This interview analysis is sponsored by Poka and was written, edited, and published in alignment with our Emerj sponsored content guidelines. Learn more about our thought leadership and content creation services on our Emerj Media Services page.

Manufacturing output increasingly depends on a resource no balance sheet captures: the operational knowledge held by experienced workers.

In the United States, more than 25% of the manufacturing workforce is aged 55 or older, according to data compiled by the U.S. Bureau of Labor Statistics — a cohort approaching retirement and taking decades of process expertise with them. Facilities have no reliable system to capture what those workers know.

The consequences are measurable. Research from the National Institute of Standards and Technology found that process variability in manufacturing directly increases defect rates and rework costs. New hires are onboarded against formal procedures that often omit the informal knowledge that actually drives performance.

The underlying problem is structural: expert knowledge lives in people, not systems. A survey of 1,000 organizations conducted by APQC found that 92% of organizations do not consistently capture knowledge from soon-to-be retirees — even as 58% of C-suite leaders describe the risk as a very serious concern.

Emerj recently hosted a series on scaling frontline knowledge with AI in manufacturing on the AI in Business Podcast, featuring Antoine Bisson, CEO and Co-Founder at Poka; Sebastian Dykas, Director of Manufacturing, Engineering, and Maintenance at Smith + Nephew; and Anand Gnanamoorthy, Director of Corporate Strategy and AI at Ingersoll, examining how manufacturers can capture, standardize, and transfer critical operational knowledge before it walks out the door with a retiring workforce.

This article examines three critical insights from industry leaders on how manufacturers can close the expertise gap before it becomes an operational ceiling:

• Generative AI for operational knowledge conversion: Converting tribal knowledge and legacy processes into validated digital work instructions reduces the time, cost, and effort of maintaining accurate operational content across shifts and sites.
• Capture and standardize expert knowledge as structured digital assets: Converting frontline expertise into repeatable processes reduces production risk, minimizes waste, and sustains output quality as experienced operators retire.
• Worker-centered AI deployment as an accelerator for adoption: Anchoring AI to the operator rather than the process drives frontline adoption and builds the momentum needed to scale.

Generative AI for Operational Knowledge Conversion

Episode 1:  Solving the Expertise Gap with AI in Manufacturing  – with Antoine Bisson of Poka

Guest: Antoine Bisson, CEO and Co-Founder at Poka

Expertise: AI Strategy, Software Engineering, Manufacturing Technology, Industrial Knowledge Management

Brief Recognition: Antoine Bisson co-founded Poka and spent more than a decade as CTO, helping build the company into a pioneering social industrial platform focused on training, knowledge retention, and real-time information for manufacturers before its acquisition by IFS in 2023. Beyond Poka, he serves as a Board Member at Can-Explore and previously held board leadership with Fondation de BAnQ, while also acting as a Limited Partner with Inovia Capital and Blank Ventures. Bisson holds a Bachelor of Software Engineering with Distinction from the University.

Manufacturers broadly understand the knowledge‑capture problem. Few have a reliable path to solving it at scale. Antoine Bisson’s argument is that generative AI now provides that path — not by replacing the expert, but by dramatically reducing the effort required to convert what experts know into structured, usable guidance.

Antoine puts forward a straightforward proposal:

• Capture a video of an experienced operator performing a task.
• Feed that video into an AI engine, which converts it into a complete work instruction.
• Generate structured guidance automatically — step‑by‑step instructions, safety checkpoints, proof points, and validation gates
• Replace weeks of documentation with a single expert review.
• Eliminate the creation bottleneck that has historically blocked knowledge capture at scale.

This is the operational logic that connects knowledge capture to AI value. When that factory‑specific knowledge is captured and structured, an operator facing a broken machine can ask the platform what to do — and the AI reasons across everything every expert ever documented, returning a contextual answer in milliseconds. Without it, the same question returns no useful results.

Gnanamoorthy adds a dimension most manufacturers are actively overlooking. Decades of files, process records, and documents sitting across employee drives represent recoverable institutional knowledge — but when workers retire or leave, that material is routinely deleted on the assumption it is too messy to be useful. His position is direct: AI handles messy data well, and waiting for clean inputs before acting is itself an operational risk.

The human validation gate remains non‑negotiable throughout. Dykas reinforces this from a regulated manufacturing perspective — in environments where an error can injure someone or compromise a product, no AI‑generated content reaches the floor without expert sign‑off. The review step is not a bottleneck to be engineered away; it is the control mechanism that makes AI‑assisted documentation safe to deploy.

Capture and Standardize Expert Knowledge as Structured Digital Assets

Episode 2: Capturing Tribal Knowledge to Solve the Manufacturing Skills Gap – with Sebastian Dykas of Smith+Nephew

Guest: Sebastian Dykas, Director of Manufacturing, Engineering, and Maintenance at Smith + Nephew

Expertise: Manufacturing Engineering, Operations Leadership, Lean Manufacturing, Medical Device & Pharmaceutical Manufacturing

Brief Recognition: Sebastian Dykas is a manufacturing and engineering leader with experience spanning the automotive, medical device, and pharmaceutical sectors across major Fortune 500 companies. Prior to his leadership role at Smith+Nephew, he held engineering and operations leadership positions at Pfizer and Stryker, where he led manufacturing strategy, maintenance, capital projects, automation, and continuous improvement initiatives across medical and pharmaceutical production environments. Earlier in his career, Dykas completed Chrysler’s Institute of Engineering rotational development program, gaining hands-on expertise in advanced manufacturing, robotics, welding, lean manufacturing, and production leadership. He holds both a Master of Science and a Bachelor of Science in Mechanical Engineering from Oakland University.

