Industry research · 2026 edition

A survey of 167 manufacturing leaders

The State of AI Data Readiness in Manufacturing

Explore the data from 167 manufacturing leaders and download the full report.

Download the report Explore the findings

Scroll

Foreword

Why we ran this survey

“Your AI system is not going to walk out to the floor and ask the operator what they saw, or why it happened.”

83% of the leaders we surveyed are already piloting or running AI. Those systems are reading production data, proposing root causes and predicting failures, and they are working from the frontline data their plants have today, in whatever state it is in.

Almost every conversation I have with a plant leader goes the same way. The leader is struggling to get data from frontline operators in a form structured well enough to support critical decisions. Those leaders know their frontline teams hold the insight, but the systems corporate put in place do not reach the operator, so supervisors and managers build workarounds with paper, Excel and whiteboards.

There are two main reasons operator knowledge does not get captured. First, ERP, CMMS, QMS and EHS software typically does not reach the operator and relies on a supervisor or manager to enter the data instead. Second, where that software does have an operator interface, the interface is not easy enough to use, which is what sends people back to the workaround.

With AI now reading data, formulating root cause and predicting future failures, frontline context matters more than ever. The missing context is never flagged, so a manager acting on what AI tells them is acting on incomplete information.

I expected to find leaders who knew their data was not ready. Instead, we found the opposite. 141 of the 167 leaders told us their data was trustworthy enough for AI. Then, a few questions later, 127 of those same 141 described at least one of the gaps that undermine that confidence.

I spent my first ten years in manufacturing as a licensed industrial mechanic, long before AI made data readiness a strategic necessity. This survey found the same four gaps I worked around back then: paper logs and Excel (we got Excel in 1992), systems that do not line up, the reason a thing happened staying in someone’s head, and no two people recording an issue the same way.

Every one of those gaps can be closed, and your frontline team is the key to closing them. Your operators know why your plant runs the way it does, and how it could run better. Give them tools easy enough to use mid-shift, and what they know reaches the record. That record is what your AI system needs to help your team make a bigger impact.

Steve McBrideChief Executive Officer and Founder, Weever

Key findings

Four gaps in the data

We asked 167 manufacturing leaders 28 questions about their frontline data. Four specific gaps came back, and any one of them can derail an AI deployment.

Gap one

Paper, spreadsheets and homegrown tools

94%

named at least one core frontline program still running primarily on paper, spreadsheets or homegrown tools

Programs still on paper, spreadsheets or homegrown tools

Quality

66%Safety

63%Autonomous maintenance

41%Sanitation

29%

Gap two

Disconnected records

81%

agree or strongly agree that disconnected systems limit how effective AI can be in their operations

Call their own systems connected80%

Yet say disconnected systems limit AI81%

Gap three

Missing human context

62%

say important operational knowledge stays only in the experience of operators or inspectors, frequently or very frequently

The “why” stays on the floor

15%

47%

31%

8%

Very frequentlyFrequentlyOccasionallyRarelyNever · 0%

Gap four

The same issue, recorded differently

51%

say the same issue is recorded differently depending on the person or shift logging it, frequently or very frequently

How often the same issue is recorded differently

16%

36%

35%

13%

1%

Very frequentlyFrequentlyOccasionallyRarelyNever

Q16, five-point frequency scale, n = 167.

Other key findings

AI is already here. Trust in the data is not.

A key finding of the survey: confidence in AI does not predict data readiness. More than 90% of leaders who described themselves as confident in AI deployment also reported at least one of the four gaps above. Most also said they have limited trust in the frontline data feeding their AI. Deployments are moving forward regardless, while 94% still capture data on paper.

Limited trust is already capping AI use73%

say limited trust in frontline data restricts how far their organization relies on AI.

Great extent · 19%Moderate extent · 54%

Confidence doesn’t mean readiness90%

of the 141 leaders who called their data trustworthy enough for AI report at least one of the three capture gaps anyway.

127 of 141 confident leaders describe a capture gap

AI in manufacturing is happening83%

are running or piloting AI currently.

Yet 94% still record at least one core program on paper.

Download the report to understand the key challenges to AI readiness, what it’s costing you and what to do about it.

Download the report

What to do about it

Make capture easy, then make it consistent

Buying a platform is the easy part. The work is making capture quick enough that an operator will do it mid-shift, consistent enough that two people log the same thing the same way, worth doing because something happens when they report, and connected so that when the same problem is reported four times you can tell it is one issue.

1

Make capture easier than paper

Closes gap one

Can an operator finish their check on a shared device, in their own language, mid-shift, without training?
Look for a platform you can configure to your lines and departments, instead of adapting to it.

2

Standardize how issues get recorded

Closes gap four

Define consistent fields: real assets, failure modes and more.
Attach photos at the point of capture to turn a thousand separate observations into a data set.
Use free text to capture the operator’s account of why it happened.

3

Close the loop

Closes gap three

Every issue needs an owner assigned when it is logged, a due date, and follow-through tracked to closure.
Review open items as a team at the start of the shift.
Without follow-through, the operators who report problems stop reporting them.

4

Connect it

Closes gap two

Your CMMS, QMS and MES stay.
Use one system to capture frontline data for quality, safety, sanitation and maintenance, so issues are visible to every department at once, and to your AI through one integration.
Keep it portable through an open API.

Assess your own plant