Sema4.ai’s autonomous agent building platform gets simpler to use, adds deeper business context and more

Sema4.ai Inc. a startup that provides tools for building and managing artificial intelligence agents, has announced a massive revamp of its platform, with big changes coming to every layer of the agent development stack.

The refreshed platform will impact and improve everything from the way agents are built to how they capture and understand business context, and the way they’re deployed within customer’s computing environments.

Sema4 first started getting attention last year when it raised $25 million in a Series A round of funding. It was founded by its Chief Executive Rob Bearden, who once held the same role at big data company Cloudera Inc., and offers a platform that enables non-technical workers to create AI agents with natural language prompts. Users can build the agents they need and then set them to work on tasks such as extracting information from documents. The agents can also use external software applications through a set of integrations known as “Actions.”

Users create their AI agents using an interface called the Work Room, which provides access to all of the AI features their agents need, while a second tool known as the Control Room enables administrators to monitor those agents.

With today’s updates, Sema4 says it’s trying to fix a number of issues that have caused agentic AI adoption to stall instead of scale. Things like the fragmented systems, disconnected data and tools that make more sense to software developers than Sema4’s actual target users.

For instance, Sema4 has introduced a new Agent Builder tool that eliminates the need for any technical expertise at all. Because it has more organizational context, users can simply explain using their voice or by typing, or alternatively upload a standard operating procedure they’re looking to automate. The Agent Builder will then generate a functional agent runbook that enables autonomous agents to perform that specific task.

Along with the support for voice, text and document inputs, users will also be able to access a more extensive library of pre-built agent skills, and the agents now get “persistent memory.” This means they can learn from their experiences, retain corrections and surface workflow recommendations, essentially compounding organizational knowledge over time. There’s also a new MCP Access Gallery, which enables Sema4 agents to utilize third-party software such as Snowflake, Slack, GitHub, Google Workspace and HubSpot via the Model Context Protocol. Finally, the Agent Builder gets support for federated and verified queries, allowing users to ask a single question and have their AI agents search for the answer across every database, spreadsheet and enterprise system it has access to.

Sema4 has also reimagined agent context with a new Business Context Layer that allows them to quickly understand how enterprise data is connected across multiple databases, systems and workflows. One of the key features is business ontologies, which can map the relationships between individual customers, purchase orders, invoices, shipments and so on. With this, agents can reason across the entire business instead of being restricted to a single database or system.

Meanwhile, there are new semantic layer enhancements within the Business Context Layer that help to improve agent performance as they work across data siloes.

“This release makes enterprise AI agents dramatically easier to build and deploy while giving them a much deeper understanding of how businesses actually operate,” said co-founder and Senior Vice President of Product and Customer Experience Paul Codding. “Agents should be able to reason and adapt to business concepts, not just columns and rows.”

On the architectural side, the release introduces improvements that expand the platform’s availability and simplifies agent deployment across Amazon Web Services, Google Cloud Platform, Microsoft Azure and Snowflake.

“We’re making enterprise AI agents more accessible to business users who understand the work, more connected to the systems where data lives and more trustworthy for enterprise operations,” Codding added. “Every improvement compounds: agents that are easier to build get deployed faster, agents with deeper business context deliver more accurate outcomes, and agents running on a simpler architecture reach production sooner.”

Image: Sema4