Honeywell is using AI-driven workflows to assist manufacturers in addressing issues like tighter labor markets, fewer skilled workers, product complexity, and pressure to meet deadlines, prioritizing asset optimization and reliability through AI agents.  

The multinational maker of mission-critical technologies, automation systems, and aerospace solutions is helping manufacturers redesign workflows to be more connective, emphasizing the benefits of AI-enabled maintenance workflows that are built on strong data foundations and connectivity.  

Podcast: With outdated and isolated OT, your AI strategy may be built on a blind spot 

AI plays a prominent role in this, according to one company expert, in automizing workflows.       

According to Omar Sayeed, digital reliability leader at Honeywell, the next stage of that evolution will be defined by AI, agent-based workflows, and increasing levels of automation. 

Speaking at the 2026 Honeywell User Group Americas Conference in Phoenix last month, Sayeed outlined how manufacturers can progress toward autonomous asset optimization by strengthening data foundations, deploying predictive technologies, and redesigning maintenance workflows. 

"It's very clear. To drive more autonomy in asset optimization, we're going to have to leverage more technology and leverage it differently," Sayeed said.   

Leveraging AI for ensuring asset reliability 

Asset reliability has become a strategic focus area for Honeywell, Sayeed added, with recent acquisitions in the last few years of turbo machinery equipment manufacturer Sundyne and Compressor Controls Corp., which provides machinery train optimization services for the oil and gas industry.   

The acquisitions build upon Honeywell's existing asset performance management platform and expand its expertise in equipment and reliability services, but the long-term goal is to extend beyond asset monitoring.  

"This presentation is really to talk about where we're headed,” Sayeed said, namely how AI is impacting the way customers are taking care of their assets.

Honeywell is emphasizing this as manufacturers across verticals invest in AI agent implementation. Meanwhile, industries grapple with supply chain issues, workforce problems, and economic instabilities as they navigate how to implement AI agents effectively. 

According to Sayeed, connected workflows could help companies continue to automate amidst these challenges.  

Workflows and asset optimization  

Sayeed emphasized that connected workflows should involve connecting “the insights that come from your applications to actions that happen in the field.”  

He described autonomous asset optimization as the integration of multiple technologies designed to reduce human intervention while still improving asset performance.   

Achieving asset autonomy requires several foundational elements, he added 

"It requires really robust data collection. It requires good analysis and prediction. Aid in decision-making requires some ability to take action, ideally autonomously, or make recommendations to the human in the loop," Sayeed said. The important infrastructure and components needed include sensors, the automation and control network, and analytics platforms.  

Some examples of industrial autonomy, Sayeed said, might be self-calibrating sensors, self-calibrating analyzers, or automatic load sharing between compressors to improve fuel efficiency and energy.  

See also: AI stokes debate over cloud-powered compute vs. on-prem 

“If I was to give an example of what autonomous might look like to me, a self-calibrating sensor, [or a] self-calibrating analyzer is an example of autonomous load sharing, having automatic load sharing between compressors to improve, let's say, fuel efficiency and energy,” he said.   

“It's another example of an autonomous system where humans necessarily don't have to be militant in the decisions that the machines are taking actions for,” he said. 

Sayeed added that moving forward, the questions begin to revolve around companies they can execute that at a bigger scale while still retaining reliability. 

He also identified several barriers standing in the way of some organizations advancing toward autonomous operation, including: 

• Sustaining those programs requires resources 

• Disconnected maintenance workflows present unique challenges  

• Operational trade-offs between equipment maintenance and production  

He added that to drive more autonomy in asset optimization, companies are going to have to leverage more technology—and leverage it differently. 

Six key workflows  

To address this, Honeywell has identified six workflows that underpin effective autonomous asset optimization.  

Sayeed said that for some customers, Honeywell monitors their assets, and then when they talk to our engineers who are in the monitoring centers, looking at the assets, they ask them what their process is. 

This involves asking questions like, “How do they take the insight from [Honeywell] and how do they help convert that for the customer into action in the field? And the breakdown of those steps are actually the breakdown of what a human being is doing to reason and to resolve.”  

The first workflow is asset surveillance, focused on triaging increasing volumes of alerts generated by predictive systems. "When you try to move to be more proactive, you naturally get a lot more alerts," Sayeed said, which require investigation and prioritization.  

