Accel and Google back Dodge AI’s ambitions to automate enterprise software maintenance with agents

Dodge AI has raised $2.65 million in seed funding to push a simple but ambitious pitch: enterprise software upkeep should not require a small army of consultants forever. The round was led by Accel and the Google AI Futures Fund, with support from Schema Ventures, New Build Ventures, Antler and angels from SAP’s ecosystem.

The startup is going after what it calls a $600 billion “enterprise firefighting” market. That is the messy, expensive layer of work that sits on top of systems like SAP, Salesforce and Microsoft Dynamics after companies have customized them for years. Install the software and the real work starts. Every deployment bends into its own shape, filled with custom rules, local workarounds and fixes that may never have been properly documented.

That complexity is why many companies lean on systems integrators such as Accenture or IBM, with teams of 20 to 50 people assigned to a single deployment. Dodge AI co-founder and chief executive Rebhav Bharadwaj isn’t saying that model failed. He’s saying it became the default because people were the only tool available. His argument is that AI agents can now do the same maintenance work faster, while also recording the logic that used to live in consultants’ heads.

The startup’s system is built around a control plane that sits across enterprise software and uses agents to analyze configurations, trace failures back to root causes, recommend fixes and carry them out. Dodge AI says its software also creates an “exception intelligence context graph,” tying together business processes, ERP settings and legacy systems so the company can see why one warehouse behaves differently from another, or why one pricing rule overrides the rest.

Dodge AI says several large enterprises have already shifted work from human consultants to its agents, though it isn’t naming them. The company says it is handling hundreds of queries per hour and resolving problems in minutes that would otherwise take hours or days. It cites one case where a warehouse loading failure was traced to an incorrect Goods Receipt Note across multiple systems, and another where a SAP inventory planning process was made 132 times faster. The longer play is bigger still: once the custom logic is understood, Bharadwaj believes, digital transformation gets a lot less scary.