Product1 publisher3 min readPublished
Accel and Google lead a $2.65 million round for Dodge AI's SAP maintenance agents
Dodge AI raised $2.65 million, led by Accel and Google's venture arm, for agents that fix SAP incidents and record the custom rules behind them. Buyers who pay integrators for that work have two unnamed customer stories, both told by the company, to judge it by.
The Product Desk · Product desk
What happened
- Integrators such as Accenture, TCS and IBM have long run this work with offshore teams of 20 to 50 people handling incidents, change requests and background jobs.
- The Next Web puts the cost of keeping these enterprise systems running at more than $600 billion a year.
- Dodge AI says it works with more than a dozen enterprises, half of them publicly listed, on incident management and process optimization across SAP, Kinaxis and Microsoft Dynamics.
- Schema Ventures, New Build Ventures, Antler and angel investors from the SAP ecosystem also joined the round.
Compiled by The Product DeskSomething wrong?How this is made
Why it matters
- exposure An integrator whose hold on an account depends on what its consultants remember loses some of that hold once the customer's custom rules are written down in a form it could hand to another partner.
- decision Buyers have to settle whether the agent only recommends fixes or also applies them. An agent that changes a system of record needs a sign-off process that the recommending version can do without.
- constraint For now, Salesforce teams would be buying mostly on the pitch. Salesforce is on Dodge's target list, but none of the customer work the company described runs on it.
A truck could not load at a warehouse because a Goods Receipt Note was printing incorrect information [9]. Dodge AI says its platform traced the fault across SAP, Kinaxis and the customer's internal warehouse software and delivered a fix within minutes [9].
The second customer story is about a slower problem. That customer had been running inventory planning overnight because SAP kept crashing when the job ran in the morning [10]. Dodge says it rebuilt the process to run 132 times faster, improved order allocation time by 8 hours and freed the team of 10 people who maintained it [11]. Neither customer is named, and the figures are the company's own. Ten people is between a fifth and a half of the 20-to-50-person offshore team in the integrator playbook [1], if those 10 were integrator staff at all.
The pitch runs well ahead of that work. Dodge describes its platform as a control plane across enterprise applications [5] and as a future "source of truth for agents operating in production" [6]. It treats maintenance as the entry point to modernization [13]. "Dodge AI captures the necessary rules and exceptions for this, while self healing your systems instead of an endless transformation project," said Aditya Thakur, Dodge AI's co-founder and chief technology officer [14]. The work the company describes is narrower and easier to check: the platform connects to ERP customizations, ITSM systems and legacy configurations to pinpoint root causes and recommend fixes [5].
The platform fields hundreds of queries every hour, according to the company [12]. That figure counts questions asked. A pilot is better judged on time-to-fix across a full incident queue, and on whether the same incidents come back, which are the repeat-maintenance patterns Dodge says its query volume lets it see [12].
The documentation half of the product goes after a different cost. According to The Next Web's account, the offshore model leaves fixes undocumented while customizations pile up, and the knowledge of how a system works ends up spread across tickets, consultants, configuration layers and memory [4]. A closed ticket is easy to mistake for a documented fix. The knowledge Dodge says it captures sits in the exceptions: why one warehouse allocates inventory differently, why one pricing rule overrides another, why a background job runs only at night [15]. "For a long time, the only way to maintain enterprise systems was to add more people," said Rebhav Bharadwaj, Dodge AI's co-founder and chief executive [7].
An operator can sort recent incidents with two tests: whether the fix crossed more than one system, and whether the reason for the rule is written down where a new consultant would find it. Single-system incidents with documented rules are what the integrator's runbook already handles, and an agent saves minutes there. Documented rules that span systems make a speed trial, scored on time-to-fix. An undocumented rule inside one system is a documentation trial: compare the agent's write-up with what the senior consultant says. Cross-system incidents with undocumented rules are the last box, and the one where a buyer leans hardest on the maintenance partner. The truck story crossed three systems [9].
I'd start a pilot in the two undocumented boxes, because a written record of the exceptions stays useful even if the agent's fixes disappoint. The tradeoff is that documentation alone does not remove anyone from an offshore contract, and the 10-person case is the only headcount figure Dodge has offered [11].
What to watch
- A named Dodge AI customer, or any Salesforce deployment, reporting before-and-after time-to-fix and repeat-incident figures.
- Dodge AI's pricing model, since per-ticket, per-system or headcount-replacement pricing would show which maintenance budget line it expects to draw from.
- How Accenture, TCS or IBM respond, whether by partnering with maintenance-agent vendors or building similar tools into their own offshore contracts.