Global AI in Financial Services Report
A · Strong evidence Cambridge Centre for Alternative Finance survey of 628 organisations in 151 jurisdictions, run with the BIS, IMF, World Bank, WEF and others; real cross-stakeholder data, though self-reported and not a random sample.
Summary based on the full text·Original: about 4 hours to read
In short
A 140-page report published in April 2026 by the Cambridge Centre for Alternative Finance (CCAF) at Cambridge Judge Business School with Financial Innovation for Impact, and research partners including the BIS, IMF, World Bank, WEF, IDB, CGAP and Arab Monetary Fund. It surveys fintechs, traditional financial institutions, AI vendors and regulators on AI adoption, value, risks and regulation, extending a 2020 CCAF-WEF study. Main finding: adoption is wide (81% of firms) but mostly internal and back-office, value is hard to measure, and regulators trail industry by a wide margin.
Key numbers
- 52%industry respondents in active adoption of agentic AI; 23% at Scaling or Transforming
- 14%industry respondents who see AI as transformational to strategy and competitive advantage
- 55%industry respondents finding it difficult to measure the value of AI, rising to 76% among large financial institutions
- 24%industry respondents expecting a net reduction in roles by 2030
As reported by the source.
English·PDF·opens jbs.cam.ac.uk
What it reports
- 81% of surveyed financial firms adopt AI at some level; 40% report advanced adoption versus 20% of regulators. Fintechs lead incumbents in advanced adoption, 47% to 30%.
- Agentic AI is in active adoption at 52% of industry respondents (fintechs 57%, traditional FIs 45%); 81% expect agentic AI to be meaningfully achieved by 2030.
- Top use cases at pilot or beyond are internal: process automation (79%), data visualisation (75%), software engineering (75%), data and knowledge management (69%). Customer support leads the front office at 74%.
- OpenAI is the most-used foundation model provider (76% of industry), followed by Google (57%) and Anthropic (35%); DeepSeek 15%.
- 40% report higher profitability from AI, 43% no change. Among organisations spending more than USD 100,000 a year, 62% report higher profitability versus 39% of lower spenders.
- Data quality is the top barrier: cited by 66% of vendors, 46% of regulators and 40% of industry. Vendors cite client data quality and completeness at 72%.
- Top risks for all groups: data privacy (74% of industry, 80% of regulators) and hallucinations (70% of industry and regulators). About two-thirds of industry are not monitoring for bias.
- 79% of regulators rate explainability critical or important; 50% of industry use explainable AI methods.
- Workforce: 24% of industry expect net job reduction, about 25% expect reskilling without large net losses, 10% a net increase. 44% of all respondents expect AGI by 2030.
How they got it
Three parallel online surveys run on Qualtrics between October 2025 and January 2026 in ten languages: industry (203 fintechs, 149 traditional FIs), 146 AI vendors and 130 regulators and central banks, 628 in total. Distribution ran through partner networks, direct outreach and LinkedIn; responses were cleaned for duplicates, bots and outliers. Partners contributed commentary chapters using anonymised data only.
Limits
- Voluntary, network-distributed sample; firms active in AI are likely over-represented.
- Profitability and productivity effects are self-reported perceptions, not measured financial outcomes.
- The spend-profitability link is a correlation and may reflect firm size or maturity.
Why it matters for companies
Financial services runs a few years ahead of most traditional sectors, so its pattern is a preview: wide adoption, use concentrated in back-office efficiency, and few firms able to show enterprise value. More than half find value hard to measure, which argues for setting baselines before deployment rather than after. Data quality remains the main blocker, as it was in 2020, so data work should be budgeted as part of any AI programme. The finding that a quarter of firms expect net role reductions, while more expect reskilling, gives HR a realistic planning range.
Summary based on the full text. Read the original for the full argument and data.Open PDF on jbs.cam.ac.uk
- Tags
- Operating model, Risk