The retirement of experienced operators does not just create a headcount problem — it creates a process quality problem. Sebastian Dykas explains how he has seen this play out directly in regulated manufacturing environments where the performance gap between senior and newer operators is not a matter of effort or aptitude, but of accumulated, undocumented know-how.

The scale of that gap is often larger than leadership realizes until it is too late to close it. Dykas describes what that looks like in practice:

“Some of the older, more senior workforce were able to provide double the quantity in a shift as someone who was only doing it for a short period of time. They had developed their own best practices — they knew just because of quantity and time and hours on the equipment how they could produce almost no scrap, how they could produce higher throughput. But it’s very difficult to put that and ingrain that into someone who’s starting off.”

— Sebastian Dykas, Director of Manufacturing, Engineering, and Maintenance at Smith + Nephew

The operational consequence is traceable. Yield, scrap rate, and throughput vary not because of equipment or materials, but because the knowledge required to perform consistently has never been standardized. Dykas argues the fix starts with establishing a replicable ceiling built from the best practices of top performers — and training every operator to that standard.

He also identifies a compounding problem most manufacturers underestimate: training quality is itself inconsistent. In a 24/7 operation, the best trainer is rarely available across all shifts. Shift-to-shift variability in yield and scrap is the diagnostic signal that this is happening — and most plants are not reading it that way.

Bisson reinforces this from an onboarding perspective. The new generation of frontline workers entering manufacturing expects digital-first guidance. Paper-based SOPs and classroom-style training are not just inefficient—they are misaligned with how newer operators learn and retain information, further compounding the knowledge transfer problem.

Antoine suggests that leaders look to close this gap:

• Identify top performers before they retire and document their process knowledge as the training baseline.
• Build sign-off procedures that verify true comprehension — not just completion of a digital quiz.
• Monitor shift-to-shift scrap and yield as a direct diagnostic for knowledge standardization gaps.
• Treat lengthy onboarding timelines for manual processes as an engineering problem, not an operational given

Worker-Centered AI Deployment as the Accelerator for Adoption

Episode 3: Why Manufacturing’s Most Valuable Data Isn’t in Any System — with Anand Gnanamoorthy of Ingersoll Rand​

Guest:Anand Gnanamoorthy, Director of Corporate Strategy and AI at Ingersoll Rand

Expertise: AI Strategy, Digital Transformation, Industrial Innovation, Business Strategy  

Brief Recognition: Anand Gnanamoorthy is a strategy and technology leader with more than a decade of experience spanning manufacturing, industrial innovation, and digital transformation. Prior to his current leadership role, he spent over 12 years at Frost & Sullivan, rising from Senior Research Analyst to Industry Director and Global Leader within the Industrial Practice, where he advised Fortune 500 executives on digital transformation, sustainability, energy transition, M&A, and Industrial IoT strategy. He has authored syndicated research, white papers, and industry articles focused on industrial markets and emerging technologies. Beyond corporate leadership, Gnanamoorthy serves on the advisory board of QuarkX AI, helping guide AI strategy and enterprise implementation. He is also pursuing a Doctorate of Business Administration focused on Business Intelligence and AI at Marymount University.

The most common reason AI initiatives stall on the shop floor is not technical — it is organizational. Anand Gnanamoorthy draws a sharp distinction between two deployment framings that produce very different outcomes: AI anchored to the process, and AI anchored to the worker.

When organizations lead with process optimization as the rationale for AI deployment, workers experience the initiative as something being done to them, and resistance follows predictably. Gnanamoorthy argues that reorienting the objective around making the individual operator’s job measurably easier changes the dynamic entirely:

“If you go and ask any employee and say: we want to capture your tribal knowledge, your expertise into a system they’re going to be very resistant. There’s a natural tendency of employees to resist those kinds of things. The challenge not just exists at the knowledge capture level, but also at the operational level — and that’s where most organizations underestimate what they’re dealing with.”

— Anand Gnanamoorthy, Director of Corporate Strategy and AI at Ingersoll Rand

The path through that resistance is sequencing, not messaging. Gnanamoorthy’s framework is grounded in organizational behavior: identify the innovators and early adopters in every workforce, equip them first, let them demonstrate value to peers, and allow peer influence to drive broader adoption organically.

Bisson echoes this from a platform design perspective. Tools built specifically for the shop floor — intuitive, real-time, and designed for the environment operators actually work in — generate their own adoption momentum. When a tool makes someone’s job visibly easier, the case for adoption makes itself.

Together, Antoine and Anand suggest a framework for enterprise leaders designing frontline AI rollouts:

• Start with one or two measurable use cases with clear ROI before expanding — complexity kills momentum.
• Anchor the use case to a specific worker outcome — faster answers, less rework, reduced cognitive load — not a process efficiency metric.
• Identify innovators and early adopters first; do not lead with a full-scale rollout.
• Let peer influence drive adoption; top-down mandates consistently underperform in frontline environments.