Enterprise IT shops more likely to buy GenAI than build it

GenAI power requirements, the cost of computing and storage, and the high salaries demanded by AI specialists make it unlikely enterprises will take a do-it-yourself approach.

Cloud providers and software makers will play an important role in enterprise adoption of generative AI as customers use their products and services to embed the potentially transformative technology into operations.

Experts agree that enterprises will find it much more cost-effective to use GenAI services rather than build the capabilities themselves. What makes a do-it-yourself approach unlikely are the power requirements of running the LLMs underpinning GenAI; the cost of necessary computing and storage resources; and the high salaries demanded by data scientists, AI developers and system architects.

"I don't think people will consume generative AI as a discrete product or a discrete function," said Tim Crawford, CIO and strategic adviser of advisory firm AVOA. "I think they will consume GenAI as an embedded function within core higher-level functions that they already use or will be using."

CIOs are under pressure from CEOs and boards to deploy GenAI quickly. The pressure makes it even more likely that they'll turn to trusted vendors while covering the basics for any technology deployment: security, data protection, available talent and ROI. The last item is not yet clear with GenAI, experts said.

"[Enterprises] are trying to build the use cases," IDC analyst Jennifer Cooke said. "They're looking for their technology vendors to help prove the return on investment."

Isaac Sacolick, founder and president, StarCIOIsaac Sacolick

Cloud providers and the most prominent software makers, including Microsoft, Oracle and Salesforce, claim to provide the guardrails and data protection tools to satisfy corporate policies, protect sensitive data and appease government regulators. The vendors also already have access to customers' data.

"It's just low-hanging fruit [for tech buyers]," said Isaac Sacolick, founder and president of StarCIO and a CIO adviser. "It's buy versus build."

Software makers using LLMs

LLMs providers include Cohere, a partner of Salesforce and Oracle, and OpenAI, which Microsoft uses in its public cloud Azure and the Copilot assistant in the Office 365 productivity suite. Microsoft has invested $13 billion in OpenAI.

The billions of dollars of revenue generated each quarter by the largest enterprise vendors place them in a position to innovate on GenAI services much faster than most enterprises could, experts said.

For example, the retrieval-augmented generation (RAG) concept is a cutting-edge feature available through Microsoft Azure and AWS. The architecture creates an information retrieval system that controls the data used to generate a response from an LLM, ensuring that corporate data stays out of the public domain.

Microsoft offers RAG in preview with OpenAI. AWS offers RAG in its Amazon SageMaker tool suite for building, training and deploying LLMs.

In the future, cloud providers will likely offer enterprises the option of choosing from various LLMs and provide recommendations on the best ones for specific tasks, Crawford said. "That's the next stage I expect to see."

AI pricing

[Enterprises] are trying to build the use cases. They're looking for their technology vendors to help prove the return on investment.
Jennifer CookeIDC analyst

Based on conversations with IT executives, Crawford doubts many enterprises will agree to pay a monthly fee per user for GenAI, similar to Microsoft's $30 monthly subscription for Copilot. Instead, enterprises will expect vendors to include GenAI services in the price of the business software.

In its 2023 State of Manufacturing survey of 500 manufacturing professionals, software maker Parsec found that more than half of the respondents believed AI and machine learning should be standard features in enterprise software.

A single-price model will make more sense over time as GenAI is incorporated in most software functions, making it difficult for vendors to separate the two.

That model is already starting to play out in the market. For example, SAP introduced low-code/no-code Gen AI tools in which some features will be available at no cost, others within a premium tier. The tools, built specifically for SAP applications, include SAP Joule, a GenAI developer assistant.

Room for small vendors

Experts expect smaller business software makers to fill GenAI market niches with specific features covering, for example, logistics, sustainability and human resources. Early examples include SAP partnering with data and AI companies Collibra, Confluent, Databricks and DataRobot to help build its new Datasphere data management portfolio.

"I would fully expect to see more of those kinds of companies doing really interesting things with data," Crawford said. "A small company has the ability to leverage cloud and [its] tools to jumpstart into the space and be able to do some novel things that, otherwise, you couldn't do."

Revenue projections indicate a market large enough for many vendor types. IDC forecasts enterprise spending on GenAI software, infrastructure hardware, and IT and business services to increase 73.3% annually to $143 billion in 2027.

Advice to CIOs

For CIOs ready to introduce GenAI into the work environment, advisers recommend letting employees experiment with the technology with guardrails in place and a well-defined purpose. A company doesn't want to impose blunt restrictions that prevent finding value in the technology.

Tim Crawford, CIO and strategic advisor, AVOA; host, CIO In the Know podcastTim Crawford

"I think it's important to help educate people and have them experiment and learn what they can and can't do," Crawford said.

Companies must also teach employees how to use prompts to get the most valuable answers to questions.

"Learning how to prompt as a skillset and a new way of working is the other edge of it," Sacolick said.

Employees also need a training data set and instructions on using it effectively.

"Don't give people the license to drive before they start understanding what they're driving," Sacolick said.

Companies should start using GenAI for tasks with the best chance of an immediate return. For most companies, that begins with improving customer service through better chatbots or consolidating information in searches to help service reps find more complete information on products and customers.

Finally, CIOs must look honestly at the organization and determine how best it can experiment with GenAI based on the enterprise's organizational capacity, culture and capabilities. Those factors are important because, ultimately, CIOs are dealing with people.

Getting people onboard as well as letting cloud providers and software makers do most of the work for GenAI will significantly improve the chances of getting the most out of the technology, experts said.

"At the end of the day, it's the business outcomes that matter the most," Crawford said.

Antone Gonsalves is an editor at large for TechTarget Editorial, reporting on industry trends critical to enterprise tech buyers. He has worked in tech journalism for 25 years and is based in San Francisco. Have a news tip? Please drop him an email.

Dig Deeper on AI business strategies

Business Analytics
Data Management