MongoDB’s Internal AI Journey with Credal

MongoDB is a leading database and data platform company. Internally, David Vainchecker’s team is responsible for MongoDB’s generative AI adoption, covering internal use cases across the company.

MongoDB’s path started like many enterprises: a basic chat agent not connected to sensitive internal sources. But quickly, the company ran into a blocker. If AI tools were going to answer real internal questions, they needed a reliable way to ensure employees only saw what they were allowed to see. That challenge led MongoDB to partner with Credal, which became not just a secure way to access internal data, but a much larger foundation for how MongoDB would build and govern all its internal AI.

MongoDB had a few practical needs:

1. Permission-aware access to internal content. Leadership needed to ensure AI tools could answer internal questions without exposing documents, policies, or knowledge to the wrong employees, using the same permission rules that already governed access to internal systems.
2. Scalability and hardening. MongoDB attempted to build permission retrieval themselves but encountered repeated challenges integrating diverse permission models, handling edge cases, and keeping access rules up to date as teams and roles changed. The ongoing security and maintenance burden made this difficult to scale in-house.
3. Self-service building blocks. MongoDB wanted a shared, approved foundation for internal AI, so individual teams could build their own assistants using common security, permissioning, and retrieval capabilities, without relying on a single central team to build or approve every use case.

Credal helped MongoDB move from “safe but limited” chat (public info only) to tools that could work with internal, sensitive content in a way leadership could trust.

MongoDB GPT: a companywide alternative to public ChatGPT

MongoDB shipped MongoDB GPT, an internal ChatGPT-like experience, because employees were already using public tools, creating a shadow IT risk. By providing an approved internal alternative, MongoDB reduced data exposure to external services and centralized AI usage under existing identity, security, and auditing controls and built features tailored to MongoDB-specific needs over time.

With a foundation in place, teams could create targeted assistants without rebuilding from scratch each time. Examples include:

‍Governance, Risk and Compliance Agent

This agent helped employees navigate MongoDB’s internal governance, risk, and compliance requirements without escalating every question to the compliance team. It was developed to answer questions like:

• Which security or compliance controls apply to a specific system or project
  
  
• How internal policies map to external frameworks (e.g., SOC 2, ISO, internal security standards).
  
  
• What approvals, documentation, or reviews are required before shipping or changing a system

Because the agent was permission-aware, it only surfaced policies, assessments, and guidance the user was authorized to see, reducing back-and-forth with compliance teams while maintaining trust and auditability.

‍Production Workflow Agent

This agent was created to support engineers working across development and production environments to help with:

• Understanding internal deployment processes, runbooks, and production standards
  
  
• Answering “How do I…” questions related to dev/prod workflows, incident response, or operational best practices
  
  
• Pointing engineers to the right internal documentation, tools, or next steps when something goes wrong

The agent made institutional knowledge easier to access, especially for engineers working outside their usual domain or ramping into new areas.

Go-To-Market Agent‍

MongoDB built Coach GTM, a technical question responder trained on internal technical documentation and related knowledge. The tool was aimed mainly at technical sales and solutions teams, but was also used by customer success, product, and engineering teams when they needed fast, authoritative answers during customer issues, escalations, or internal discussions.

Before Credal, many high-value technical questions from customers and internal teams flowed through Solutions Architects and engineering, creating delays, higher cost-to-serve, and risk in customer engagements. After deploying Credal-powered systems, most notably CoachGTM, MongoDB saw measurable improvements across cost, efficiency, and risk.

On the customer success side, CoachGTM acted as a real-time technical knowledge assistant during customer interactions, drawing from internal documentation, code repositories, and Google Drive while respecting permissions and guardrails. This produced clear, quantifiable outcomes such as $2M in annualized cost savings, based on 60,000 technical questions answered per year with an average time savings of 30 minutes per question.

In sales onboarding and productivity, Credal-powered briefings and research tools materially accelerated ramp and day-to-day efficiency, including 3–4 hours saved per week per sales rep in research and preparation time as well as a 76 percent increase in marketers using new business intelligence reports, reflecting broader, more consistent adoption of data-driven workflows.

MongoDB’s experience with Credal reflected the benefits of working with a smaller, technically strong partner. Credal’s team was responsive, closely engaged on product direction, and willing to adapt as MongoDB’s use cases expanded. As David Vainchenker noted, “that took us on that quicker trajectory of shipping more valuable applications.” The collaboration clearly accelerated progress.

Have questions on the above or want to know how these apply these use cases more specifically to your needs? We’re happy to chat. Drop us a line at sales@credal.ai!