn’t whether a conversation ended. It’s whether the person got what they needed, were informed what would happen next, and didn’t have to repeat the request.

Give each request a way forward

For the employee waiting on software access, Computer can help find the applicable policy and start the approved process. If access needs a manager’s approval or an IT administrator’s action, the request goes to that person with the relevant details. The employee can check its status, while the existing IT system remains the system of record.

For a customer, the next step depends on the problem. An account-specific question may have an immediate answer. Another issue may need a specialist to step in. The conversation and account history should travel with the handoff so the customer can pick up where they left off.

Measure the result, not just the response

In the Enterprise-Bench benchmark, Computer was 48% more accurate than a leading AI model on identical enterprise tasks, while using 4.4x fewer tokens per correct response – demonstrating the precision and efficiency customers need from AI at work.

Both systems were tested on the same tasks and data with the same underlying model. These results speak to the quality and efficiency of the answers in the evaluation; each deployed workflow still needs its own checks for action quality, safety, and resolution.

The BILL customer story offers one example of measuring the outcome. In a proof of concept using real customer queries, BILL reported a 70% AI resolution rate. That figure reflects BILL’s evaluation. Other organizations need to define and measure resolution for their own requests and workflows.

Start by answering the questions people are already asking

Begin with a request that regularly slows people down. Identify the systems involved, the approvals required, and the exceptions a person should handle. Agree on what a completed request looks like, then test the workflow with real cases.

Customers and delivery partners know the work behind the request: which systems have to agree, where approvals stall, and which exceptions demand human judgment. Their expertise helps turn AI capability into a service that fits the customer’s environment and keeps improving after launch.

For the person asking for help, the experience should be straightforward: explain what you need once, see what happens next, and get a clear resolution.

Powering self-service through the OpenAI Marketplace

Customers can already explore solutions in the DevRev Marketplace. OpenAI’s Partner Network brings companies together to build and deliver AI solutions. As an OpenAI Marketplace launch partner, DevRev aims to give eligible enterprise customers another way to discover its employee and customer self-service offerings. The listing will let customers find DevRev in the Marketplace Directory and contract directly with DevRev. Eligible customers may also be able to apply part of an existing OpenAI commitment toward qualifying partner products, subject to the program terms.

Finding a solution is the start. Customers and partners still need to choose a request worth solving, connect the right systems, and agree when a person should step in. That work determines whether employees and customers get the resolution they came for.