AI Is Everywhere. What About Operational Intelligence?
Artificial intelligence has become every B2B technology vendor's favourite word. The promise for smarter planning, predictive maintenance, autonomous scheduling, AI-powered insights is everywhere. But manufacturers don’t need to know whether AI is available, they need to know whether the business has the operational foundation underneath it to make that AI worth anything.
One question gets skipped in almost all of the hype:
What is the AI actually making decisions from?
2026 hasn't given manufacturers much room to figure that out at a leisurely pace. Skilled-labour shortages are still the top-ranked operational challenge in industry surveys this year (ahead of tariffs, ahead of input costs) with the workforce gap projected to exceed two million workers over the next decade as experienced operators retire faster than they're replaced.
At the same time, trade policy keeps moving under manufacturers' feet: new U.S. tariff actions on steel, aluminum, and a broader list of Canadian exports have landed through the first half of 2026, and CUSMA itself is under formal review starting July 2026 (this month!). Manufacturing output in the tariff-exposed segments, such as autos, wood products, primary metals, has taken the sharpest hits, while everyone else absorbs the uncertainty of not knowing what the next change will be.
None of that is solved by adding AI.
If anything, it raises the stakes on getting the fundamentals right first.
Without connected, reliable, real-time operational data, AI isn't making smarter decisions. It's making faster decisions based on incomplete information. Which, on a factory floor, is often worse than making no decision at all.
What Manufacturers Need
Most manufacturers already have access to an overwhelming amount of information:
- production schedules,
- inventory levels,
- machine performance,
- supplier deliveries,
- quality inspections,
- sales forecasts.
And as headcounts start shrinking, the small stuff that an experienced supervisor used to notice on a routine round is more likely to go unnoticed until it's expensive.
Every day, small operational issues occur across the business:
- Supplier shipment is delayed.
- A production order falls behind schedule.
- Scrap rates tick up.
- Inventory drops below safety stock.
- A quality inspection fails.
Left unnoticed or untreated, these disruptions compound into missed deliveries, downtime, unplanned costs, and dissatisfied customers.
Operational intelligence continuously monitors these activities, flags exceptions as they happen, and directs attention to the issues that require a decision, instead of burying them in a report nobody reads until Monday.
Operational intelligence: The
real-time monitoring of operational data that flags deviations from plan as
they happen. A monitoring layer, not an AI model itself; AI can sit on top of
it to catch subtler patterns, but the core goal is getting the right signal to
the right person, fast.
ERP Provides the Foundation
An ERP system is the backbone of a modern manufacturing business. It centralizes production, purchasing, inventory, finance, quality, and customer data into one source of truth. That part hasn't changed.
But a single source of truth isn't the same as an early warning system. Most ERP implementations are very good at recording what happened and reasonably good at telling you what's supposed to happen next. They're less good at telling you the moment something starts to drift off plan.
Operational intelligence sits on top of that ERP foundation, continuously watching for deviation. Instead of waiting for an end-of-day report or a weekly production meeting, managers get a meaningful alert when a job, a supplier, or a machine starts falling out of line with the plan while there's still time to do something about it.
That gap is already showing up in how manufacturers are spending. CADDi's 2026 Manufacturing Outlook Study found planned investment in ERP and MES systems fell to 33% of manufacturers this year, down sharply from 60% in 2025. Meanwhile investment in robots and other floor equipment climbed to 69%. It's an understandable instinct: a new robot is visible on the floor; an ERP upgrade isn't. But pouring money into equipment while starving the systems that make sense of what that equipment is doing is exactly the pattern that leaves AI initiatives running on incomplete data.
That shift, from reviewing what happened yesterday to responding to what's happening right now, is what separates operational intelligence from traditional reporting.
AI Is a Tool, Not Your Next Employee
A lot of AI marketing implies that businesses should be looking for ways to replace certain jobs. Given the labour shortage most manufacturers are actually living through, that pitch is a little tone-deaf. The more urgent problem for most plants isn't too many workers, it's not enough of them.
Let us set the record straight: AI's value isn't in replacing experienced planners, supervisors, or production managers. It's in helping the people who are already stretched thin make faster, better-informed calls. AI can scan through operational data far faster than a person can, surface patterns, and flag exceptions worth a human's attention. What it doesn't have is the shop-floor judgment that comes from years of knowing this customer, this machine, or this supply base.
People still make the decisions. AI just helps them get to a decision faster, with more of the picture in front of them. Some in the industry describe this as the core idea behind Industry 5.0.
Industry 5.0:
technology that augments the people running the operation rather than trying to
remove them from it.
Whether or not that label sticks, the underlying point holds: on a floor that's already short-staffed, a tool that makes the people you do have more effective is worth more than one that promises to need fewer of them.
AI Is Only as Good as the Data Behind It
This is where a lot of AI rollouts run into trouble. Companies adopt AI tools while still running on disconnected spreadsheets, manual updates, siloed departments, and inconsistent reporting.
That's a bit like installing a top-tier navigation system in a vehicle running on an outdated map.
If inventory counts aren't accurate. If production updates lag by a shift or a day. If quality data lives in a separate system from scheduling. If purchasing and production aren't talking to each other, AI can't fix any of that. It can only generate recommendations from whatever data it's actually given, accurate or not.
Operational intelligence is the layer that makes sure the data is connected, current, and trustworthy before AI starts analyzing it. Without that layer, AI is just another dashboard nobody fully trusts. With it, AI becomes something planners will actually act on.
The Competitive Advantage Isn't AI. It's Faster, Better Decisions.
Manufacturers can't hire their way out of a labour market that's short two million workers.[1] They can't tariff-proof a trade relationship that's still being renegotiated. What they can control is how quickly they notice a problem and how well-informed the response is when they do.
As the operating environment gets more volatile (e.g. tighter labour, shifting trade rules, thinner margins for error) the advantage shifts to the manufacturers who can spot exceptions early, prioritize the ones that matter, and get the right information to the right person before a small issue becomes an expensive one.
ERP provides the operational foundation. Operational intelligence turns that data into something actionable. AI accelerates the analysis and the recommendations on top of it.
Together, they build a manufacturing operation where technology supports the people running it, instead of asking them to keep up with it.
In 2026, operational intelligence isn't just the missing layer between ERP and AI. For manufacturers navigating a tighter labour market and a moving trade landscape, it's what lets both the people and the technology actually perform.