


Back to Case Studies
How AI-Native Tekion Is Securing AI Agents at Scale with Britive

Tekion runs AI agents in production for auto dealers, where one request can pass through many agents and no person can approve each step. Tekion chose Britive’s runtime authorization and access enforcement to put AI agents, non-human identities (NHIs), and humans under one policy model, with access created for the task and removed when the work is done.
550Identities
in the first rollout: 300 human, and 250 non-human
0Standing Privileges
The target Tekion set for every identity
1Policy Model
for AI agents, NHIs, and Humans
Fast growth left access standing
Tekion serves customers in North America and the UK and runs on more than one cloud. As the company grew, access grew with it. Teza describes what fast growth does to identity.
“Identity is a boundary that cannot be defined with a single scale. You’ve got to be very, very agile about how you define it and how you provision it.”
Tekion’s team called the old way of working “brute force.” Four things defined it.
- Standing access by default. Administrator rights and long-lived cloud keys stayed in place because they kept work moving.
- An indirect path to data. To run a database query, an engineer signed in to a bastion host, a shared server in front of the database, and worked from there.
- Manual requests and reviews. People handled access requests and group mappings by hand. Quarterly access reviews were heavy because so much access was permanent.
- Slow emergency access. During one incident, a teammate waited six minutes for emergency (break-glass) access.
Then One Request Became Many Agents
By the second half of 2025, Tekion was deploying many AI agents for its customers. The access process built for people did not fit them. A developer’s request can wait for a manager’s approval. Agents act on their own and start other agents, so that step does not work for them.
“Agents work in a very different way. They are autonomous. They have the ability to spin up other agents, talk to other agents, and take action with other agents, too. So the identity has to traverse across agents. A human in that loop is just not possible given the scale of it. You’re talking about millions of these.”
With T1 Pro, one person’s question can require agents in sales, service, and parts to act and return a single answer. The primary agent starts task-based agents to do the work, and one agent often gives access to the next. Teza calls this “this whole inheritance of permissions” and says the security team needs an “absolute view into what is going to happen.”
Auditors noticed too. Teza says one of the most common questions auditors ask today is, “how do you know what the agent has done?”
Fixed rules did not solve it. “You could put in hard rules,” Teza says, “but you actually restrict what the agent can actually perform.” Tekion needed access that is decided for each task, that expires, and that leaves a full record, for every kind of identity in one place. Teza’s term for that place is “a single funnel.”
The Solution: Zero Standing Privileges as the Target
Tekion set zero standing privileges as the goal for every identity, which means no identity keeps privileges between tasks. Britive is built on that principle. Britive authorizes each request at runtime, creates the privilege inside the target system only when an authorized action requires it, and removes the privilege when the work is done.
Tested Alongside Production Before Any Cutover
Tekion’s proof of concept had to show four things.
1. Tekion could reach zero standing privileges without breaking production services.
2. Britive-managed access could run in parallel with existing access before any cutover.
3. Engineers could get access to AWS, databases, and Kubernetes on request, with sessions recorded and access expiring automatically.
4. Engineers would use the new workflow and would have no reason to work around it.
One Runtime Control Plane Under the Tools Tekion Already Had
Tekion kept its identity provider and its access request tool. Britive sits underneath them and creates, scopes, and removes access in each cloud. Requests, approvals, and alerts arrive in the chat, ticketing, and monitoring tools Tekion’s teams already use. Britive requires no endpoint software.
“I liked Britive’s approach,” Teza says. “They gave us one single view into how we provide identities the access to perform in our environments.”
Purchased through AWS Marketplace
Tekion purchased Britive through AWS Marketplace and is rolling it out to 300 human and 250 non-human identities, starting with AWS and Azure.
Eliminating Standing Privileges at Scale
The rollout changes how access works for each group at Tekion.
One place to see every identity. Tekion’s security team sees what each human and non-human identity can do, and what each one did, in one console.
Access that expires. An engineer or a service receives access for a task and loses it when the task ends, so standing access stops building up between quarterly reviews.
A direct, recorded path to data. Engineers connect to databases and terminals without the bastion host, and Britive records the session.
Faster emergency access. Engineers request emergency access in the tools they already use.
Evidence by Architecture: Because access is created at runtime and removed when the session closes, the audit trail stays current by design. Every session is built on dynamic checkouts, so granular audit records are generated automatically at runtime, giving compliance teams clean verification data directly alongside native cloud infrastructure logs.
An answer for auditors. Britive records who requested access, who approved it, what Britive granted, and when Britive removed it.
“Whether it’s a SaaS platform that we’re using, a human identity, a compute resource, or an agent identity, bringing all of this into a single funnel gives my team the right kind of visibility into what is happening across our identity stack. That’s where we win with Britive in place.”
Next: Bringing AI Agents Under the Same Policies
Tekion’s team is extending the same policies to its AI agents. Britive’s platform creates and removes privilege for each task, and Britive ARC™ (Agentic Runtime Control) extends that model to AI agents. An agent receives access for one task, Britive removes the access when the task ends, and the audit record ties each action to the identity that took it. That matches what Teza says an agent’s task has to be: short-lived, limited to the task, and traceable.
A Partner that Already Knew Automotive
Britive worked with global automakers, including Toyota, before Tekion’s evaluation began.
“Auto has its own complexities, and that was a huge validation when we were working with [Britive.]”
