75% of companies see positive ROI with AI
MIT focuses on custom AI development projects aimed at deep transformation within companies. These projects are complex, have long timelines (development + extended optimization feedback loops), and face many challenges such as inflexible systems, high failure rates, and organizational resistance. So, MIT reports a high failure rate (95%) for these ambitious custom AI pilot programs.
Wharton takes a broader view, including more general AI adoption such as using off-the-shelf generative AI tools (ChatGPT, Copilot, etc.) for augmentation of workflows like data analysis, summarization, and editing. Wharton finds 74-75% of companies report positive ROI from these kinds of AI uses, which are easier to implement and deliver quicker, though more incremental, gains.
The key difference is between deep custom AI transformations (hard, slow, risky) versus more accessible AI augmentation (easier, faster ROI, human empowerment). Wharton also highlights that positive ROI perception varies by company size (smaller companies see ROI faster), industry sectors, and seniority roles.
For AI businesses and agencies, the practical takeaway is to recognize multiple opportunity paths:
Training companies to use generic AI tools effectively across departments (lower risk, quicker wins).
Recruiting and upskilling AI talent, addressing the current high demand and talent shortage.
Pursuing custom AI development projects that can deliver true transformational ROI but require longer timelines, iteration, and higher pricing strategies.
External partners and agencies are shown to have double the success rate over internal builds, so partnering with skilled agencies is critical.
Companies are optimistic about AI budgets in 2026 with many expecting revolutionary impacts in the next 2-5 years, even as middle managers remain cautious compared to executives.
Overall, there is no outright contradiction—both studies reflect different parts of the AI adoption spectrum: the challenging journey to deep AI transformation and the more widespread, positive impact from adoption of general-purpose AI tools used to augment human work.
This nuanced understanding helps AI businesses position their offerings appropriately and set realistic client expectations around AI ROI and timelines.