Manufacturing: AI, the New Competitive Race, and What It Means for Owners

Expect a sophisticated buyer to ask whether the plant is becoming more competitive—and what it will cost to keep it that way. Be ready to connect changes on the floor and in the office to your numbers.

• Is revenue growing, flat or declining? Are gross and EBITDA margins improving or deteriorating?
• Is labor cost growing faster than revenue? What is happening to revenue per employee?
• Are quotes winning on price, speed or capability? Are competitors quoting and delivering faster?
• Are lead times, scrap, rework and capacity utilization improving?
• Are customers demanding faster responses or more customization without paying more?
• Are competitors producing more with fewer people?
• What is already automated? What has management invested in, what worked, what didn’t, and what comes next?

A buyer should distinguish results already earned from savings that still require equipment, people, downtime and money.

Start with your own numbers

The order book is growing, but overtime and work in progress are growing faster. Or the machines are busy, yet quotes are taking longer to win. Neither proves an AI problem. Both are reasons to examine where profit and cash are getting stuck.

• Are revenue, gross margin and EBITDA margin improving? How much is price or product mix rather than operating improvement?
• What is happening to revenue per employee, labor cost as a share of revenue and headcount growth?
• Are we using available capacity better? Are scrap, rework and on-time delivery improving?
• Are inventory turns improving—or is more working capital tied up in materials and unfinished orders?
• Is quote-to-order conversion changing? Is the sales cycle lengthening?
• Are we becoming more dependent on a few customers?

Are you seeing the early signs of AI competition?

• Are competitors answering RFQs and quoting faster, with more engineering work already done?
• Are they producing more with fewer people or reducing administrative headcount?
• Do customers want faster answers, shorter lead times or more customization?
• Are customers bringing more complete specifications or designs, changing the work they need from us?
• Are they asking for digital interfaces, instant configuration or self-service?
• Are competitors using AI or automation to include something we still provide manually?

Are our competitors getting better faster than we are?

A manufacturer can remain profitable and grow while its competitive position deteriorates. Existing programs and customer relationships may sustain this year’s results while the next RFQ favors someone else. Use customer feedback and lost-order reasons; do not infer a competitor’s technology from its price alone.

Different businesses face different races

A custom fabricator waiting on an estimator has a different constraint from a chemical plant losing yield. Our economic grouping below is a way to read the evidence—not a set of rigid industry labels. A company can belong to several groups.

• Engineered-to-order and design-heavy products: engineering hours, design changes and the time from requirement to release drive economics. Faster analysis matters only if the design remains manufacturable and accountable.
• High-mix, low-volume specialists: quoting, setup, scheduling and scarce know-how matter disproportionately. Lower preparation costs can make smaller orders worthwhile; a robot dedicated to one part may be a poor fit.
• Repetitive discrete production: cycle time, first-pass quality and repeatable labor content offer opportunities. Automation must run enough hours, at the right mix, to justify its cost.
• Continuous or batch process manufacturers: yield, energy, consistency and downtime often matter more than office headcount. Existing sensors and controls may offer a starting point; process changes carry operating risk.
• Contract manufacturers: customer-owned designs and competitive rebids can send productivity savings back to customers. Quoting discipline, program mix, utilization and purchasing determine how much benefit the manufacturer retains.
• Proprietary component and product manufacturers: selection, configuration, applications engineering and after-sales support affect the value of the product. Defensible designs and approvals matter; a catalog and slow manual answers alone may not.

The contrast is visible in the named factory cases below. BCG’s 2026 research also finds that benefits vary with labor, energy, materials, logistics and automation potential. These are conditional opportunities, not a universal savings rate. BCG, May 2026 manufacturing research.

Engineering and design

A customer changes a specification. AI-assisted engineering can help explore designs, support simulation and analysis, and prepare documentation or change-order material. Engineers still have to verify tolerances, safety and manufacturability. Siemens’ 2025 Altair acquisition expands the available simulation, data and AI tools; it does not establish a standard reduction in engineering hours. Siemens, March 2025 Altair acquisition.

How much engineering time does it take us to get from customer requirement to manufacturable product? The opportunity may be more products with the same team, rather than fewer engineers.

Quoting and estimating

An RFQ arrives with drawings, specifications and exceptions. Someone interprets it, checks materials and tolerances, estimates labor and machine time, and prepares a price. Software can automate repeatable geometry checks and calculations; AI can help extract and organize less structured requests. The remaining judgment includes missing information, difficult setups and commercial risk.

Protolabs describes returning automated quotes with manufacturability feedback in hours. This is a competitive service expectation already in the market, not proof that a general-purpose AI tool can price any job correctly. Protolabs, July 2025.

