There is a widening gap between the sophistication of manufacturing data models and the reality of the production line.
During a recent planning session, I noticed a common paradox: a data science team presenting a masterful demand forecasting model that was entirely inaccessible to the operations team standing three feet away. When asked for a simple stock-level check, the system couldn’t deliver. This captures the “complexity trap” perfectly—an industry spending billions on AI adoption yet leaving 95% of its workforce unable to use the tools.
The complexity trap
Manufacturing is in the middle of an analytics boom. The global manufacturing analytics market grew from $16.6 billion in 2025 to over $20.6 billion this year — a 24% jump. AI adoption in manufacturing leapt from 87% to 93% in a single year, and 95% of manufacturing and supply chain leaders now consider AI a requirement for competitiveness.
The problem is the tooling has become extraordinarily complex. Modern data architectures often involve stitching together multiple programming languages, orchestration layers, and specialised platforms that demand deep engineering expertise to operate. For many organisations, only the people who loaded the data into the platform can do anything meaningful with it.
We’ve been here before — it was called Hadoop. A decade ago, organisations poured millions into Hadoop clusters, only to discover that the complexity of getting data in and out meant that insights remained locked behind a technical priesthood. The promise of “big data for everyone” became “big data for the three people who understood MapReduce.”
Today, a different generation of complex architectures risks repeating that same mistake. And in manufacturing — where 74% of operators report an acute shortage of skilled workers, and technicians waste up to 30% of their time just searching for information — complexity isn’t merely inconvenient. It’s a genuine barrier to competitiveness.
Designing for the 95%, not the 5%
The instinct in our industry has been to build platforms for power users and hope that adoption trickles down. It rarely does. What actually works is the opposite: design for the 95% of use cases that businesses genuinely need, and make those capabilities native, intuitive, and accessible.
Take time series forecasting. Most manufacturing planners don’t need to deploy a custom neural network. They need a reliable forecast for next week’s demand, callable without a PhD. Platforms that embed forecasting as a built-in feature — rather than requiring users to assemble a pipeline from components — dramatically widen the pool of people who can actually benefit.
“The value of an insight is zero if nobody outside the data team can reach it. The companies getting ahead are those that treat simplification as a design principle, not a compromise.”
SQL: the great equaliser
One of the most powerful — and often overlooked — decisions in democratising analytics is the choice of interface language. Many modern platforms require proficiency in specialized programming languages. But there’s a language that millions of professionals already speak: SQL.
When advanced capabilities like AI inference, anomaly detection, and forecasting are accessible through SQL, the population of people who can benefit expands by orders of magnitude. You’re not training new specialists; you’re unlocking people you already have.
Cortex Code: the “No-Code” frontier
SQL lowered the barrier to data, but Cortex Code removes it entirely. While SQL was the bridge, Cortex Code is the open door.
On a factory floor, the real challenge is unifying fragmented PLC tags, historian signals, and ERP data. Even SQL can be a hurdle for a busy supervisor who just needs a trusted view of the line. Cortex Code—a Generative AI layer in Snowflake—removes this barrier by translating natural language into technical implementation.
Instead of writing complex code, a user can simply ask:
“Create a unified namespace for Line 4 that links my historian data with SAP production orders.”
Cortex Code handles the schemas and pipelines automatically. This shifts data modelling away from the “coding elite” and creates a shared OT/IT language that anyone on the plant floor can use.
Democratise access, not chaos
Of course, opening up analytics to more people creates a legitimate concern: governance. If the operations lead, the quality manager, and the plant director all have access, how do you ensure security?
The answer lies in a robust role-based access framework that makes both possible simultaneously:
- Need-to-know restrictions: Define them once, enforce them everywhere.
- Data subsets: Share only what is relevant to specific teams or partners.
- Centralised management: Clear role definitions across thousands of users.
- Encryption: Using external keys to protect sensitive intellectual property.
What this means for UK manufacturing
The UK manufacturing landscape is under real pressure. Rising costs, global competition, and a workforce gap that shows no sign of closing. Against that backdrop, 80% of manufacturing executives are planning to invest 20% or more of their improvement budgets into smart manufacturing initiatives this year.
But the return on that investment depends entirely on how many people in the organisation can actually use what gets built. A predictive maintenance model that only the data science team can interpret is an experiment. The same model, accessible to every shift supervisor via a simple natural language prompt, is a competitive advantage.
The difference between those two outcomes isn’t more technology. It’s less complexity. Simplicity scales; complexity doesn’t.
Michael Gerstlauer: Manufacturing Field CTO, Snowflake
Connect with Michael here.
References & data sources
- Fictiv (2026). 2026 State of Manufacturing Report.
Key Finding: AI adoption has reached a critical mass, with 93% of manufacturers now integrating artificial intelligence into their core workflows. - Fictiv & The Robot Report (2026). The AI Mandate in Robotics and Automation.
Key Finding: The industry has shifted from viewing AI as an “advantage” to a “necessity,” with 95% of executives classifying it as a formal requirement for operations. - HM Government / GOV.UK (2025). UK Industrial Strategy 2025-26: Building Resilience.
Key Finding: Labor remains a primary constraint on industrial growth; 74% of firms report a persistent workforce shortage impacting production capacity. - Rockwell Automation & Snowflake (2025). Cloud-Scale Analytics: A Manufacturing Case Study.
Key Finding: Through the implementation of unified data platforms, organizations achieved a 65% reduction in the time required for system diagnosis and troubleshooting.
For more articles like this, visit our Industrial Data & AI channel.
