 |
VOOZH | about |
|
VOOZH | about |
We’re so glad you’re here. You can expect all the best TNS content to arrive Monday through Friday to keep you on top of the news and at the top of your game.
Check your inbox for a confirmation email where you can adjust your preferences and even join additional groups.
Follow TNS on your favorite social media networks.
Become a TNS follower on LinkedIn.
Check out the latest featured and trending stories while you wait for your first TNS newsletter.
OutSystems was one of the early players in the low-code/no-code space. Like many of its competitors, including Microsoft’s Power Apps or Mendix, it’s betting that AI and AI agents will give the entire space a new boost. With Mentor, OutSystems has long offered an AI-powered tool to quickly get basic internal and customer-facing apps off the ground, and today, OutSystems is launching Agent Workbench, its service for building and deploying AI agents, into general availability.
Virtually every company is now trying to figure out how to best use agents to increase productivity by automating workflows, no matter whether that’s about business processes or managing the software development lifecycle. But it’s no secret that many businesses also struggle to take their AI pilots from flashy demos to production.
The promise of Agent Workbench is that it will give developers an end-to-end low-code platform to build, test and deploy agents securely, with unified data access across a business’s data tools to ground the agents for improved results.
To access this data, the service uses Outsystem’s existing Data Fabric to pull in data from virtually all standard databases and data lakes, SaaS applications like Salesforce, and SAP enterprise services.
“When you start a new agent on our platform, historical data is already pre-prepared for you,” OutSystems co-founder and AI product manager Rodrigo Coutinho told me when I asked him about the Data Fabric. “And even though it’s all customizable, because in low-code you can change and customize whatever you want, just this step of having these things prepared, it’s a big jump in terms of productivity.”
Often, the reason why these AI pivots fail isn’t technical — though there are some hurdles there, too — but about trust and compliance. In these non-deterministic systems, you need the right tools to ensure that the agents can only access the data they are supposed to have access to, after all. OutSystems has long featured automated governance enforcement. The core security features on the platform are automatically enforced and built into it.
In his keynote at the company’s annual ONE conference, OutSystems CEO Woodson Martin today stressed this end-to-end nature of the service. “You’re slowing down because of all the friction that’s inherent in dealing with multiple technology platforms, multiple vendors,” he said. “I can’t blame you at all, because you have to worry about multiple agentic development frameworks. You have to worry about orchestrators, different connections to data. Don’t even get me started about different UI toolkits.”
He also noted that a lot of the existing tools were built for agents that work completely autonomously, but in the enterprise, “you want to put the human in the loop,” he said. That’s why Agent Workbench offers a number of features that ensure that the agent can pass information back to the human for validation.
Agent Workbench allows developers to pick which large language model they would like to use, with all of the usual suspects, including models from Anthropic, Cohere, Google, IBM, Mistral and OpenAI, available to them. There is also support for custom-built models from Google’s VertexAI and open-source models on HuggingFace, as well as support for AWS’s fully-managed Bedrock LLM platform.
Other new features in Agent Workbench, now that it is generally available, include support for the Model Context Protocol (MCP) and an agent marketplace.
