How RapidCanvas automates 70% of data tasks for gen AI projects

Ever since ChatGPT emerged, enterprises have been all about AI and how it can help them address critical business challenges. It all started with large language model (LLM)-powered chatbots and search tools, which allow users to find answers and insights quickly. But the trend has now shifted to compound AI agents — systems capable of performing multi-step reasoning and handling tasks like support ticket management, responding to emails and making reservations. 

Salesforce triggered the wave of AI agents with the announcement of AgentForce a few months back. Now, these systems are moving deeper into the enterprise stack. Case in point: RapidCanvas, a Texas-based startup, claims its context-aware AI agents can automate 70% of data tasks during custom AI deployment.

The company has raised $16 million in series A capital to further accelerate the expansion of its agent-based platform. In fact, enterprises such as PayPal, Suzlon and MTE Thomson are already using it across their workflows, accelerating the time to value by ten-fold and reducing the implementation costs by as much as 80%.

RapidCanvas AI agents address AI deployment bottlenecks

When executing an AI project, organizations are often bogged down by tech the talent shortage (due to high demand). Even if they manage to hire skilled engineers or external consultants, those teams have to spend a lot of time on coding and data science tasks — from integrating data assets, to preparing, transforming and modeling them, to productionize downstream use cases. This stretches out implementation by several months, affecting ROI and business growth. 

To fix this, former PayPal executives Rahul Pangam and Uttam Phalnikar — who handled risk strategy and architecture — teamed up to launch RapidCanvas. 

“Our goal with RapidCanvas is to revolutionize how businesses build reliable, customizable AI solutions without requiring teams of technical experts; our platform empowers business and operations teams by using a hybrid approach combining AI Agents and an expert in the loop,” Pangam told VentureBeat.

At the core, RapidCanvas’ platform provides enterprises with content-aware AI agents that can be prompted in natural language to handle several data engineering and science tasks, from data ingestion, orchestration and preparation to enabling analytics, applications, pipelines, automation and modeling. 

According to Pangam, agents execute these tasks on behalf of users by enriching their prompts with contextual information gathered directly (business terminologies fed by users) as well as from connected systems (CRMs, data platforms, support ticket systems). It also takes into account the problem the user is trying to solve, as well as context gathered from previous projects to ensure the task is optimally run.

This, Pangam says, enables enterprises to handle up to 70% of data tasks faster and more cost-effectively than humans. And, they can use the prepped data in combination with a visual canvas to deploy the application in question.

But here’s the catch. While the offering reduces the dependency on technical talent, such as data engineers, it does not eliminate their need. The remaining 30% of the job in the workflow — covering aspects like system design, hypothesis testing and problem-solving — goes to human experts. Pangam says a company that may have previously employed 10 expert engineers would just need one or two when using RapidCanvas agents to build AI projects.

Taking on DataRobot, Dataiku

RapidCanvas is taking on the likes of leading players like DataRobot, Dataiku, Palantir and Alteryx. However, the company says its human-agent hybrid approach is a key differentiating element.

“In any of the legacy data science machine learning vendors, the primary way for non-coders to build end-end AI solutions is using no-code templates,” Pangam explained. “For example, if I want to merge two datasets, I have to pick the ‘join’ template from the UI, add datasets, join conditions to instruct which columns to match for index, set the join type and then define output columns. On the other hand, with RapidCanvas, the user instructs the agent to merge two specific datasets and it auto-generates the code to merge them. This is because the agent already has the prior context of the type of tables, index and schema, size, join types, data types, etc.”

Further, the CEO noted that the company offers a human expert as part of its subscription. This individual works as an advisor, helping teams at key decision points with ideas, as well as support for performing complex operations, verifying outcomes and understanding industry best practices. Users can go either for this human expert-backed plan or a self-service platform-only offering at a fixed monthly fee per user.

Several enterprises, including Fortune 200 companies, across manufacturing, retail, infrastructure and financial services domains have already begun adopting RapidCanvas for their AI development pipelines. The company counts among its early customers PayPal, SFR, Suzlon, AutoFi and MTE Thomson.

Looking ahead, the company plans to grow its customer base and further enhance its AI agents to ensure they can work together to automate and simplify complex workflows in a multi-agent, human-in-the-loop-backed setup.