Most surveyed companies prefer a mix of public and private generative AI models. Security risks and technical complexity can slow adoption down.

Most IT decision-makers pursuing adoption of generative AI choose a hybrid approach of a mix of public and private models (38%), a recent study found. The hybrid approach is appealing because it matches companies’ needs to protect data, maintain control over AI models and results and cost. This illustrates what IT decision-makers prioritize when making decisions about generative AI adoption.

On behalf of Dell Technologies, Morning Consult surveyed 500 IT decision-makers involved in generative AI initiatives. Survey respondents were located in the U.S., U.K., France and Germany. Results were gathered in August and September 2023.

Jump to:

Leading factors for companies’ strategic approach to generative AI

Of the respondents whose organizations have moved beyond a pilot stage with generative AI, 80% use centralized decision-making and/or a center of excellence in their strategic approach. 87% of people whose organization has moved past a pilot program believe generative AI is on track to deliver meaningful results, and 76% are increasing their budgets to include AI.

SEE: What factors should go into choosing between public or private generative AI models for business? (TechRepublic) 

Other factors respondents value highly when making decisions about how to buy and implement generative AI are:

• Security and protecting the value of data.
• More control over models and better output results.
• Cost.

“It’s really a brand new technology that now humanity has to play with, and now we’re looking at whether it operates similarly to anything I’ve used in the past, especially when it comes to access and identity,” said Ryan Orsi, worldwide partner lead for security at AWS, in an interview with TechRepublic.

“What are providers doing with their prompts and their responses is another big question, which touches on data privacy and data security,” Orsi said.

Security, control and cost influence whether IT decision-makers use public models, build their own mode or choose something in between.

• 38% plan to approach generative AI by classifying their data and using a hybrid approach.
• 21% plan to retrain an existing model using their own data in their own environment.
• 16% prefer to purchase public models in the cloud.
• 14% prefer to use open source or other models on-premises for inferencing.
• 9% prefer to build their own model from scratch.

How far are organizations in the generative AI adoption journey?

44% of respondents are at an early to midpoint in the adoption of generative AI. This means they either have no strategy around generative AI or established core use cases, but they have not yet deployed solutions.

42% of respondents say their organization is not too hesitant about generative AI adoption. Another 29% are somewhat hesitant. On the extremes, 8% are very hesitant, and 21% are not hesitant at all.

Many respondents whose organizations moved beyond pilot programs (49%) expect value within six months to a year.

76% of people surveyed felt generative AI impact will be “significant if not transformative.” In particular, they expect it to:

• Provide productivity gains.
• Streamline processes.
• Achieve cost savings.

Concerns slowing generative AI adoption

The top reasons why respondents are hesitant to implement generative AI are:

• Security risks such as data or intellectual property leakage.
• Technical complexity.
• Data governance concerns such as regulations or compliance.
• Cost of implementation.
• Concerns around ethical or responsible implementation.

A small number of organizations surveyed (5%) ban the use of generative AI. Of the four countries surveyed, the number of organizations which ban generative AI is highest in the U.S. (6%) and lowest in the U.K. (2%).

Dell generative AI product news and competitors

Dell has been forward-looking in adopting generative AI within its own products and providing generative AI-related services. On October 4, 2023, Dell announced it would add pre-trained models and inferencing to its generative AI services with Dell Validated Design for Generative AI with NVIDIA for Model Customization globally in late October. A suite of Dell Professional Services for Generative AI will be available in select countries starting in late October.

Competitors to Dell’s generative AI hosting and professional generative AI services include Snowflake, Amazon’s SageMaker, Google Cloud Platform’s AutoML and Vertex AI, and Microsoft Azure.