How much of the work between an RFQ and a quote can now be automated? Faster bad quotes simply win unprofitable work sooner.

Production planning

An urgent order displaces tomorrow’s schedule. Better planning can compare machine capacity, materials, sequencing, labor and bottlenecks before making a promise. Forecasting helps anticipate demand; a digital twin can simulate alternatives. Neither creates missing material or capacity. Scheduling and process improvement feature among the priorities in McKinsey’s 2025 COO research. McKinsey, December 2025 manufacturing COO survey.

Quality

A defect is cheaper to catch before more work is added. Machine vision can inspect images; process monitoring can flag drift; analysis can help investigate root causes and assemble quality records. These tools need representative examples and reliable measurements. NIST warns that idealized or unrepresentative data can fail to reflect actual factory conditions. NIST, February 2025 industrial AI guidance.

Maintenance

A warning about a failing pump has value when the team can act before the line stops. Monitoring and predictive tools can support maintenance timing and spare-parts planning. The economics depend on the cost of downtime, useful warning time and the resources to make the repair—not the number of alerts. The Sachsenmilch example below illustrates that distinction.

Purchasing and inventory

Better demand and production information can improve purchase timing, supplier follow-up and inventory allocation. The commercial opportunity is fewer shortages, expedites and unnecessary buffers. Inventory reduction releases cash; it is not automatically an equivalent increase in profit. Poor forecasts or unreliable suppliers can still make apparently lean inventory expensive.

Customer service and sales

Product selection, technical questions, order status, documentation and follow-up may require fewer manual handoffs when systems and product data agree. Configuration and self-service can reduce the cost of serving smaller orders. Answers about fit, tolerances and delivery still need a dependable source; a quick incorrect answer can create a costly return.

The factory itself

AI is not another word for robotics. AI analyzes or generates information. Automation executes a process; robots perform physical tasks; machine vision interprets images. A digital twin models a product or operation. Some applications combine these, but conventional controls, tooling, handling equipment and physical process changes remain separate investments.

Better inspection software may work with existing cameras. Robotic handling may also require fixtures, guarding, layout changes and commissioning. Buy the capability that improves the constraint—not every technology in the category.

Consider two shops with similar machines. One can repeatedly hold a difficult tolerance, document it and deliver a qualified part quickly. The other owns the equipment but relies on one person to make it work. Their assets may look similar; their transferable capability does not.

• Proven processes and know-how: proprietary methods, specialized tooling and equipment matter when they produce repeatable results competitors cannot readily match.
• Engineering and designs: application knowledge, defensible intellectual property and the ability to turn a requirement into a reliable product can support pricing and retention.
• Customer qualification and trust: certifications, approved processes, traceability and reliable delivery can make replacement costly. A certificate by itself does not ensure superior economics.
• Useful proprietary data: linked records of materials, settings, failures and results can help the company improve. A large archive without usable context is not the same asset.
• Local availability and short lead times: proximity can be valuable when it reduces delivery risk or enables rapid changes. It must compete on total delivered cost and service.

BCG’s 2026 analysis suggests upgraded production can strengthen the case for some higher-cost locations, with different outcomes by sector and logistics exposure. This is scenario analysis, not proof that domestic production or physical assets are automatically protected. BCG, May 2026 manufacturing research.

Which parts of our manufacturing advantage are actually difficult for a better-capitalized competitor to reproduce? Also ask what would remain if a key employee left.

Evidence reviewed through October 1, 2026. These are operating examples, not promised returns for your plant. Where a source omits project cost or an independent evaluation, that limitation matters. Lighthouse sites are selected leaders, not a representative sample of middle-market factories.

Protolabs: speed starts before machining

Its July 2025 account describes automated quoting and manufacturability checks connected to production, plus robotic preparation of machining blocks. The objective is faster, more consistent delivery with less manual work. This requires proprietary software, captured manufacturing rules and physical equipment. Protolabs reports higher productivity from block preparation but gives no quantified project return or cost. The lesson: a fast quote is valuable when the operation can fulfill it. This combines established digital automation with newer capabilities; it is not all generative AI. Protolabs, July 2025.

Schneider Electric: improve a complex order’s journey

At El Paso, growing data-center demand strained an engineer-to-order operation. Schneider reports that connected systems, analytics and AI aligning engineering, production and supply chain increased on-time delivery from 61% to 97%, reduced lead times by up to 35% and cleared $43 million in backorders. Its June 2026 release does not disclose the project’s full cost. These are company-reported outcomes of a combined program—not an isolated AI effect. For a custom manufacturer, coordination can matter as much as machine speed. Schneider Electric, June 2026.

