Manufacturers are changing how they buy industrial AI

Manufacturers are showing growing interest in small-scale AI deployments that can demonstrate a financial return before companies commit to wider investment. The approach is beginning to influence how industrial software is purchased, with customers seeking evidence of practical improvements rather than relying on the benefits projected at the beginning of a major technology program.

Mark Moffat, CEO of IFS, discussed the changing buying process during a press briefing at IFS Unleashed in Orlando, Florida. He told journalists that customers were increasingly interested in minimum viable products and limited implementations that could be expanded once the business had established confidence in the results.

“I think adoption comes from capturing people’s imagination,” he said. “When you can quickly demonstrate a return on investment by solving a problem that matters to the customer, particularly one that is attracting pressure from the board, adoption can happen very quickly.”

His observations raise questions about the established approach to industrial software investment, particularly where manufacturers commit substantial resources to enterprise-wide projects before they can evaluate the operational benefits. AI-assisted development offers opportunities to introduce capabilities more quickly, although manufacturers still need to establish whether the resulting improvements justify the investment.

A changing buying process

Moffat was asked whether industrial AI would follow earlier enterprise technology transitions, where predictions of rapid adoption had frequently proved optimistic. Changes such as the migration from on-premises applications to cloud platforms have often taken longer than suppliers anticipated, particularly in organizations with extensive existing systems and established operating procedures.

His response focused on allowing customers to experience the technology within their own businesses before deciding how extensively to deploy it. “We are seeing a change in the buying cycle, with growing demand for minimum viable products and solutions that can be scaled. Customers are increasingly starting with smaller deployments, sometimes at just a few sites, to establish confidence in the technology and demonstrate its ability to deliver meaningful improvements to their business.”

Starting with a restricted deployment gives manufacturers an opportunity to examine whether an application can address a specific operational requirement. It also allows the business to establish what information is needed, how employees will use the technology and whether improvements remain consistent during normal operations.

The decision to expand can then be informed by experience rather than assumptions made during the initial purchasing process. Manufacturers may find that an application performs well at one facility but requires additional development before it can be introduced elsewhere, particularly when equipment configurations, information structures and working practices differ.

Moffat also sees closer customer engagement influencing how software providers decide which applications to develop. Rather than relying exclusively on established product development processes, direct discussions with industrial businesses are revealing opportunities that might otherwise have been overlooked. During the briefing, he recalled a recent conversation with Tampa Electric about an aging application that the company needed to replace, but which carried a substantial price tag. After exploring what the application did, he began considering whether the same requirement existed elsewhere in the utilities industry and whether a more economical alternative could be developed.

“I went back to our team and asked them to investigate the opportunity and validate whether other customers faced the same problem,” he said. “We needed to establish whether it was something we should be developing and whether it fitted the industries and customers we serve. That is increasingly how we are making our investment decisions.”

Although the example came from the utilities sector, the commercial considerations are familiar to manufacturers managing extensive portfolios of operational applications. Aging systems can remain in use because replacement projects involve considerable expenditure, even where the software performs a relatively narrow function. The ability to develop more economical alternatives could make some of these projects commercially viable, provided suppliers can demonstrate sufficient demand and a sustainable business case.

Developing alternatives more economically could make some of those projects viable, but Moffat’s account also demonstrates why identifying a requirement does not automatically justify developing a new application. The supplier must establish whether the problem is sufficiently widespread and whether the proposed capability can be supported commercially.

Rethinking how industrial software is priced

The growing use of AI agents is also challenging the commercial model traditionally used to sell enterprise software. As intelligent applications take on activities previously performed by employees, charging according to the number of individual users becomes increasingly difficult to justify. IFS has responded by introducing a pricing structure linked more closely to the scale of its customers’ industrial operations.

“We have moved away from the traditional per-user pricing model towards something that better reflects the economic realities of our customers,” he explained. “I am an accountant by background, and I have always believed that costs should be aligned with business outcomes and revenue. We wanted a pricing structure that reflects how our customers operate and where they generate value.”

