OutSystems launches Agent Workbench for companies seeking to maximize the value of their digital processes with AI agents. The generally available version of Agent Workbench is more comprehensive than the previously announced test version, including support for a broader range of LLMs and an agent marketplace.
Agent Workbench enables organizations to build AI agents that can independently perform tasks across various business processes. While many companies are stuck in AI pilots without concrete results, OutSystems offers a platform that can be used immediately with existing workflows. According to the company, this results in AI agents whose business impact can be quickly observed by users.
Results from early access program
The results from the early access program show what is possible. Axos Bank developed an intelligent agent that analyzes error logs and generates practical recommendations. This saves the team considerable time in resolving technical issues.
Thermo Fisher Scientific deployed a Customer Escalation Agent that interprets unstructured customer data. This eliminates the need for human intervention during initial contact with customers.
“We are going to further expand the use of AI capabilities within OutSystems to achieve immediate results without having to invest in specialized AI roles,” said Kevin Hearn of Axos Bank during the presentation.
Expansion of support
The number of supported LLMs is expanding in the generally available Agent Workbench. In addition to AWS Bedrock, Azure OpenAI, and Anthropic models, Agent Workbench now also works with Gemini, Cohere, Mistral, Databricks, and IBM WatsonX.
The platform also offers support for custom models via VertexAI and open-source alternatives via HuggingFace. This flexibility enables organizations to develop their own AI strategy without being locked into a vendor.
Prebuilt agents
An essential addition is the agent marketplace. Here, organizations can find prebuilt agents for common business processes. For example, TeamWork developed a multi-agent system that automatically handles routine tickets, allowing support teams to focus on more complex issues.
The marketplace also supports the Model Context Protocol (MCP), giving agents direct access to business systems and external tools. This significantly speeds up implementation.
Next step in AI
The launch of Agent Workbench is the latest step in OutSystems’ AI strategy. The company previously released AI Builder, a tool for developing AI applications within the low-code platform.
OutSystems Mentor has already made it possible to generate complete applications using natural language. Agent Workbench complements this by explicitly addressing the need for independently operating AI assistants. The platform offers control through policy, traceability, and compliance throughout the entire lifecycle of AI agents. For organizations that want to introduce agentic AI without the risks of uncontrolled AI experiments.
Read also: Time for the backlog: How are low-code and AI changing software development?
Brand-new CEO
Paulo Rosado, founder of OutSystems, led his company for 24 years through developments in the cloud, DevOps, low-code, and AI, among other things. The latter development is still in full swing with a move towards agentic AI. Rosado has decided to step down to enjoy his retirement, marking the first CEO change in OutSystems’ history.
His successor, Woodson Martin, has a background at Salesforce, where he worked for 18 years and was responsible for scaling AppExchange, among other things. This was a significant expansion to the world’s largest enterprise cloud marketplace. According to OutSystems, Martin is the right person to steer the company through the AI era. In the coming days, we will discover during the ONE Conference in Lisbon what course Martin wants to take with the agentic AI developments.