In the case of this asset, they ask how to best deal with those alerts and how they can have an AI actually dispose of those alerts faster than a human being can. There are different techniques, he said, adding that they would then consult with their monitoring-center engineer regarding how they can put assets into autonomous workflows.   

The second workflow is root-cause analysis. "In an autonomous workflow, we would allow the agent to actually perform a 5-Why or a fishbone and present the evidence to a human being, rather than a human being have to go and fetch all of the information, so that’s the root-cause analysis,” he said. 

After identifying the problem, the next question is how to specify what actually needs to be done.  

“That's where having prescriptive models that can isolate a particular failure once it's identified and provide recommended actions, and it's also linked to where your subject matter experts can update those recommendations, the prescriptions that come out of these kinds of systems much more actionable.”   

The third workflow involves actionable insights through prescriptive recommendations. Having prescriptive models that can isolate a particular failure once it's identified and provide recommended actions helps organizations respond more effectively, Sayeed said.  

The fourth workflow focuses on optimizing maintenance strategy by incorporating dynamic risk information into reliability planning.

"If we want to actually reduce our maintenance costs in the long run, what would be helpful would be to update our maintenance strategies with insight that's coming from dynamic risk. So, evaluating risk that’s coming based on our sensing and our analytic models. That's another very important workflow that could be powered from asset optimization, improving the asset operation,” he said.  

The fifth area involves improving asset operation by evaluating the production and reliability trade-offs. “Being able to make the trade-offs between the reliability of the asset and the process requirement, and evaluate that quickly to make decisions, is really important as far as extending the asset operation,” Sayeed said.  

The sixth supports field workers by ensuring technicians have the information they need at the point of execution, ensuring that the workers have the right instructions at the right time when they need to go and carry out maintenance. 

Maturity model and data Integration 

Central to Honeywell's approach is an asset-management maturity model that describes how organizations evolve in their reliability practices.  

“The foundation of Honeywell's approach is predictive health and performance,” Sayeed said. He explained that the model is driven by increasing levels of data quality and automation.  

Increasing data sophistication can involve deploying IIoT technologies, self-calibrating smart sensors, and intelligent analyzers capable of self-diagnosis, progressively shifting decision-making and actions away from manual intervention.  

Sayeed said that organizations hoping to adopt autonomous workflows first need reliable sensing technologies, strong control infrastructure, and analytics platforms capable of transforming raw data into useful information. For many facilities, the biggest barrier is fragmented or poor-quality data, which AI cannot help.  

"[Asset autonomy] requires really robust data collection. It requires good analysis and prediction,” he said. 

See also: Report: Manufacturers split on whether to prioritize industrial AI 

Data integration goes beyond Honeywell’s model, as data can determine how successful AI agent training and implementation will be on factory floors. With strategic integration of data across company teams, agent training is much more effective.  

This is seen in manufacturing companies such as Wolfspeed, which has its own data repository it uses to train agents.  

Honeywell’s 'North Star’  

For organizations beginning this journey, Sayeed stressed that success starts with fundamentals. "The first step is to have a foundation in place. That foundation includes identifying critical assets, strengthening data infrastructure, and deploying predictive models against the different asset classes. Establish a robust data foundation."  

Organizations must also establish standardized work processes and adapt their operating models to support increasingly remote reliability functions while remaining centralized across the enterprise. “It's really important to consistently gather this information in a centralized manner,” Sayeed said.  

The progression then moves from predicting failures toward prescribing actions, basically transitioning from ‘when it's going to break’ to ‘what should I do about it.'  

See also: Stories of AI adoption: Wolfspeed all-in with 22 agents across key company teams    

Risk-based prioritization remains key in determining the response window to issues and orchestration between departments. Eventually, AI agents, Sayeed said, may coordinate activities across surveillance work groups, subject matter experts or maintenance personnel.  

"The last one is the North Star: implementing autonomous action," Sayeed concluded, describing a future in which systems can make reliability-informed control decisions with minimal human involvement.  

For process manufacturers struggling with labor shortages, increasing asset complexity, and growing reliability expectations, the path to autonomous operations begins not with replacing people, but with equipping them through better data, stronger maintenance workflows, and AI-enabled decision support.