A new AI agent and no-code AI Studio, built on 12 years of frontline data, are launching in the Poka early access program to bring managed industrial AI to the worksite of the world’s largest industrial operators.
Pokathe leading industrial AI platform for connected work, is setting the bar even higher with the announcement of its Industrial AI Roadmap. No-code AI Studio and a Model Context Protocol (MCP) server are added to managed AI agents already in operation for designated industrial customers. MCP Server gives any AI assistant access to Poka functionality, and code-free AI Studio lets customers build their own agents to solve and manage critical shop floor processes. Both are currently available to selected customers in an invite-only early access program and will be available to the public in October 2026.
The world’s largest industrial operators are increasingly choosing Poka as their connected work solution. New customers acquired in 2026 include Viega, First Solar and The Coca-Cola Company, as well as expansions across existing locations. Poka has expanded across Europe with new growth in DACH, the UK and Ireland, and France. The company’s ARR grew significantly by 63% in 2025, followed by a 60% increase in the first quarter of 2026.
“Industrial AI is not enterprise AI in a helmet,” said Antoine Bisson, co-founder and CEO of Poka. “This belongs in work. We’ve spent 12 years building the foundation on which it works. Now we’re defining industrial AI for connected work.”
This is the company’s first major strategic move since Antoine Bisson took over as CEO earlier this year. He previously served as Chief AI Officer at Poka, where he led the creation of the data foundation built into the DNA of the new agent. Poka’s 12 years of real-time, frontline workflows across thousands of sites and dozens of languages is a foundation that no AI-first entrant can match.
Also read: AiThority interview with Matej Bukovinski, Chief Technology Officer, Nutrient
The roadmap addresses three forces converging on the manufacturing floor. Experienced operators are retiring faster than new operators can be trained, robbing them of decades of plant knowledge. Regulators are tightening rules around audit trails and deterministic behavior in the environments in which industrial AI currently operates, and the first wave of horizontal AI on the factory floor has had limited operational impact.
Poka’s answer is a complete knowledge flywheel. Consists of AI agents. AI agents take what the best operators know and put this industry expertise into the hands of every operator, shift by shift, managed in a regulated environment, and based on real-world workflow data. Rather than forcing a horizontal AI fixed to the floor, deliver managed industrial AI with tribal knowledge that equips your employees.
Industrial AI in production
Poka Industrial AI agents are currently in operation with selected industrial customers, covering key use cases such as:
- A problem-solving agent that guides operators to a solution
- Analytics agent that analyzes trends across operations and reveals insights in seconds
- A knowledge agent that searches across work instructions, troubleshooting guides, and important documents to provide clear, actionable answers.
Poka announces early access program
Now, Poka is also launching an invite-only early access program for Poka AI Studio, a no-code environment where operations teams can build their own agents on the same foundation or use templates that cover use cases such as:
- Work order comparison agent that compares two steps and accurately highlights differences
- A formless issue agent that handles issues end-to-end in plain language (create, update, resolve tasks, assign tasks, delete comments, etc. without opening the issue itself)
- Similar Issues Agent searches open, resolved, and archived issues to uncover currently relevant issues and reduce problem resolution time.
These AI agents will be generally available in October 2026.
Agents do more than just answer questions. Through Poka’s extensibility framework, trigger workflows in connected systems like MES, CMMS, ERP, etc. and execute steps with human approval. All Poka agents are managed, end-to-end auditable, and can be deployed by operations teams without any coding effort.
Today, Poka is also shipping production-ready MCP servers to enable enterprise AI assistants to access Poka data and workflows. Industry organizations building their enterprise AI stacks with Copilot, ChatGPT, Claude, or their own solutions can connect directly to Poka without custom integration. An employee who asks “Show me the third open issue from last week” within an AI tool provided by the company will receive a live response from Poka, with authentication, permission scoping, and governance performed on every call.
The early access program will begin as a founding cohort. This will be limited to select manufacturers who will deploy with Poka engineers along with their teams to capture baselines and outcome metrics from day one, shape the roadmap for Poka AI Studio, and transform based on founding members’ terms upon general availability. Manufacturers can apply at poka.io/ai.
“On the factory floor, a wrong answer from AI could hurt someone and get regulators involved,” Bisson said. “Industrial AI is what works there and drives the numbers on the board pack. Across all plants, on the fly. No configuration sessions required. No compromises. This is the AI we built.”
Also read: AI Systems – Interoperable AI Systems: Connecting models across platforms
