AI and automation's adoption gap: why most manufacturers haven't crossed the line yet

Eighty percent of U.S. manufacturing facilities operate with zero automation today. That figure, attributed to Intrinsic Chief Technology Officer Brian Gerkey and reported by Manufacturing Dive, puts the scale of the gap in stark relief at a moment when the industry's headlines are dominated by AI and robotics investment.

The disconnect is not about awareness. Deloitte's 2025 Smart Manufacturing and Operations Survey found that 92% of manufacturers believe smart manufacturing will be the primary driver of competitiveness over the next three years, according to Manufacturing Dive. Yet only a small fraction say AI is widely deployed across their operations. Jeff Burnstein, president of the Association for Advancing Automation, told Manufacturing Dive that interest is uniformly high, but execution is where things consistently break down.

Labor pressure is raising the operational stakes

The urgency is intensifying because the workforce math is worsening. Honeywell CEO Vimal Kapur told CNBC in June that his customers across hospitals, airports, data centers, semiconductor facilities, and LNG plants are struggling to find qualified operators and technicians. Aging populations and slowing workforce growth mean the problem compounds over time.

Kapur framed the response not as headcount replacement, but as a redefinition of what automation is for. Customers, he told CNBC, are treating AI-enabled systems as revenue-generation tools rather than pure cost-reduction plays. That framing has direct implications for how procurement and operations teams should be writing business cases internally.

Honeywell is positioning itself at the center of this shift. The company completed the spinoff of its aerospace division on June 29, leaving behind a pure-play automation business spanning sensors, controls, and software. Kapur told CNBC that the company's competitive advantage lies in the domain expertise and operational data already flowing through its installed systems, which AI can now convert into optimization decisions that previously required human judgment.

New tools are expanding the entry points

Part of what has kept adoption low is that many automation investments historically required significant infrastructure commitment. A cluster of product launches in late June is beginning to change that calculus at the device and software layer.

ABB introduced its Ability Field Information Manager 3.5 platform, a vendor-neutral system designed to automate bulk firmware updates and manage mixed fleets of field devices across multiple manufacturers, according to Automation International. For operations teams running heterogeneous environments, the ability to manage multi-vendor device fleets through a single platform reduces one of the persistent friction points around automation: the cost and complexity of managing legacy equipment alongside newer systems.

On the sensing side, Sonair announced what Automation International describes as the world's first safety-certified 3D ultrasonic sensor. The ADAR One has achieved SIL2 and PL d ratings and meets the requirements of the European Machine Directive for detecting humans and objects. That certification matters operationally: SIL2 and PL d compliance is the bar for deploying sensors in proximity to human workers in regulated environments, and until now, 3D ultrasonic technology has not cleared it.

Contrinex is also expanding in the sensing space, adding miniaturized smart measurement sensors and 3D vision systems aimed at real-time monitoring applications, per Automation International. Compact form factors open automation options in environments where space constraints previously ruled out sensor deployment.

The platform layer is consolidating too

Below the device level, the control software tier is also evolving. Congatec and CODESYS announced a partnership combining congatec's hypervisor technology with CODESYS control software to create virtualized real-time control platforms for mixed-critical industrial workloads, as reported by Automation International. Virtualizing real-time control functions on shared hardware is a meaningful architectural shift for facilities trying to consolidate OT infrastructure without sacrificing determinism.

Watlow, separately, introduced an edge process management platform built for regulated thermal manufacturing environments, targeting the challenge of maintaining precise process data and secure digital records across temperature-driven operations, according to Automation International. Regulatory compliance documentation has been a persistent manual burden in pharmaceutical, food, and specialty chemical manufacturing; edge platforms that automate record integrity are directly addressing an operational cost center.

What this means for your team

• Revisit your automation business case framing: Honeywell's Kapur is telling CNBC that the strongest argument is revenue generation, not cost reduction. If your internal proposals are still primarily cost-out narratives, that framing may be limiting approval velocity.
• Evaluate vendor-neutral device management before your next field device refresh: ABB's FIM 3.5 targets the multi-vendor fleet problem directly. Operations teams managing mixed-manufacturer environments should pressure-test whether a neutral management layer reduces integration costs before committing to proprietary upgrade paths.
• Check safety certifications on any 3D sensing shortlist: Sonair's SIL2 and PL d clearance for the ADAR One is a first in 3D ultrasonic. If you are deploying collaborative or autonomous systems near workers in regulated settings, confirm your sensing stack meets the applicable functional safety standards, not just performance specs.
• Assess virtualized control readiness: The congatec-CODESYS partnership points toward consolidating OT compute through hypervisors. Facilities planning infrastructure refreshes in the next 12-18 months should evaluate whether real-time virtualization fits their mixed-criticality workloads before locking in dedicated hardware architectures.