AI in Manufacturing Is Not a Software Project

Interview by Nishkam Batta, Editor-in-Chief, HonestAI Magazine

Manufacturers are under growing pressure to adopt artificial intelligence, but simply investing in AI tools does not mean a company is becoming more intelligent.

That distinction was at the center of a recent conversation between Nishkam Batta, Editor-in-Chief of HonestAI Magazine, and Cosmin Badea, a manufacturing consultant and mechanical engineer with more than 25 years of experience across electronics manufacturing, supply chain, procurement, quality, engineering, program management and factory leadership.

Having worked with major electronics manufacturing organizations and across OEM, EMS and ODM environments, Badea has seen manufacturing evolve across countries, technologies and operating models. Today, his consulting work focuses on helping companies improve profitability, strengthen manufacturing partnerships, digitalize operations and identify where AI can create meaningful value.

His central message to manufacturers is straightforward: AI should not begin with technology. It should begin with a business problem.

AI Is Becoming a Trend, but That Can Be Dangerous

Nishkam opened the discussion with a problem he has encountered while speaking with manufacturing and enterprise leaders: companies are spending heavily on generative AI subscriptions and tools without necessarily knowing what they want those investments to accomplish.

Badea sees the same pattern.

AI, he explained, has become a powerful acronym and, in some organizations, a source of competitive pressure. Leaders see competitors announcing AI projects and conclude that they need an AI initiative of their own.

That can lead companies to approach AI primarily as a software investment.

But the more important questions are different: Where does the business have a problem? Where can AI influence that problem? And how will solving it create measurable value?

For Badea, buying technology before answering those questions puts the project in the wrong order.

Before AI, Get the Foundation Right

Successful AI implementation requires much more than selecting a platform.

Badea highlighted several factors manufacturers need to consider: budget, internal expertise, governance, available resources, data quality and organizational culture.

Data is particularly important.

Manufacturing operations may generate thousands of data points, but having large quantities of information is not the same as having reliable information. Poor-quality or fragmented data limits what an AI system can deliver.

The familiar principle still applies: garbage in, garbage out.

There is also a broader operational lesson. AI cannot automatically repair a company with fundamentally broken processes.

If data is disorganized, processes are inconsistent and employees are not adequately trained, adding AI can amplify those weaknesses rather than eliminate them. The foundation therefore needs to come first.

AI Should Help People Make Better Decisions

One concern continues to follow almost every conversation about AI: Will it replace jobs?

Badea does not see that as the most useful way to think about AI in manufacturing.

He sees AI primarily as a mechanism for bringing together information from different systems, processing it faster and helping people make more informed decisions.

Consider the information distributed across quality, procurement, supply chain, engineering, R&D, sales, marketing and HR. Traditionally, these departments often optimize their own KPIs independently.

That creates silos.

An improvement in one department does not necessarily translate into better overall business performance.

Badea believes AI has the potential to connect information across those functions and help leaders understand the wider consequences of individual operational decisions.

That is why he argues that AI should not be treated simply as another automation tool.

Its greater value may lie in improving the quality and speed of decision-making across the organization.

The CEO Must Own the AI Vision

Who, then, should be responsible for an organization's AI strategy?

For Badea, the answer is clear: the CEO.

The CEO sits at the intersection of customer expectations, market conditions, internal operations, budgets and organizational resources. That makes leadership responsible for defining where the company wants AI to take the business.

But CEO ownership does not mean CEO-only execution.

Badea recommends establishing a cross-functional AI committee that can oversee investments, milestones and progress. Depending on the project, that group might include data analysts, data scientists, compliance leaders, quality leaders and managers from the operational functions affected by the initiative.

The objective is to prevent AI from becoming an isolated IT experiment.

As Badea put it, the CEO must turn the vision into a mission because a vision without action remains only a dream.

Don't Ask "Where Can We Use AI?" Start With the Bottleneck

One of the most practical parts of the conversation centered on how manufacturers should identify their first AI use case.

Badea connected the question to the Theory of Constraints: first identify what is limiting the business.

