Everyone Wants AI/Gen AI in Their Factories, but Manufacturers Need to Be Aware of the Hidden Costs Associated with AI/Gen AI

OpenAI recently began floating the idea of adding advertisements to ChatGPT in a bid to drive greater revenue as its Artificial Intelligence (AI) development costs mount. The shift in business model highlights a question that many have been asking in the manufacturing market: what is the actual cost of AI/Generative AI (Gen AI)?

Standalone AI/Gen AI solutions are far easier to identify a price for, as they tend to handle specialized tasks such as closed-loop process optimization or targeted data analytics, with these tasks having much clearer Return on Investment (ROI). However, many AI/Gen AI tools, primarily industrial copilots, are integrated into existing products and have far more ethereal productivity benefits. Many technology vendors are simply adding these in for free, despite massive development costs, to encourage user adoption and drive increased value for existing users of a given software solution, whether it be a Manufacturing Execution System (MES), Product Lifecycle Management (PLM), etc. The primary use cases for Gen AI indicated from respondents in ABI Research’s Industrial and Manufacturing Survey 1H 2024: Utilization of Generative AI (PT-3332) were:

• Helping operators get to the root cause of a production issue more quickly.
• Rapidly creating work instructions.
• Empowering individuals with limited coding skills to create sophisticated software solutions.

However, it is highly likely that, at some point, these “free additions” could result in notable sticker shock for manufacturers. The reality is that AI/Gen AI solutions are not getting cheaper to develop and run; in fact, costs are rising fast, all while the jury is still out regarding concrete ROI.

There are several notable costs associated with adoption new AI/Gen AI solutions that are not included in the upfront price:

• Collection and Storage of Clean Data: The cost of ensuring access to clean data for powering AI can be high. Fixing up data architecture that is currently unprepared for support AI is expensive and essential, or else manufacturers risk “garbage-in, garbage-out” AI.
• New Hardware: Manufacturers often require new hardware to run AI/Gen AI tools, especially if the solution is operating at the edge. This can result in the need for investment in new Graphics Processing Unit (GPU)-powered Instructions per Cycle (IPCs). Such a factor will never be costed into the software tool that manufacturers are purchasing, or even adopting for “free.”
• New Staff and Training: All new technology deployments have associated training costs to acclimatize workers with the new solution. In the case of AI/Gen AI tools, it might also require additional specialized staff or System Integrator (SI) contracts to manage them. Even companies with large Information Technology (IT) teams don’t necessarily have employees specialized in AI model tuning and management. The current labor market for this type of digital talent is incredibly tight, and therefore, expensive.
• Free Tools Not Remaining Free: Copilots and other new functionalities integrated into existing solutions that are currently embedded in the cost of the license might not remain that way. As manufacturers adopt and become reliant on these new tools, a likely price hike will follow, either for the solution as a whole or for access to the AI/Gen AI tool.

Technology vendors can no longer ride the hype of AI/Gen AI to justify blank check investment into new solutions. As manufacturers become more critical of the adoption of AI/Gen AI for their factories and its impact, vendors can take several actions to build transparency and trust in their AI/Gen AI solutions’ business models.

1. Show Concrete ROI: This will be essential to justify the increasing costs of these solutions for customers. This is particularly the case for industrial copilots, with wide use cases and ranging productivity benefits. Technology vendors will need to be clear about what use cases are being measured and how they are associated with a solution’s pricing. Vendors could consider offering different tiers of solutions (standard or premium) based on expected offerings’ ROI.
2. Create Transparent Road Maps: If there are plans to increase the costs of AI/Gen AI tools, technology vendors need to be open with customers so they can better plan for these changing Operational Expenditure (OPEX)/Capital Expenditure (CAPEX) costs.
3. Build Solutions, Rather Than Selling Tools: Technology vendors should look to provide AI/Gen AI solutions that account for the hidden costs associated with deployment. For example, if the solution requires additional GPU-powered hardware, the technology vendor should partner with an industrial hardware provider for their Go-to-Market (GTM) strategy, clearly pricing this element into the overall cost of adoption.

The onus is not just on vendors to build transparency, but manufacturers also need to ensure they are doing their due diligence on costs when adopting new tools. Companies should not simply adopt new AI/Gen AI tools because they believe they need them to remain competitive and take their cost at face value. The technology being adopted must align specifically with challenges faced on the factory floor and drive measurable impact.