Tines has added an interaction layer optimized for artificial intelligence (AI) agents to its workflow management platform.
The AI Interaction Layer added to the Tines Platform provides a single interface to manage AI agents, copilots and model context protocol (MCP) clients and servers in a way that makes it simpler to incorporate them into existing workflows managed by humans, says Tines CEO Eoin Hinchy.
The overall goal is to reduce the current level of friction organizations encounter when trying to incorporate AI into workflows by, for example, automatically executing a playbook that was created by an AI agent or copilot tool, he adds.
The Tines automation framework presents end users with a graphical interface for invoking an automation engine that can be applied to a range of workflows. The AI Interaction Layer extends the reach of the platform to incorporate AI agents.
IT teams can now build MCP servers directly into the Tines platform to define exactly what AI can access and how those capabilities are used. Those MCP interactions are then executed through governed workflows, producing controlled, auditable, end-to-end work that security and IT teams can fully orchestrate, says Hinchy.
That approach provides end users with a simpler means of operationalizing AI without having to necessarily master prompt and context engineering techniques. Additionally, organizations gain more control over how probabilistic outputs generated by a large language model (LLM) are infused into deterministic business workflows that need to be generally completed the same way every time, notes Hinchy. The goal is to make it easier to observe and guide the AI agent, he adds “We’re trying to accelerate AI adoption,” says Hinchy.
Organizations of all sizes have been employing a wide range of frameworks to automate tasks. The rise of generative AI provides an opportunity via a natural language interface to make automation platforms more accessible to a wider range of end users.
Each organization depending on the use case will need to determine to what degree to rely on the reasoning capabilities enabled by an LLM to automate a task versus using an automation framework. It’s not likely LLMs will eliminate the need for automation frameworks any time soon, especially when the total cost of injecting LLM reasoning into a workflow is considered. In fact, as more organizations push the limits of LLMs many of them might soon begin to view automation frameworks as a natural extension that makes it simpler to enforce governance and security policies across workflows spanning AI agents and the humans interacting with them.
In the meantime, there is likely to be a lot of continuing trial and error as organizations continue to experiment with AI technologies. The challenge, as always with any emerging set of technologies, is moving beyond a proof-of-concept that can reliably be deployed in a production environment at scale. The issue that many organizations may be overlooking is the degree to which organizations that are achieving that goal are actually relying on a range of other technologies and platforms to ensure the return on their AI investments.
