A new report by the Deloitte AI Institute shows that investment in generative artificial intelligence (generative AI) is growing, with early results proving promising. However, organisations are struggling to overcome data management issues and risk-related challenges that hinder their ability to scale AI initiatives effectively. The data and risk management hurdles are tempering initial enthusiasm, too, with only 30% or fewer of their AI experiments fully moving into production for 68% of the respondents.
Despite these misgivings among respondents, Deloitte AI expert Adam Cichocki said that the benefits of generative AI deployments still outweighed the drawbacks. “Our latest research indicates that the top benefits of Gen AI are extending beyond improved efficiency, productivity, and cost reduction,” said Cichocki, who also acts as the organisation’s advisory partner in Jersey. More than half of respondents, he added, pointed to “increased innovation, improved products and services, enhanced customer relationships and other types of value” being produced as a result of investing in AI.
Corporate investment in generative AI continuing to rise, says Deloitte
The third quarterly edition of Deloitte’s “State of Generative AI in the Enterprise” report reveals that 67% of the 2,770 director to C-suite-level respondents across 14 countries are increasing their investment in generative AI. Among those surveyed, data management emerged as a critical issue, with 75% of organisations increasing investments in data lifecycle management to support their AI strategies. However, despite these efforts, 55% of respondents reported avoiding specific AI use cases due to data-related concerns, including data quality, privacy, and security.
To address these challenges, organisations are focusing on enhancing data security (54%), improving data quality practices (48%), and updating governance frameworks (45%). The need to ensure accurate, secure, and properly governed data is seen as essential for scaling AI deployments effectively.
Deloitte’s study additionally identifies risk management and regulatory compliance as major barriers to the successful deployment of generative AI. Only 23% of respondents feel highly prepared to manage risks, such as model bias, hallucinations, and privacy concerns.
The top challenges highlighted by the Deloitte survey are regulatory compliance (36%), difficulty in managing risks (30%), and the lack of a governance model (29%). To mitigate these risks, 51% of organisations have established governance frameworks for AI, 49% are actively monitoring regulatory requirements, and 43% are conducting internal audits and testing.
Enthusiasm gap evidenced among execs
Measuring the success of generative AI deployments was also proving difficult for many of the executives Deloitte canvassed. The report indicates that 41% of organisations find it challenging to define and measure the impact of their AI efforts. Only 16%, meanwhile, produced regular reports for CFOs on the value created by AI. To address this, some organisations are adopting specific KPIs for evaluating AI performance (48%), developing frameworks for assessing AI investments (38%), and tracking changes in employee productivity (38%).
Interest in generative AI from senior executives and board members also shows signs of waning, down 11% and 8% points respectively since Q1 2024. While enthusiasm for the technology remains relatively high, clear, value-driven use cases seem to be needed to sustain investment and support from leadership.
Deloitte’s survey emphasises that while generative AI offers significant potential, successful scaling will require a balanced approach that addresses both opportunities and obstacles. Organisations are encouraged to focus on integrating AI into their business processes, enhancing data management, and implementing robust governance and risk management frameworks. Deloitte also argued strongly for a dedicated, senior executive in each organisation tasked with managing risks surrounding generative AI deployments.
“This… is something very few organisations currently have,” says the report. “This executive should also carefully consider pursuing Generative AI applications using more sensitive data – and not altogether avoiding those use cases.”