According to the report, only 5% of organizational initiatives deliver measurable business returns. Those who succeed are the ones who focus on solving one specific problem and align it with genuine customer needs.
The primary reason Generative AI fails is the learning gap within these systems: they do not adapt or improve over time, making it difficult to integrate them into existing workflows and processes. The report exposes what it calls the “AI shadow economy” — people using personal AI tools such as ChatGPT to complete tasks, even when the organization has purchased an official enterprise subscription.
⚠️ Important note: The research has been subject to criticism regarding its reliability, due to a small sample size that includes only 300 publicly disclosed AI initiatives, interviews conducted with just 52 organizations, and a survey of only 153 senior executives.
Contrary to common assumptions, the highest business value actually comes from investing in back-office process automation — driving down costs, reducing outsourcing, and improving operational efficiency in ways that ultimately benefit customer service. Despite this, 70% of investments continue to flow toward sales and marketing.
Another key to success is the “buy before build” strategy, which indicates that vendor-partnered solutions succeed in 67% of cases, compared to just 33% for internally built solutions.
The two industries showing the most meaningful AI adoption are technology, media, and telecommunications. Healthcare, financial services, and manufacturing show a great deal of experimentation but relatively little actual transformation.
Ultimately, I think the report should be taken in proportion given its small sample size — but there are several findings worth carrying forward.
Five Practical Steps from the Research
- Map the critical organizational processes where AI offers clear, demonstrable business value.
- Select one process, define it end-to-end, and identify integration points.
- Engage frontline managers as early adopters — ask them to identify for themselves where AI fits in.
- Act like a BPO client — demand deep customization from your vendor and tie them to measurable business outcomes.
- Focus on adaptive learning systems that improve over time.
Recently, we developed a bank-wide AI adoption program that addresses two vectors simultaneously: advancing the level of AI adoption among employees, while investing in deep learning processes in the areas most affected by the change and with the highest business impact on the bank.
So if you are evaluating an AI initiative, start with one critical process — not with marketing. Ask yourself: where will we save the most, or improve efficiency in a measurable way?