MUMBAI: Artificial intelligence is delivering average net productivity gains of 8.2 per cent to 12.3 per cent across industries, but companies are still working to build governance frameworks capable of managing the risks associated with wider AI adoption, according to a Morgan Stanley Institute for Sustainable Investing report.
The report, Governance and AI Adoption, is based on interviews with C-suite professionals and a survey of 200 executives involved in AI governance at global companies.
Morgan Stanley Research’s Alphawise team conducted 808 interviews with C-suite professionals in April 2026 across banking, software and services, technology hardware and equipment, semiconductors and professional services. This was combined with 935 interviews conducted in October 2025 across transportation, autos, consumer staples, healthcare and real estate.
Across 10 industry groups, companies estimated average net productivity gains from AI implementation over the previous 12 months at between 8.2 per cent and 12.3 per cent.
The gains extend beyond technology-focused sectors, with productivity improvements also reported in consumer staples distribution and retail. Customer service, marketing, operations, supply chain, sales and human resources were among the functions benefiting from AI adoption.
However, governance frameworks are yet to fully catch up with the pace of adoption. Among the 200 executives surveyed, 48 per cent said their existing risk-management structures would be sufficient with some changes, while 39 per cent said their frameworks should work but were still being assessed.
Nearly 90 per cent of respondents said their organisations had either defined or were in the process of defining responsibility for AI governance and risk management.
Specific model controls showed lower levels of implementation. Only 41 per cent of companies had a formal inventory of AI models with assigned ownership fully in place, while the same proportion had formal testing and validation protocols.
Defined standards for AI model design and data use were fully implemented at 39 per cent of companies, while 33 per cent had a model-risk classification and tiering system. Structured mechanisms for identifying unintended or emerging risks had the lowest implementation rate, at 27 per cent.
Data-related risks, including privacy, cybersecurity and discrimination, were identified as the biggest current AI concern, with 56 per cent of executives describing them as the most material risk.
Over the next two to three years, 35 per cent of executives expect regulatory and legal risks to become the biggest concern, followed by execution risks at 32 per cent, including challenges in generating returns from AI investments.
Regulatory complexity was also identified as the biggest barrier to effective AI governance, cited by 38 per cent of respondents.