Rockwell: high mix does not rule out improvement

Its Singapore plant faced frequent changeovers and dependence on worker knowledge. The June 2026 Lighthouse account describes more than 50 digital and AI applications, including flexible automation, quality control and maintenance. Reported results include 43% more units per person-hour and 35% fewer defects; project cost is not disclosed. The lesson is flexibility and repeatability, not a universal labor-cut target. World Economic Forum, June 2026 Lighthouse cases.

DCM Shriram: the biggest cost may be energy

At its Gujarat caustic-soda site, margin pressure centered on energy-intensive production. The Lighthouse account reports 45 applications, including AI process control and maintenance support, with power costs down 32% and an 11-percentage-point EBITDA improvement. Full investment cost and AI’s separate contribution are not disclosed. This reported site result points a process manufacturer toward its dominant cost driver, not an office-headcount benchmark. World Economic Forum, June 2026 Lighthouse cases.

Sachsenmilch: one avoided failure can matter

A dairy operating continuously added vibration monitoring, integrated existing machine data and trained its maintenance team with Siemens. In June 2025, its technical leader reported that early detection of a failing pump saved a low six-figure euro amount and paid for the pilot. The full pilot cost and an independent audit were not provided. Further SAP maintenance integration was planned. The lesson is a specific downtime problem, supported by sensors, expertise and a team able to act. Siemens / Sachsenmilch, June 2025 (German).

The wider evidence is encouraging, but uneven

Deloitte’s 2025 publication surveyed 600 large manufacturers in August–September 2024. Respondents reported production-output improvements of 10–20% from smart manufacturing initiatives. This is self-reported, covers more than AI and does not establish what a smaller plant should expect. McKinsey’s late-2025 research similarly finds many large manufacturers still exploring or using AI in limited areas. Strong individual cases do not mean adoption or returns are uniform. Deloitte, 2025 Smart Manufacturing Survey; McKinsey, December 2025 manufacturing COO survey.

A shop wins more small orders because quoting and setup take less time. A process plant sells more output because fewer batches fail. Both may improve, but the cash comes from different places. The following are economic possibilities to evaluate against this company’s constraints—not forecasts of savings.

• Direct labor: output can grow without proportional headcount growth. Savings may first appear as less overtime, fewer vacancies or avoided hiring. They require useful work and demand for the output.
• Engineering and quoting: less repetitive preparation can let the same team quote more jobs or handle customization. Better conversion and realized job margin matter more than quote volume.
• Quality and maintenance: less scrap, rework, warranty expense and downtime can increase sellable output. Sensors, inspection systems, maintenance work and technical support cost money too.
• Purchasing and inventory: better coordination can reduce expedites, excess stock and work in progress. Lower working capital improves cash; inadequate buffers can damage delivery.
• Sales and customer service: faster configuration, documentation and order answers can lower the cost to serve. Customers may also expect these services without an additional charge.
• Software, data and infrastructure: licenses, integration, computing, cybersecurity and ongoing support become recurring costs.
• Automation, robotics and capital equipment: tooling, installation, safety work, commissioning and training add to the purchase price. Underused equipment can reduce returns despite a faster cycle time.

For a high-mix specialist, reducing engineering and preparation time may make smaller orders profitable. For a repetitive producer, utilization may decide whether automation earns a return. In a process plant, yield or energy may dominate. A contract manufacturer may have to pass part of the benefit to its customer at the next bid.

Capital requirements can rise. McKinsey’s survey of 101 operating executives at manufacturers with at least $1 billion in revenue found plans for materially greater digital and AI spending. Those are large-company intentions, not a middle-market spending target. McKinsey, December 2025 manufacturing COO survey.

The test is cash earned or released after the full cost of change. Faster machines do not help if the bottleneck moves downstream or the extra output cannot be sold.

Can this company fund and absorb the changes required to remain competitive? A promising project is only part of the answer. The company must still deliver current orders while the work happens.

• Capital: funding for equipment, software, integration, commissioning, training and working capital—with room for delays and follow-on investment.
• People: engineering and automation judgment, reliable outside expertise where needed, and employees who understand how the process actually behaves.
• Data and systems: dependable drawings, routings, costs, machine and quality records, connected where the business needs them. More data is not automatically better data.
• Management attention: authority to set priorities, resolve conflicts between production and change, and stop initiatives that do not improve economics.
• Learning capacity: employee training, time to integrate new tools and an ability to run and fund further experiments.

