About CarNow
CarNow is a mid-sized automotive SaaS company providing digital retailing and customer-engagement software to more than 4,000 dealerships. Its engineering team runs its day-to-day work through GitHub, Jira, Confluence, Datadog, Slack, and Microsoft 365.
The problem
Agentic AI adoption across CarNow’s engineering team was stalled. Some engineers didn’t see the value yet and hadn’t tried. Others had tried and gotten stuck. To connect a coding agent to a real system, an engineer needed to configure an MCP server for that app. That required admin permissions in the source system (ex. Slack, Atlassian), which most engineers lacked. For these engineers, the barrier wasn’t a lack of trying. Their attempts to set up their own MCP servers were blocked because the access they needed lived in someone else’s admin console. Multiplied across every app they wanted to connect, that meant a new permissions request and a new configuration effort for each one, with no guarantee any of it would be approved.
The Solution
To address this problem, CarNow consolidated access into one governed point of entry. Its platform team stood up a single Dtwo Gateway and wired in every MCP server the engineering team needed: GitHub, Jira, Confluence, Datadog, Slack, and Microsoft 365. Instead of each engineer needing source-app access to configure their own MCP servers one integration at a time, the platform team configured that access once, centrally, and exposed all of it through the gateway. Setup took five to ten minutes: connect, authenticate, done. The sanctioned path became the easy path, so it’s the one engineers took.
Dtwo did more than unlock adoption. Because every agent action now ran through one governed path instead of a dozen individually configured MCP servers, CarNow’s platform team gained something it didn’t have before: a single point where agent behavior could be observed and controlled.
The Impact
With the tools they needed available in one place, engineers moved quickly. Within days, teams went from little or no automation to workflows driven largely by agents.
The results were immediate:
- Reduced the overhead of creating Confluence documentation from days to minutes by automatically gathering context from across CarNow’s tools.
- Streamlined Jira workflows so developers no longer need to create or manage tickets manually.
- Accelerated code reviews by drafting pull requests, sharing them in Slack, and incorporating feedback automatically.
- Improved post-deployment reliability by monitoring Datadog and identifying—and sometimes resolving—issues before developers need to investigate.
Engineers can now spend more time designing and writing software instead of managing administrative work. That reclaimed time has helped them ship multiple pull requests per day.
The Next Phase
CarNow’s engineers are already asking for additional MCP connections, like Miro for architecture diagrams and AWS log access for debugging deployments beyond what Datadog captures.
On the governance side, CarNow’s platform team is taking an observe-before-enforce approach: using Dtwo to gain visibility into how agents are actually being used before writing the policies that will govern them. That gives them the evidence to write guardrails that fit real behavior, not just theoretical risk.
The Takeaway
The lesson from CarNow is simple: make the sanctioned path the easy path. Once engineers had easy, governed access to the tools they needed, adoption followed on its own, and the productivity gains came right behind it. If CarNow’s results sound like what your team needs, start a free trial at dtwo.ai.