Relevance AI Agents Deliver 3,000% ROI for Clients with MongoDB

AI is set to define our working lives for decades to come. Businesses across the world see its potential and are looking at how best to embrace it. However, few are using it to reach the staggering levels of growth—both internally and for its clients—that Relevance AI is achieving.

Based in Sydney, Australia, Relevance AI provides a no-code platform that allows its clients to build customized AI workforces on demand. Customers use large language model (LLM)-based automated processes to define and build AI agents quickly and effectively. These AI agents handle repetitive, low-value workloads in areas ranging from customer service to sales, business development and lead generation, and conversion.

“We’re a team of 30 people looking to build the automations that we truly believe are set to power next-generation companies,” explained Relevance AI Co-Founder Jacky Koh. “We want companies to only be limited by the size of their ideas, not the size of their teams.”

Founded in 2020, Relevance AI’s users span an impressive roster of clients, from dynamic tech startups to Fortune 500 giants. However, all have one thing in common.

“These companies all have strong—and growing—demand,” said Koh. “There’s this old equation that scaling a business means growing your team. Our job is to decouple that, which is why these growing companies are such a sweet spot for us.”

Central to Relevance AI’s operation is its database. The business launched on a proprietary cloud-based NoSQL platform, however quickly realized that a lack of vector search capabilities was a severe limitation. Relevance AI needed a more comprehensive solution that would meet its requirements for handling large and growing volumes of unstructured data and corresponding vectors quickly, reliably and effectively.

“We initially looked at developing our own, but then we saw that MongoDB Atlas Vector Search had developed that capability,” Koh said. “We already knew how seamless MongoDB’s cloud capabilities are, and its auto-scaling and observability benefits, so it was an easy decision.”

Relevance AI’s decision to switch to MongoDB Atlas was underpinned by a variety of key factors. With a global client base that stretches from Australia to the US, Relevance AI operates separate regional clusters—both client-specific and multitenant—in Azure and AWS.

“That cross-regional capability was also an important selling point for MongoDB,” Koh said. “We spin up servers for every region in which we operate: Europe, Asia-Australia and the US. With MongoDB, managing clusters for each region is seamless.”

While Relevance AI chose MongoDB primarily for its qualities as a database, other features stood out that proved popular with the business’s staff.

“A lot of our team come from a full-stack engineering background, so MongoDB is definitely a top choice for them when it comes to developing applications,” explained Koh. “The developer experience is excellent; being able to drop in lots of unstructured data with no negative effects in areas like latency or cost is a very competitive proposition.”

MongoDB’s maturity as a cloud-based platform was also highly attractive to Koh and his team. “MongoDB offers good observability and high-performance query profilers,” he added. “The user interface to manage indexes was also a top selling point, adding to the vector search and scalability features.”

The vector capabilities are particularly important to Relevance AI, giving teams the flexibility to develop a series of different client applications and LLM-based use cases.

“Just as an experiment we tried using a separate vector database and MongoDB as an operational database. But the separate vector database fell short on multiple different benchmarks,” Koh said.

“Having both the operational and vector database in the same platform makes managing scale much simpler and managing performance much more effective. It also allows us to just use it more—instead of just very specific cases, we can use it whenever we need it. We also don’t have to learn another query language, which is a big help.”

MongoDB is now a critical component of Relevance AI’s platform. It enables the business to build out its offering horizontally and manage extraordinary levels of growth without missing a beat.

“When we started, we were processing around 200,000 tokens a day,” said Koh. “Now that figure is close to a billion, and we’re looking at growth of between 2,000% and 5,000% in the coming year. That wouldn’t be possible without the flexibility of MongoDB.”

The use cases extend beyond Relevance AI’s client offering. The business uses the platform to manage its own operations ranging from content generation to qualifying potential sales leads.

“Some of the most exciting features are actually the simplest,” said Scott Henderson, Growth, Product and Marketing Lead at Relevance AI. “ We can take a lot of the repetitive tasks that have thus far been impossible to automate and just make everyone’s life a bit easier as well as, of course, increasing their output. That’s pretty cool.”

While Relevance AI’s internal successes are impressive, Koh is keen to focus on the successes the company enables for its clients.

“A typical AI agent that’s running well can deliver an ROI of around 3,000% for the customer,” said Koh. “These are the figures we like to talk about. Are we delivering value for our customers? Are they growing as much as we want them to? That’s ultimately what we care about, and our success as a company is very much tied to that.”

The company’s customer-first philosophy is set to deliver yet more growth and to play a key role in broadening Relevance AI’s scope and popularity even further to new users, verticals, and use cases.

“We’re building our levels of maturity in different business functions – going beyond sales, marketing and customer success into other industries and business areas,” Koh said.

Jacky Koh, Co-founder, Relevance AI

Jacky Koh, Co-founder, Relevance AI