ORLANDO, Fla.—IFS at its Unleashed conference here last week announced the company’s industrial AI Harness, which is designed to help enterprises implement AI within the policies, controls and operations that govern their business.  

See also: New IFS product grabs industrial data for a single, AI-powered interface 

AI Harness extends its existing security, safety, governance and trust controls. The platform provides context, permissions, governance, and controls that allow its agentic “digital workers” to execute work within established business rules, according to IFS. 

AI Harness also serves as the operational foundation to IFS Intelligence, which the company said can collate data from all types of industrial assets, systems, and processes all around an operation and turn that information into actionable tasks that flows through one conversational interface. 

According to new research from IFS, 43% of frontline teams have trained and introduced agentic digital workers into their day-to-day operations over the past 12 months. However, only 11% have validated AI outputs for safety, quality, or mission-critical processes, and 11% have established control mechanisms for AI use with sensitive or export-controlled data.  

As reported by Smart Industry last month, research from Futurum also found that while 66% of enterprises plan to invest in digital workers within the next 12 months, just 5.7% trust AI to act fully autonomously.  

See also: How data, governance and organizational change define AI success

To help companies address these control gaps, IFS designed the Harness so that AI operates through the customer's existing controls. It inherits permissions and access controls from connected systems, using yet another IFS product, IFS Cloud, as the foundation source, so no raw data is sent to external AI providers.

The harness has six key components, according to IFS:   

• Planning and orchestration break goals into steps and coordinate agents and tools.  

• Skills and connectors define and validate the systems agents can use.  

• Memory and context carry progress forward and assemble only the information relevant to each task.   

• Guardrails and permissions enforce identity, approvals, and limits on risky actions.   

• Verification and escalation check work for correctness and policy compliance, escalating when needed.  

• Observability and auditability logs every decision and action for review, so nothing happens that cannot be traced back after the fact.   

The harness is also designed to work across AI models and services, so customers are not locked into a single provider. It also allows customer requirements, like where data can be processed, to remain part of how AI is deployed, IFS said.  

Podcast: Do manufacturers dream of 'digital workers'? 

Other features, according to the company, include:

• Sovereignty: All data stays within the system of record; no raw data sent to external AI providers.  

• Staged autonomy: Workers start with co-pilot assistance, progress to recommendations, then to autonomous action, with escalation on uncertainty.  

• Industrial context: Domain expertise and curated tooling that prevents agents from being paralyzed by 70,000 available APIs.  

• Multi-layer governance: Security at the cloud level, the harness level, and the model level; not a single control, but aggregated assurance.