Enterprises double their generative AI deployment efforts, Bloomberg survey says

The rate at which companies deploy generative artificial intelligence (Gen AI) copilot programs doubled between December last year and July 2024, according to a Bloomberg Intelligence report on AI. 

Also: A third of all Gen AI projects will be abandoned, says Gartner

The report, which surveyed 50 CIOs at US-based companies in July, found 66% of respondents are working on deploying generative AI copilots, compared with 32% in December’s survey, according to lead author Mandeep Singh, Bloomberg Intelligence’s senior industry analyst.

<!–>

–>

The principle use case for Gen AI cited by over half of respondents is chatbot agents, such as for customer service applications. 

<!–>

Companies also increased their evaluation of the training of foundation models, the large language models that form the basis of most Gen AI applications. The number of respondents “working on” training foundation models rose from 26% in the December survey to 40%, and half of respondents said they are “evaluating” the training of models.

Sing said these deployments may lead to a surge in AI inference work among companies. As many as 60% of respondents said their organization intends to increase spending on Microsoft’s Azure for AI inference work, the asking of predictions, up from 41% in the December survey.

Azure leads the pack in cloud providers for inference choice, with Amazon’s AWS cloud service dropping from 55% to 42% of respondents between the December and July surveys. Google Cloud is in third place, with 36% of respondents intending to increase spending for inference. 

Singh said demand for Azure’s inference is expected to continue to increase, partly because of the appeal of partner OpenAI’s Gen AI models, such as GPT-4, which are unavailable on AWS or Google Cloud. 

Also: There are many reasons why companies struggle to exploit Gen AI, says Deloitte survey

“The integration of Microsoft’s Azure platform with OpenAI models continues to be an advantage over public cloud rivals for hosting inferencing workloads,” wrote Singh in the report.

The percentage of respondent companies using OpenAI models jumped from 41% in December to 70% in the latest survey. “Google Gemini use was just 18%, compared to OpenAI’s 70%,” wrote Singh, referring to Google’s top Gen AI offering. 

Also: Google’s new Gemini models achieve ‘near-perfect recall’

The survey also suggests Gen AI’s appeal is helping Microsoft close the gap with Amazon in cloud provision. “At the end of 2023, Microsoft’s share of the cloud infrastructure services was 16%, compared to AWS’ 47%,” versus 48% to 12% in 2018, the report states. “We expect this gap to narrow even more,” wrote Singh.

Aside from the big three cloud specialists, Snowflake and MongoDB ranked high among preferred vendors for developing “retrieval-augmented generation” (RAG), an increasingly popular Gen AI technique, where the AI model taps into an external database. 

MongoDB and Snowflake were the top choices for RAG for 14% and 10% of respondents respectively, behind Azure (26%) and Amazon (20%).

The progress towards Gen AI deployment highlighted by the Bloomberg survey contrasts with other recent research. Analyst Gartner predicted at least 30% of generative AI projects will be abandoned after the proof-of-concept stage by the end of 2025 as companies struggle to prove and realize value from the technology. 

Meanwhile, a recent Deloitte study of Gen AI found most organizations had moved only a third or less of their Gen AI projects into production.

It’s important to note that Bloomberg’s report uses tentative language such as “working on” and “evaluating”. In short, progress is required before Gen AI is embedded in the enterprise.