Deloitte’s survey identifies talent and operational disruption as material constraints. NIST emphasizes data that represents real operating conditions. Both point toward a capability requirement beyond buying technology. Deloitte, 2025 Smart Manufacturing Survey; NIST, February 2025 industrial AI guidance.

The company does not need to own every skill. It does need enough internal judgment to choose partners, assess results and retain responsibility.

The risk isn’t simply spending too little. It is spending enough to disrupt the business without spending enough to finish the job.

• Under-investing while faster competitors raise the customer’s expectations.
• Buying technology without fixing the process—or automating a bad process more quickly.
• Overestimating labor savings when people must still cover exceptions, setup and service.
• Failing to achieve the volume or utilization needed to pay for equipment.
• Choosing tools that become obsolete, do not integrate or cannot support the real product mix.
• Disrupting production, quality or delivery while systems and equipment change.
• Failing to recruit the required people, exhausting management or running out of capital halfway through.

These risks should affect the size and timing of the owner’s commitment. If the first attempt disappoints, can the company fund the correction while protecting customers and payroll? A vendor demonstration is not evidence of a return across your shifts, materials and order mix.

Two plants with the same EBITDA may need very different amounts of cash after closing. One has proven demand and a clear next productivity investment. The other needs substantial modernization simply to keep its customers.

A path to profitable improvement and a requirement for defensive catch-up spending are different investment cases. A buyer should separate maintenance capital, competitive catch-up and optional growth investment.

• How automated is the operation today, and how much of its economics depends on labor?
• Where can labor be redeployed or avoided—and what spending, disruption and utilization does that assume?
• What automation has already been attempted? What was the realized return, including failures?
• What is the next logical investment, and how much capital is still required?
• How dependent is the company on a few skilled employees? Can its manufacturing know-how transfer?
• What protects the process or customer relationship? What can competitors now do more cheaply, and what can this company do more cheaply?

A buyer’s model should connect the opportunity to jobs, machines, customers and cash. An AI label does not establish an improvement opportunity—or prove the company is at risk.

Independence may make sense when the required investment is manageable, the company has a credible team and the owner wants to lead the next phase.

• Do we have the capital and access to the right people?
• Do we have the management bandwidth to keep operating while changing?
• Can we keep investing as customer expectations and technology move?
• Can we tolerate failed experiments and still finance the next attempt?

Can we make the investments required to remain competitive without a partner? Compare a realistic independent plan with the resources another party could actually commit. The question includes the owner’s appetite for risk, time and continued responsibility.

• Respond independently when the economics, resources and appetite are there.
• Bring in outside expertise when the missing capability is specific and management can direct the work.
• Bring in capital when the plan and team are sound but funding limits the response.
• Acquire capabilities when a team, process, product or facility fills a gap the company can integrate.
• Partner when another business brings engineering, automation, technology or market access.
• Join a larger platform when shared people, purchasing, systems or investment capacity improve the opportunity.
• Sell when another owner is better positioned to finance and execute the next phase.

Selling is not failure. An owner may have built a very good manufacturing company and conclude that its employees and customers would benefit from capital, technology, people or scale another owner can provide. The choice depends on the actual gap and the terms—not on a presumption that every manufacturer should stay independent or sell.

Capabilities exist, but they vary by buyer

Blackstone’s July 2025 Copeland account describes support for cost reduction, manufacturing expansion, pricing and employee engagement, alongside increased investment. This is a sponsor-reported example of operating support, not independently measured AI or automation returns. It does not tell you what a different portfolio company would receive. Blackstone, July 2025 Copeland account.

On the strategic side, Schneider’s own El Paso results show manufacturing coordination experience. Siemens’ completed Altair acquisition shows investment in simulation and industrial AI capabilities. Neither proves that a particular buyer will assign its best people or sufficient capital to your business. An equipment supplier’s expertise may also be available through a commercial partnership without selling the company.

Ask what this buyer will commit to this company

• Which operating, engineering, automation and AI specialists will actually be available?
• Have they improved a comparable manufacturing operation? What did it cost, how long did it take and what return was realized?
• What capital will be available after closing, beyond the acquisition financing?
• How quickly can investment decisions be made? How much autonomy will management retain?
• What happens when an initiative fails or needs another round of funding?
• What can they bring that the company cannot reasonably obtain on its own?

Who is actually better positioned to take this manufacturing company through the next phase? Ask for examples from comparable plants and speak with the people who ran them.

With an earnout, rollover or retained equity, the seller is also betting on the buyer’s execution. Future budgets, production disruption and management authority can affect the seller’s outcome. Understand who makes those decisions and how the transaction terms address them.