AI agents are gaining traction, but building them to be useful, governable and grounded in real business data remains a challenge. That's why KNIME, the open source data analytics and AI company, recently demonstrated the software's AI agent building capabilities at their annual Spring Summit event in Berlin last month. CEO Michael Berthold shared a vision for how organizations can leverage their existing infrastructure and the KNIME platform to create scalable, intelligent, and continuously evolving AI agents.
Agent design typically comes in two flavors either through blackbox prompt chains or through code-heavy environments accessible only to deep experts KNIME offers users something in between. The visual workflow-based software is as transparent and flexible as a coding language, but intuitive allowing for wider accessibility and easier collaboration. It's also modular, allowing for faster time to value and easy reuse.
What's more, organizations have worked tirelessly to figure out how to make the most of the large quantities of structured and unstructured data they have been storing. Enterprises build up data practices and upskill data science citizens but both approaches still only scratch the surface of making the most of the troves of proprietary data every organization sits upon. Agents that are built with KNIME can integrate the know-how of data experts and the context of domain experts allowing the organization to build more agentic data workers that surface information or perform actions based on the needs of the organization.
KNIME Software, historically, has been used for data access and integration, building analytical and predictive AI models, and automating data processes. The same intuitive, modular design of the platform allows existing users to turn their workflows into tools that can then be accessible by an agent (also built via KNIME workflow). For example, a customer churn model or a report generator, can now act as callable tools inside an AI agent's decision flow. KNIME Software also easily supports standards, like the recent Model Context Protocol (MSP) an open standard for connecting AI agents and their tools.
At the event, Michael Berthold defined agents as orchestrators systems that dynamically select and sequence tools (some powered by GenAI, some not) to solve problems in context. These tools include everything from predictive models and LLM-powered components to basic data aggregation workflows most of which many data teams are already creating. For KNIME's existing userbase of nearly half a million users, turning their existing repository of workflows into "tools" for an agent is straightforward and opens the door for agents to make use of their large collection of existing workflows.
Live demo'ing these advanced capabilities, the KNIME team showcased two internal agents. One was an internal 'Ask Me Anything' agent a chat-based AI assistant that pulls from customer, community, and employee data across multiple systems like Salesforce, Zendesk, and HubSpot. Another focused on enforcing communication consistency across public-facing content. KNIME's "Style Checker" agent scans blog posts and marketing material for tonality and terminology. It can even update its own guidelines based on human feedback via email, allowing it to learn and evolve with each interaction.
"This demonstration evidenced that building agents doesn't mean a tradeoff between transparency and complexity," says Michael Berthold. "KNIME gives enterprises a visual, modular, and governable way to build intelligent agents that they can trust and scale."
KNIME's agentic AI approach embraces flexibility. Agents can adapt broader scopes and new capabilities. Tools can live anywhere and be customized. Prompts can be edited and specified. Users (from data engineers to business analysts) can collaborate on intelligent systems without reinventing the wheel. Looking ahead, as agentic systems evolve, KNIME's infrastructure will continue to make it possible to combine flexibility with governance rooted with good data.
Ready to build your first AI agent? Get started at https://www.knime.com/.
About KNIME:
KNIME makes working with data and AI accessible, intuitive, and transparent for everyone.
The free and open-source KNIME Analytics Platform is a visual workflow tool that allows users from all backgrounds-business, technical, or data-focused-to explore, analyze, and design data solutions of any complexity from spreadsheet automation to advanced machine learning to AI-augmented analytics work with intuitive visual workflows.
KNIME's commercial products include the KNIME Team Plan, a cloud-based SaaS offering that supports teams who want to collaborate on and automate data work in the cloud. And KNIME Business Hub, the enterprise software that offers advanced capabilities for automation, collaboration, and data and model governance allowing organizations to drive up AI innovation while driving down any associated risk and ensuring scalable and responsible management of data and AI projects across organizations. KNIME, the company, has over 400 customers, including Audi, AMD, Lilly, Novartis, Bayer, Sanofi, Genentech, the FDA, P&G and Mercedes-Benz and nearly half a million open source users.
KNIME supports the complete data science lifecycle, allowing teams at all levels of analytics readiness to build a scalable data science and AI practice.
View source version on businesswire.com: https://www.businesswire.com/news/home/20250430335835/en/
Contacts:
For media inquiries, please contact:
Marissa Pasillas
Walker Sands for KNIME
knime@walkersands.com
