The answer to that question depends on the scale and objectives of a company’s AI project. Building and training a proprietary large language AI model is a very expensive endeavor, with steep costs in hardware, energy, and engineering talent.

The good news is that, for most businesses, there’s no need to build your own model. A broad industry of vendors already exists that allows businesses to leverage the latest AI innovations from the likes of Microsoft, Google, Amazon, and various AI startups. Many of the offerings also include open-source AI models, bringing the capabilities of these powerful generative AI tools to organizations that might not have the technical resources or skills to deploy it on their own.

“The costs of the AI tools themselves, they really aren’t genuinely prohibitive,” says Phil Gilchrist, chief transformation officer of AI and sustainable materials at TE Connectivity, a maker of EV sensors and other electronic components. “What’s much more challenging is to recognize that we’re going to live in a world that will be an AI world going forward, and we have to recognize that we need to organize ourselves.”

For most companies looking to deploy generative AI, there are three main pricing models to consider: consumption-based models that are tied to the compute power needed;  subscription-based fees in which a company pays a regular monthly fee for each user; and outcome-based models that set pricing when tasks are deemed a success by a customer.

Subscription-based models include offerings like Microsoft’s Copilot AI assistant, which charges $30 per month for enterprise users, enabling them to do things like create data visualizations in Excel and more quickly clear out an overstuffed Outlook inbox. A business that wants to roll out a customized chatbot to handle customer support, however, will probably be paying a fee that varies based on how often the bot is used.

Boston Consulting Group’s John Pineda advises clients to think through if the AI being deployed is meant to augment the work humans do or to fully replace tasks. If tasks are expected to be completely automated, “it starts to make sense to price either consumption-based or more outcome-based,” says Pineda. Conversely, for use cases that boost the effectiveness of how people work and assist workflows, a subscription, user-based model might be better.

But Pineda also advises that businesses shouldn’t be so focused on the costs of AI that it would impede innovation. “Experimentation drives value,” he says. “Get people using it, trying it, and testing it, [to] come up with their own ideas of what they can do with the technology.”

One size does not fit all

As a business recognizes the value of generative AI throughout an organization, various teams and departments are likely to cook up their own projects, with each project calling for different combinations of AI models—and different pricing plans.

“Businesses need to adopt the right model for the right task,” notes Ritika Gunnar, general manager of data and AI at tech giant IBM. She says 80% of enterprises are utilizing a multimodal approach today, leaning on open-source and commercial AI offerings. As businesses continue to pilot and put AI into production, she estimates they’ll continue to utilize different models to figure out what’s best. 

The first step for any business is to determine whether AI is actually the best tool for the job. “Focus on the right use case,” says Gunnar. “Because the AI and the generative AI capabilities are really an ingredient to be able to help accelerate the outcomes that the business is trying to focus on.”

For TE Connectivity, a big upcoming project involves sorting through some 200 million documents spread across dozens of databases. Gilchrist says decisions will need to be made to determine which AI models will help extract that information to boost TE Connectivity’s competitive advantage, training employees on how to deploy and use the tech, and getting workers comfortable with using and trusting AI.

TE Connectivity always pilots new technology, whether it may be an AI chatbot created to help customers look for information on the company’s website or an internally designed tool for engineers. 

“We’ll try it first, and we’ll prove to ourselves [it] really is delivering the value … that is advertised,” says Gilchrist. “We’re very much a ‘prove it to me’ company.” 

Abe Kuruvilla, chief technology officer at software maker ACI Worldwide, says many business leaders today are looking at the power of AI and thinking about productivity in terms of reducing costs, including headcount. But he is more focused on enabling ACI’s workforce to handle tasks at a faster pace . 

“For me, I’m really still trying to figure out the monetization strategy on speed,” says Kuruvilla. “What is that value proposition for the client, versus the typical legacy models of pricing.”

ACI Worldwide and other companies, he says, are still trying to lay out the total cost of building a product on the cloud, as well as using AI. “The challenge becomes, ‘What’s the market willing to pay for that speed?’” asks Kuruvilla.

If anything is clear so far, it’s that the breadth of AI services is only likely to grow. Over the course of a decade, spending on generative AI software, services, and other products is expected to soar to $1.3 trillion, Bloomberg Intelligence estimates.