THEIR CHALLENGE
Fragmented data and sales friction
For MongoDB's go-to-market (GTM) teams, preparing for a customer conversation has never been a simple research task. It’s meant piecing together information from Salesforce, meeting transcripts, emails, and documents, then translating that information into account plans, discovery questions, and follow-up work.
The information existed, but sellers had to connect it themselves. Moving between applications meant repeatedly gathering context, and general-purpose AI tools often produced answers that lacked an understanding of MongoDB's business and sales methodology, which led to manual vetting. In fact, MongoDB account executives spent an average of 4 to 6 hours on administrative work, per account/deal cycle. Reducing that time, and allowing GTM team members to focus their time and efforts on customer happiness, could be broadly impactful.
But to do so effectively, MongoDB needed a way to bring relevant customer context into the work itself, rather than leaving each seller to bridge the gaps between systems.
OUR SOLUTION
Building a system of action with Atlas Agent Engine
MongoDB used Atlas Agent Engine to build Holly, an internal AI assistant that connects account intelligence with everyday sales workflows.
For meeting preparation, Holly brings together relevant information to create tailored briefs and surface unanswered questions. After conversations, it helps sellers capture discovery insights from transcripts and maintain Salesforce records. Account and opportunity briefs support ongoing planning, while research and correspondence tools help teams prepare targeted outreach.
Figure 1. Holly’s agent architecture.
These capabilities address related stages of a seller's work: understanding the customer, identifying what still needs to be learned, and recording information that supports the next interaction. Holly grounds that work in MongoDB-specific business context rather than treating each request as a generic research question.
