95% of GenAI Pilots Show No P&L Impact — Because It's a Culture Problem

MIT's NANDA initiative found that 95% of generative AI pilots delivered no measurable P&L impact — and the fix, according to a wave of enterprise research, has almost nothing to do with better models. When Boston Consulting Group surveyed 1,000 senior executives, roughly 70% of AI implementation challenges stemmed from people and processes, 20% from technology, and 10% from the algorithms themselves.

The pattern repeats across studies. IBM's 2025 CEO study found only 25% of AI initiatives delivered expected ROI, and just 16% scaled enterprise-wide. SAP and Oxford Economics' Value of AI Report 2026 found 69% of businesses "satisfied" with AI ROI, though more than two-thirds see room for improvement — a tension sharpening as cost per token rises. The companies extracting the most value, BCG found, invest 10% in algorithms, 20% in technology and data, and 70% in people and processes.

One CIO writing from practice — at Per Scholas, a tech-training nonprofit — argues three cultural traits separate teams that produce returns from teams that produce pilots: results-orientation, entrepreneurial license, and collegiality.

What does "results-oriented" actually mean here?

McKinsey's 2025 State of AI survey found 80% of companies set efficiency as their AI objective, but high performers add growth and innovation goals. They're nearly three times as likely to have fundamentally redesigned individual workflows — the strongest contributor to business impact McKinsey tested.

Timelines are also maturing. TEKsystems' State of Digital Transformation 2026 report found only 27% of organizations expect transformation ROI within six months, down from 42% a year earlier.

The operating rule: every AI project needs a named owner, a baseline, and a business metric before it starts. Per Scholas built ResumeBoost, an alumni resume tool, and measured time first — it cut staff time on alumni resumes by over a third. But honest measurement means tracking more than one number and acting on what the metrics disagree about.

Why do pilots stall without permission to fail?

A 2025 MIT Technology Review Insights survey found 83% of business leaders believe "experiment-friendly cultures" make AI initiatives much more likely to succeed — yet 22% had hesitated to lead AI projects for fear of failure or criticism.

MIT's researchers also documented a "shadow AI" economy: employees using personal AI tools while official pilots stall. Their prescription: move from central labs to empowered teams.

ResumeBoost itself failed small. Time savings were real, but the alumni user experience missed; Per Scholas folded it into a broader AI coaching platform rather than propping it up. The pilot did its job — it showed where to pivot.

Internal builders matter too. Ed Hernandez, who designed Per Scholas's AI-Native Classroom, spent roughly fifteen years there — learner, then instructor, then curriculum lead. Leadership's job was to say yes, set guardrails, and weigh possibility against necessity.

How does collegiality show up in shipped products?

Accenture's research on "Talent Reinventors" — organizations advancing people and AI together — found they delivered revenue growth 1.8 percentage points higher and profit growth 1.4 points higher than peers in 2025.

Per Scholas built Sherlock, a job-sourcing tool, with the staff who place graduates — not for them. Sherlock reviews over 125,000 postings a day, filters staffing-firm reposts, and exports in the Salesforce format staff already use. Staff tune the feed by flagging mismatches. Three weeks after launch, staff at nearly three-quarters of campuses had logged in, weekly users had more than doubled, and users saved over 100 hours of sourcing time. Seven weeks in, staff were reporting where Sherlock falls short, and enhancements are being scoped.

The failure mode nobody budgets for

AI is a force amplifier, not a force generator: it magnifies whatever culture already exists, bad or good. Candor is a precondition — people don't experiment freely with a technology they suspect will replace them. Equinix's chief people officer, describing early-career pathways, put it plainly: "We're not just competing for talent; we're creating it."

Employers say the differentiators are whether people can define a problem, try something new, and work across a team to solve it. As token costs climb and CFOs start asking harder questions, those same traits will decide whether AI budgets become line items or returns.