Inside Torque: How Syspro Wants Manufacturers to Trust Autonomous AI
Key Takeaways
⇨ Syspro Torque is a new industrial AI platform that detects manufacturing problems, recommends the next step and takes approved action inside ERP, MES or SCADA systems, with every decision logged and explained.
⇨ In an exclusive interview, Syspro Chief Solutions & Technology Officer Chris Lloyd explains how the Glass House principle, plain-language agent building and per-workflow cost budgets address the trust and ROI barriers holding back manufacturing AI adoption.
⇨ Syspro will showcase Torque publicly for the first time at IMTS 2026 in Chicago, following a controlled availability programme spanning food and beverage, industrial equipment and fabricated metals manufacturers.
Syspro has launched Torque, an industrial AI platform that detects operational problems, recommends the next step and takes approved action inside the systems manufacturers already use. Announced today, the platform works with any ERP and connects to MES, SCADA and warehouse systems through MCP connectors. It also logs every action it takes, including the rule applied, the data used and the reasoning behind the decision.
Torque entered a controlled availability programme in August with manufacturers and distributors across food and beverage, industrial equipment and fabricated metals. Syspro will showcase the platform publicly for the first time at IMTS 2026 in Chicago from 14 to 19 September. The company describes Torque as its most significant product launch in nearly five decades.
However, Torque is entering a market where manufacturers have been vocal about the gap between AI ambition and AI accountability. Ahead of the launch, ERP Today spoke with Chris Lloyd, Chief Solutions & Technology Officer at Syspro, about why autonomous operations have become possible now, how the platform aims to earn trust from an industry that has remained cautious about AI, and what early adopters are prepared to hand over to AI.
From Experimentation to Operationalisation
For Lloyd, the timing reflects a convergence of AI models, cloud infrastructure, industry data and manufacturers’ readiness to move beyond pilots. He also believes the nature of manufacturing itself is an important part of the equation.
Manufacturing is highly governed, regulated, time-sensitive and repeatable. Lloyd argues that earlier AI tools were too probabilistic for those conditions. The combination of probabilistic models with deterministic tools, industry knowledge and augmentation now creates what he sees as a safer environment for moving from experimentation to operationalisation.
Syspro has previously introduced AI capabilities, from machine learning and cognitive services to embedded automation that helped it secure a Leader position in the 2026 Nucleus Research SMB ERP Technology Value Matrix. What Lloyd says differentiates Torque is the way those capabilities have been packaged.
Definable and pre-packaged agents, autonomous workflows, skills, integrations, a rules engine, knowledge graph memory and cloud or on-premises deployment now sit within a single governed workspace.
The aim is to provide a complete package for autonomous operations rather than presenting separate capabilities under different labels.
Dissolving the Trust Barrier
When ERP Today previously spoke with Lloyd, he described a “trust barrier” around allowing AI to take action, particularly among midmarket manufacturers. He also highlighted the importance of keeping humans involved in the process.
Torque is Syspro’s response to that challenge, built around what the company calls the Glass House principle: everything visible, nothing hidden.
Lloyd describes Glass House as the opposite of a black box. Explainability extends across the platform’s lifecycle, from design and implementation through to operations and audit. Users can drill into individual actions to understand why they were taken, which data was considered, what guardrails were in place and which rules were applied.
For Lloyd, trust and explainability are central to moving AI into operational environments.
Manufacturers can also start at different levels of AI maturity, from human-in-the-loop assistance through to fully autonomous execution. The autonomy level can then increase as governance structures and confidence develop.
That graduated approach echoes the governance-first architecture Syspro CEO Leanne Taylor outlined earlier this year when describing AI as an “activation layer” for ERP.
Industrial AI First, ERP Second?
Asked whether an ERP-agnostic AI layer means Syspro now sees itself as an industrial AI company first, Lloyd’s answer was direct: “Absolutely yes.”
He argues that moving from ERP to industrial AI requires the unification of IT, operational technology (OT) and enterprise technology (ET). That means reaching into OT systems, warehouse management systems and manufacturing execution systems to create a unified contextual layer.
Torque can operate without Syspro ERP, although its capabilities run deepest when used alongside it. Lloyd describes ERP as the foundation of the intelligence layer, providing the underlying data on which Torque operates.
The two platforms can also provide value independently. Torque can even function as an integration and workflow mechanism ahead of an ERP transformation programme.
The platform does not replicate data, which Syspro says helps contain governance risk. The positioning also aligns with the company’s broader argument that manufacturing AI succeeds or fails at the ERP layer, where operational controls are already established.
No Code Does Not Mean No Control
Torque allows operations teams to describe an agent in plain language and deploy it without developers. This changes who can own AI initiatives inside a manufacturing organisation.
Lloyd describes this as a redistribution of labour rather than a removal of control. Domain experts can drive business value, while IT continues to control permissions, data boundaries, audit requirements and deployment standards.
The approach is based on the idea that transformation will increasingly involve a series of smaller, incremental workflow changes owned by the people who understand the desired business outcomes.
No-code therefore does not mean a lack of control. Instead, it is intended to bring AI development closer to business outcomes while maintaining tightly managed IT guardrails.
ROI a CFO Can Track
One of Torque’s key differentiators is its approach to cost governance. Manufacturers experimenting with AI can struggle to track spending, often managing costs retrospectively as they accumulate.
Torque takes a different approach. During setup, users define the KPIs that matter, whether that means fewer stockouts, reduced waste or faster disruption handling. Agents are then designed around those targets.
A built-in value calculator estimates the cost of running each workflow. Manufacturers can set the cadence and budget so that each workflow has an expected return before it begins operating.
This gives organisations visibility into where each workflow is running, how it is operating and how much of its allocated monthly budget it has consumed.
For CFOs, the key measures include cost avoidance, hours saved, scrap reduction and faster resolution times.
What Early Adopters Are Handing Over
Within the controlled availability programme, most use cases are focused on frequent, rules-based processes. These include sales order adherence, job status monitoring, supplier follow-ups and eliminating manual “swivel chair” integration work.
More complex use cases are also emerging at the enterprise end of the spectrum. Lloyd pointed to a supply chain control tower implementation as a major example, describing it as a significant implementation spanning the supply chain.
How far customers move towards autonomy will depend on the trust established in production. Visibility, transparency and explainability are central to that process. Customers want to see initial use cases working with humans in the loop before handing more responsibility over to autonomous systems.
Lloyd also highlighted two design choices intended to differentiate Torque. Deterministic workflows can be deployed on-premises for customers that require data sovereignty, while the platform is designed to be lightweight enough that customers do not need a large team of engineers to implement it.
What This Means for ERP Insiders
Autonomous AI is becoming an ERP selection criterion, and explainability is the gate. Torque’s reasoning chain, covering the rules applied and data considered, sets a standard that buyers should expect from other vendors. ERP users evaluating AI roadmaps should ask not only whether a platform can take action, but whether it can show who approved that action, why it was taken and which business rules governed it.
AI cost governance has moved from an afterthought to an architectural requirement. Torque’s per-workflow budgets and upfront cost estimates directly address the AI sprawl that has slowed manufacturing pilots. CFOs should expect workflow-level visibility into costs and ROI from agentic AI investments rather than reconciling expenditure retrospectively.
The ERP vendor category is being redrawn. Syspro’s ERP-agnostic positioning, combined with its Nucleus Leader ranking and embedded AI investments, suggests that midmarket industrial vendors are increasingly competing as intelligence layers across IT and OT rather than simply as systems of record. ERP users should reassess whether their incumbent’s AI strategy reaches the shop-floor systems where operational decisions are ultimately made.