For manufacturers, software requirements are frequently determined by the scale and complexity of their operations rather than simply the number of employees accessing an application. A new production facility may require additional planning, maintenance and asset management capabilities without a corresponding increase in software users. Moffat argued that linking costs more closely to the number of plants or assets provides a better basis for pricing, particularly as digital workers become more commonplace within industrial businesses.

“If a customer operates several plants, that becomes the basis for pricing,” he said. “When they invest in another plant, there is an incremental cost associated with the software needed to support it. We are also moving towards a world with more AI agents performing work, which makes the traditional approach of charging for individual human users increasingly outdated.”

Moffat confirmed that IFS Intelligence would carry an additional charge calculated as a percentage of a customer’s existing software arrangements, with the amount varying according to factors such as business size and the number of plants or assets involved. He did not provide a standard percentage during the briefing but acknowledged that adopting the new capabilities would increase customers’ software expenditure. Manufacturers evaluating the financial return from AI will therefore need to consider these additional charges alongside the resources required to support wider deployment, particularly when moving from an initial implementation to multiple production facilities.

The move away from per-user licensing also raises broader questions about how enterprise software suppliers will charge for applications that undertake an increasing proportion of routine business activities. Pricing based on the number of employees accessing a system becomes less representative of its operational value when AI agents can initiate transactions, retrieve information and complete defined tasks independently. As these capabilities develop, manufacturers will need to assess whether alternative licensing arrangements provide a clearer relationship between software expenditure and the benefits delivered across their operations.

The future cost of implementation

AI could have an equally significant effect on the cost of implementing enterprise software, an area where manufacturers have traditionally faced substantial expenditure beyond the initial licensing agreement. Configuring applications, integrating existing systems and preparing employees for new working practices can require lengthy programs involving software suppliers, systems integrators and specialist consultants. Moffat expects intelligent development tools to reduce much of the technical effort involved, potentially changing both the economics of implementation and the services customers purchase.

“I spent more than 15 years implementing packaged applications in major enterprises, so I understand what is involved in delivering software to customers,” he said. “The costs of implementation are going to fall dramatically, and that could fundamentally change how service partners make money. Instead of spending so much time getting software to work, they will have greater opportunities to help customers transform their operations, manage organizational change and apply their industry expertise.”

For manufacturers, lower implementation costs could make technology projects commercially viable that might previously have been difficult to justify. Replacing an aging application or introducing new capabilities across several facilities could become more affordable if development and configuration require fewer specialist resources. However, the scale of any savings will depend on the complexity of existing systems and the work required to integrate new applications into established manufacturing processes.

Moffat also raised questions about whether companies should develop and maintain their own enterprise applications as AI makes software creation more accessible. He described discussions with German businesses that had built their own field service systems, including replacements for commercial software, but subsequently encountered problems maintaining integrations and responding to changes in the underlying technology. Although these applications initially appeared to offer greater control, the companies discovered that sustaining them required software engineering expertise outside their principal business activities.

The experience illustrates why lower development costs do not necessarily translate into lower lifetime costs. Manufacturers considering internally developed applications must account for ongoing maintenance, security, integration and technical support, particularly when software performs activities essential to production or customer service. Moffat explained that the businesses concerned had begun reconsidering commercially supported enterprise systems because they needed the reliability and continuing development that a specialist provider could offer.

The combination of faster deployment, changing licensing arrangements and lower implementation costs could alter how manufacturers evaluate enterprise technology investment. Instead of committing to lengthy programs with substantial upfront expenditure, companies may have more opportunities to test individual capabilities and establish their financial value before expanding them. However, that flexibility will only deliver a commercial advantage if manufacturers consider the continuing costs of operating and supporting the applications they introduce.

Moffat’s observations at IFS Unleashed suggest that customers are becoming more demanding about the relationship between technology expenditure and business results. Smaller initial deployments provide evidence for subsequent investment decisions, while changes to implementation and licensing models could reduce some of the financial barriers associated with enterprise software. Manufacturers will still need to assess whether the operational benefits justify the full cost of deployment, but they may increasingly be able to make that judgment before committing to a large-scale program.