Kiwi powers autonomous enterprise with MongoDB

THE CHALLENGE

Unstructured data and operational drag

According to Kiwi AI Co-founder and CEO Abhishek Jha, in a typical modern enterprise, knowledge workers lose 40-60% of their time to “human API work”. These are the repetitive tasks that waste human potential, and can include manually reconciling data, switching tabs, and stitching together disparate systems.

According to Jha: “Busy work is unstructured, and so is the data behind it.”

To address this, Kiwi wanted to build a platform that understands context and acts across systems to eliminate operational drag. This involves handling messy real-world data—emails, documents, logs, and events—while maintaining the ability to move fast and scale.

OUR SOLUTION

A flexible backbone for AI agents

Kiwi selected MongoDB as its backbone database to power its vision of the autonomous enterprise. MongoDB’s flexibility made it the natural choice, as it maps cleanly to application objects and offers excellent developer ergonomics.

“We first came across MongoDB while evaluating database options, and its flexibility stood out immediately,” said Jha. “It maps cleanly to application objects, has great developer ergonomics, and gives us speed without compromising scale.”

MongoDB now allows Kiwi to store raw text, structured JSON, and AI-generated metadata side by side. This flexibility is essential for automating cross-system work, allowing the storage of UI layer data and structured logs in a single, unified execution trail.

In addition, MongoDB’s change streams allow applications to subscribe to real-time data changes within the database, giving Kiwi’s agents the power to react the moment data is updated or modified. This empowers real-time anomaly detection, seamlessly flagging errors, mismatched contracts, or revenue leakage.

With MongoDB acting as its operational core, mapping entities, customers, and contracts across siloed systems, Kiwi has created an “enterprise context graph” that allows AI agents to reason, understand context, and execute autonomously.

OUTCOME

Startup agility with enterprise scale

By leveraging MongoDB Atlas, Kiwi has been able to shift its focus from infrastructure to innovation. MongoDB is now Kiwi’s primary datastore for agent configurations, workflow state, conversation history, and raw I/O processes, while Atlas Search powers instant lookups of agents and workflows by keywords, metadata, and tags.

The implementation has enabled three critical outcomes:

• Dynamic routing and rapid prompt iteration: MongoDB allows Kiwi to avoid schema bottlenecks and iterate quickly without complex migrations. As Kiwi’s requirements evolved, the document model allowed the business to ship fast.
• Production reliability: The platform handles workload spikes seamlessly as agents orchestrate thousands of tasks, offering what Jha describes as “startup agility with production reliability”.
• Accelerated development: Through MongoDB for Startups, Kiwi received engineering guidance on data modeling for real-time automations, helping it to move quickly from prototype to production.

“MongoDB is the backbone database for companies working with unstructured real-world data who need to move fast and scale,” Jha concluded.