By: PYMNTS
Date: September 11, 2026
Main Facts
The industrial manufacturing sector stands on the precipice of a silent operational revolution. For decades, the true mechanics of heavy industry relied on a fragile human artifact: the tacit knowledge held exclusively by veteran factory technicians. When an aging workforce retires, decades of institutional troubleshooting capability walks out the door with them.
Recognizing this critical vulnerability, industrial artificial intelligence startup Squint has introduced a suite of autonomous operational tools—anchored by its flagship Lean Manufacturing Agent—designed to digitize, analyze, and optimize human workflows on the factory floor.
Rather than relying solely on high-cost robotic arms or generalized machine learning models, Squint builds proprietary, secure "context layers" for individual manufacturers. By ingesting factory floor video feeds, informal legacy documentation, and enterprise resource planning (ERP) data, Squint bridges the gap between raw human labor and digital automation.
Early deployments of the technology have yielded striking results. Industrial clients report cutting scrap rates by as much as 50% and dramatically accelerating training cycles for complex new processes. Looking ahead, Squint’s founder and CEO, Devin Bhushan, envisions a future of "instant changeovers," where factories can pivot production lines at a moment’s notice to match fluctuating market demands.
Chronology
The Pre-AI Industrial Era: The Stopwatch and Clipboard
For over a century, industrial engineering methodologies like Lean and Six Sigma have relied on manual time-and-motion studies. Industrial engineers would stand on factory floors with stopwatches and clipboards, observing workers, logging bottlenecks, and writing up manual reports that often took days or weeks to process. Simultaneously, troubleshooting relied on unwritten tribal knowledge—an informal oral history of equipment quirks and quick fixes passed down from senior technicians to apprentices.
The Foundation of Squint: Recognizing the "Context Gap"
Before founding Squint, Devin Bhushan encountered a recurring roadblock while consulting for industrial clients. He realized that generalized machine learning and applied AI applications were failing on factory floors for a fundamental reason: there was no structured digital context for the AI to analyze. Crucial procedural details existed only in human memories or fragmented paper logs. Bhushan launched Squint to solve this foundational problem, focusing first on creating a digital "context layer" before attempting to deploy active AI solutions.
Scaling the Context Layer: The Pivot to Specialized Small Models
Initially, building a comprehensive digital context layer for a single manufacturing facility was a labor-intensive, costly bottleneck. Processing massive volumes of factory floor video took large language models between 10 and 14 days per customer. To overcome this hurdle, Squint engineered a proprietary, highly specialized 2-billion-parameter model. Dedicated exclusively to watching human operations footage and mapping it against legacy maintenance records, this small language model reduced deployment times and preserved client data privacy.
The Evolution into Autonomous Agents
With individual, secure context layers established for heavy hitters like Pepsi and specialized service providers like Carolina Handling, Squint transitioned from documentation to active operational analysis. The company rolled out its Lean Manufacturing Agent, replacing the traditional stopwatch with real-time video intelligence. Following this, Squint released specialized conversational agents tailored for pre-work safety checklists and predictive forklift maintenance diagnostics, transforming static guidelines into dynamic operational support.
Supporting Data
The integration of artificial intelligence into industrial manufacturing is no longer a theoretical exercise; it is backed by measurable efficiency gains across diverse supply chains.
- 50% Reduction in Scrap Rates: Early enterprise adopters of Squint’s technology have reported slashing their material waste by up to half, directly attributable to fewer human errors and faster, more accurate diagnostic workflows.
- 10 to 14 Days to 2-Billion-Parameter Efficiency: Squint’s streamlined approach utilizes a lean 2-billion-parameter AI model to process weeks of factory floor footage, dramatically outpacing the processing timelines of generalized large language models.
- 150 Safety Checks Automated: Complex pre-work safety protocols—often involving upwards of 150 individual manual inspections performed by lone operators—have been compressed into conversational, AI-guided workflows that generate prioritized safety plans instantly.
- Immediate Efficiency Metrics: The Lean Manufacturing Agent categorizes every second of a recorded work cycle into three definitive buckets: value-added, necessary, or waste. This granular breakdown allows plant managers to intervene immediately rather than waiting for weekly or monthly review cycles.
Official Responses and Industry Insights
The philosophy driving Squint’s technological roadmap is deeply rooted in the realities of day-to-day factory management. According to CEO Devin Bhushan, the traditional approach to industrial automation often missed the human element entirely.
"There was a lot of stuff that was in people’s heads that had never been documented," Bhushan explained in an interview with PYMNTS. "No amount of ML or applied AI could actually help them because there was no context for that AI to work on top of."
Addressing the sensitivity surrounding industrial data, Bhushan emphasized that Squint avoids generalized, shared models. Manufacturers fiercely guard their proprietary recipes and operational workflows—exemplified by enterprise clients like Pepsi, whose production secrets cannot and will not be blended into a public or shared foundational model.
On the ground level, the Lean Manufacturing Agent operates as a collaborative partner rather than an authoritarian monitor. According to official company release documentation, raw footage converted into gapless time-and-motion studies yields specific, actionable recommendations. Plant managers retain full editorial control, reviewing, modifying, or rejecting AI-generated fixes before they are implemented on the floor.
For service networks like Carolina Handling, the technology bridges the gap between unstructured problem descriptions and targeted field execution—identifying exact equipment faults, required replacement parts, and the specific technician skill sets needed for the repair.
Implications
The deployment of Squint’s Lean Manufacturing Agent and context-layer architecture carries profound implications for the future of global supply chains, labor markets, and manufacturing economics.
1. Reversing the Impact of the "Silver Tsunami"
As manufacturing workforces age, the retirement of veteran technicians threatens to paralyze industrial productivity. By capturing tacit knowledge—how a seasoned worker intuitively interacts with a stubborn machine—through video ingestion and contextual mapping, companies can permanently archive and scale institutional wisdom. New hires can be onboarded in a fraction of the traditional time, effectively flattening the steep learning curve of advanced industrial engineering.
2. Moving from Lean Theory to Real-Time Execution
Traditional Lean and Six Sigma methodologies were inherently retrospective. Engineers analyzed past inefficiencies to make future adjustments, often days or weeks after the waste had already occurred. Squint’s technology shifts industrial engineering into real time. By zooming dynamically from a single workstation to an entire production line, the Lean Manufacturing Agent balances workloads the exact moment a station falls behind target cycle times, neutralizing bottlenecks before they cascade across the facility.
3. The Holy Grail: The "Instant Changeover"
Perhaps the most ambitious implication of Squint’s long-term strategy is CEO Devin Bhushan’s vision of the "instant changeover." Currently, retooling a major manufacturing plant to shift production from one product line to another—such as pivoting a snack food facility from manufacturing Cheetos to Doritos—can take months or even years of planning and physical reconfiguration.
By creating a deeply digitized, AI-governed context layer that understands every variable of factory operations, Bhushan aims to shrink that transition window to match the speed at which raw materials arrive at the loading dock. If achieved, this capability would allow global manufacturing capacity to flex dynamically in real time with shifting consumer demand, fundamentally reshaping the economics of industrial production in the 21st century.
