Businesses are quick to adopt GenAI but are struggling to productionize and scale their infrastructure, says a new study that looked into the challenges businesses face in adopting AI.
This was the conclusion of a new Economist Impact report, commissioned by Databricks, titled “Unlocking Enterprise AI: Opportunities and Strategies”. The global study surveyed some 1,100 technical executives, data engineers, data scientists, and architects, and includes insights from 28 C-Suite executives from 11 industries.
Most are using GenAI
The study found that large organizations are rushing to GenAI, with 97% of companies with over USD10 billion in revenue now using the technology. And by 2027, 99% of all respondents expect GenAI adoption across both internal and external use cases.
Indeed, only 18% of respondents think AI is overhyped, while 73% view the technology as vital to their long-term goals. The vast majority of respondents (85%) say they are using generative AI (GenAI) in at least one function.
In addition, 58% of data scientists have begun to augment their LLMs with proprietary data through retrieval augmented generation (RAG). Two-thirds of organizations recognize the potential in combining LLMs with enterprise data to enhance data intelligence.
Organizations also expect to mix and match different models and tools, spanning open source and proprietary technologies, to drive better performance. By 2027, almost all (96%) plan to deploy open source AI models.
More to be done
Despite more companies investing in AI than ever before, the report highlights challenges in delivering business-specific, highly accurate, and well-governed results at a reasonable cost. And these challenges are preventing organizations from scaling their AI efforts and achieving more transformational results.
Specifically, only 37% of executives believe their GenAI applications are production-ready. This figure falls to just 29% among practitioners, who cite key hurdles including cost (41%), skills (40%), quality (37%) and governance (33%). Moreover, few (22%) feel confident that their current IT architecture could effectively support new AI applications moving forward.
It’s worth noting that nearly half of data scientists (45%) are still using a general-purpose large language model (LLM) without contextual enterprise data, which often struggles to provide the necessary quality, governance, and the ability to evaluate outputs.
“It’s clear that AI is becoming an integral part of every business, but leaders still have concerns about quality and cost when it comes to GenAI. They’re seeking solutions tailored to their organizations, and they realize they need a platform that prioritizes data, centralizes governance and delivers efficient TCO at scale,” said Andy Kofoid, the president of Global Field Operations at Databricks.
Much work remains to be done. Ultimately, the report emphasizes the importance of data intelligence and highlights the importance of a holistic approach that encompasses data management, governance, and domain-specific expertise.
The full report can be accessed here (Free registration required).
Image credit: iStock/megaflopp