That bottleneck does not have to be a machine on a production line.

It could be falling customer orders, supplier performance, excessive costs, inventory, poor visibility or another constraint affecting profitability.

Rather than creating a long list of possible AI applications, leadership can identify its top business priorities and investigate whether AI is appropriate for solving them.

This changes the starting question from:

"Where can we implement AI?"

to:

"What is preventing us from performing better, and can AI help solve it?"

That distinction can prevent manufacturers from investing in technology simply because it is available.

Supplier Selection Shows What AI Can Actually Do

Badea offered supplier evaluation as an example.

A manufacturer may have years of information distributed across numerous suppliers. Traditionally, purchasing decisions can become heavily influenced by unit price or by KPIs owned by individual departments.

AI can potentially help consolidate multiple dimensions of supplier performance, including total cost of ownership, quality performance, delivery history and other operational information.

Instead of asking only which supplier offers the lowest price, decision-makers can evaluate which supplier has historically created the strongest overall business outcome.

This is where cross-functional visibility matters.

Optimizing purchasing price while creating inventory, quality or delivery problems elsewhere may save money on one KPI while reducing profitability at the company level.

Start Small Before Scaling AI

Badea cautioned manufacturers against attempting a massive AI transformation immediately.

His recommendation is incremental implementation.

Define a problem. Establish a budget. Build a prototype. Measure what happens. Iterate.

If repeated iterations fail to demonstrate progress, leadership needs to question whether the company has selected the wrong solution, partner or even the wrong problem.

Most importantly, the evaluation should eventually connect back to business value.

If an AI initiative does not improve profitability, throughput, inventory, operating expenses or another strategically important outcome, manufacturers should ask what the project is actually accomplishing.

Simply being able to say that the company has AI projects is not enough.

How Should Manufacturers Evaluate an AI Partner?

Batta raised another difficult issue: manufacturers know how to evaluate traditional suppliers because they have decades of experience doing so.

An automotive manufacturer, for example, understands how to assess a component supplier.

AI is different.

Many manufacturers do not yet have enough internal expertise to fully assess the technical quality of an AI solution before implementation.

Badea recommends applying the discipline of traditional supplier due diligence while adapting it to AI.

Manufacturers should investigate the provider's previous projects, results, case studies and references. They should understand whether projects were successfully completed rather than merely launched.

Commercial terms matter, but so do service levels, maintenance capabilities, commitment and the provider's ability to support the system as business conditions change.

That last point is particularly important because an AI implementation should not be treated as a one-time installation.

Production conditions change. Designs change. Supply chains change. Logistics change. Business conditions change.

The AI system and the data supporting it must therefore be monitored and maintained over time.

AI Needs Its Own KPIs

Traditional manufacturing KPIs are not going away.

Companies will continue measuring on-time delivery, yield, OEE and other operational indicators. At the business level, they will continue watching throughput, inventory, cash, margin and profitability.

But AI projects require another layer of measurement.

Manufacturers need to define what a specific AI initiative is expected to achieve, by when and with what resources.

For example, if AI is being introduced to improve supplier management, the company should define the expected outcome and then determine whether the project is actually influencing inventory, decision speed, supplier performance or another targeted metric.

The purpose is not to measure AI for the sake of AI. It is to connect AI performance with operational and ultimately business performance.

AI Should Follow the Business Problem, Not the Hype

The conversation between Batta and Badea ultimately returned to a simple principle.

Manufacturers do not need AI everywhere.

They need clarity about where their businesses are constrained, what information is missing from important decisions and which problems are worth solving.

From there, AI becomes one possible method of creating improvement rather than the objective itself.

The path Badea describes is deliberately methodical: identify the constraint, establish the business objective, assess data and organizational readiness, involve leadership, select the right expertise, start with a manageable project, measure results and scale only when value becomes visible.

For manufacturers trying to understand what AI means beyond the headlines, that may be the more useful definition of AI transformation.

It is not about how many AI tools a company can deploy. It is about whether those tools help the company make better decisions and build a better business.

About Cosmin Badea

About the Interviewer