Most factories have spent the past decade learning to collect data from their machines. The harder part has been doing something with it fast enough to matter on the shop floor. Flexxbotics, a US industrial software company, is betting that the next step is letting software act on that data directly, within limits that production teams set themselves.
At IMTS 2026 in Chicago, Flexxbotics showed an updated release of its platform for what it calls manufacturing autonomy. The headline addition is a real-time control plane that lets manufacturers decide when the software can respond to a problem on its own, which automatic responses it is allowed to carry out, and when a person has to approve a step before anything changes.
The company is pairing the release with some bold customer numbers. It says one customer, PMI, increased capacity by 125%, and that another, EIS, cut downtime by 89%. Those figures come from Flexxbotics and its customers. The company has not published the baselines, timeframes or conditions behind them, so they are best read as examples of what individual sites have reported, not as typical results.
From Collecting Data To Acting On It
The idea behind the platform is straightforward. Factory equipment, from CNC machines to robots and inspection systems, produces a steady stream of signals. Experienced technicians know how to read those signals and when to step in: adjust an offset, change a tool, stop a line before it produces scrap. The problem is that those experts are scarce, work one shift at a time, and cannot watch every machine in every plant.
Flexxbotics aims to capture those responses as software rules. When the platform detects that a process is drifting, it can determine a correction and, if permitted, send it back to the equipment through the same connections it uses to read data.
"We help with the correction of that deviation, and then we administer that correction itself through the same digital pipelines we've created to capture information," co-founder and chief executive Tyler Bouchard told Automation World at the show.
Governed Actions, Not A Free Hand
The central feature of the new release is FlexxControl, the real-time control plane. It governs how the software responds across machines, production lines and plants. Production teams define the boundaries: which actions the system may carry out automatically, which ones need escalation, and where a human must sign off.
FlexxControl also manages approvals and escalations and logs each decision, so there is a record of what changed, why and who authorised it. That traceability matters in regulated industries such as aerospace and medical devices, where changes to a process often have to be documented and justified.
This governance layer is what separates the pitch from simple automation. Few manufacturers are willing to let software change a running process unchecked, and the approach of letting teams widen the software's authority step by step is likely to be easier to accept than an all-or-nothing switch.
What The Platform Is Made Of
Flexxbotics groups the release into four solutions:
· Industrial data connectivity, which captures machine signals along with production context.
· Automated production tracking, following each order and part through machines, shifts and the people running them.
· Process characterisation, which profiles live production to set baselines and detect drift.
· Autonomous process control, which identifies risks and carries out corrective actions within the rules teams have approved.
Underneath, FlexxEdge provides edge software runtimes for factory workcells, and protocol-level drivers the company calls Transformers handle two-way communication with equipment. Flexxbotics says these drivers work with over a thousand different machine brands and models.
Working With What Factories Already Run
The platform is designed to work next to the ERP, MES, SCADA and PLC systems a factory already runs, not replace them. It can be deployed on a customer's own servers, in AWS, or in AWS GovCloud for defence-related work, with containerised edge software running on standard gateways and industrial PCs.
Earlier in the year, the company made its software-defined automation runtime, studio and API available as a free download, without time limits or disabled features. At the time, it listed support for common industrial protocols including OPC UA, MQTT, PROFINET, EtherNet/IP and FOCAS.
For background on how these layers fit together, our explainer on industrial IoT connectivity walks through the path from machine to cloud.
The Customer Results, In Context
Alongside the PMI and EIS figures, Flexxbotics lists several other customer outcomes:
· Ruland Manufacturing reached an 80% utilisation rate.
· SpiTrex Orthopedics shortened lead times by 20%.
· Orizon Aerostructures cut rework and scrap by 40%.
"Data-driven autonomy scales the impact of our most knowledgeable people so we're able to increase capacity even faster while we simultaneously raise quality," said Rick Newell, director of advanced manufacturing technology and innovation at Orizon Aerostructures.
How To Read Vendor-reported Gains
Numbers like these are useful signals, but they need care. A 125% capacity increase can reflect many things: a site that started with low machine utilisation, a move to unattended overnight shifts, a specific product line, or a combination of software and new equipment. Without the starting point and the time period, it is hard to compare one customer's result with another plant's situation.
Manufacturers evaluating the platform will want to ask for comparable cases, ideally in similar processes and at similar scale, and to run their own pilot before drawing conclusions. The same caution applies to any vendor in this space.
Why This Matters Now
The pitch lands at a time when manufacturers are struggling to find skilled staff. A 2024 study by Deloitte and The Manufacturing Institute estimated that US manufacturing could need as many as 3.8 million new workers by 2033, and that about 1.9 million of those jobs could go unfilled if the skills gap is not addressed.
"Increasing capacity isn't just about adding more plants and equipment. Companies must scale their experts as well to truly drive ROI," Bouchard said in announcing the release.
That framing, using software to extend the reach of a limited number of experienced people, is becoming a common thread in industrial AI. We explored a similar theme in our profile of Datanomix turning machine data into decisions, and in our look at how physical AI turns sensing into action on the factory floor.
The Governance Question Will Grow
As more software gains the ability to act on physical equipment, the question of who authorised what will become central. Industrial AI agents, whether built into platforms like Flexxbotics or connected through newer interfaces, raise similar issues around permissions and audit trails. Our piece on what Samsara is doing with MCP looks at how that debate is playing out in connected operations.
Flexxbotics' answer is to make the boundaries explicit and the decisions logged. Whether that satisfies quality managers, safety teams and auditors in practice is likely to depend on how each manufacturer sets up its rules.
The Bottom Line
Flexxbotics is pushing factory software from watching machines to adjusting them, with a control layer that lets production teams decide how far that autonomy goes. The customer results it reports are striking, but they are the company's own figures, without published baselines.
For manufacturers with more machines than experts, the approach is worth watching. The real measure will be whether governed, logged automation can deliver consistent gains across different plants, processes and teams, and not only at the sites where it has worked best.
