Google May Have Just Solved the AI Failure Mystery

MIT’s finding puzzled a lot of us.

95% of enterprise AI pilots deliver zero ROI. That’s $40 billion wasted. Then Google’s DORA report connected the dots… 

They surveyed nearly 5,000 technology professionals globally and found the pattern: 

> ## *AI doesn’t fix broken organizations. It amplifies whatever’s already there.*

The data shows 90% of software professionals now use AI daily. 80% report productivity gains. Yet organizational outcomes split dramatically. According to DORA’s research, teams with strong foundations see 1.5x revenue growth. Those without? They experience increased delivery instability and burnout.

 

What DORA Actually Measured

The report identifies seven capabilities that determine AI success:

 

1. Clear AI stance. The business goals are clear to everyone and well-aligned.
2. Healthy data ecosystems. Quality data that’s properly governed and maintained.
3. AI-accessible internal data. Your systems can actually share data with AI tools.
4. Strong version control. Every change is tracked, reversible, and reviewable.
5. Small batches. Work flows in manageable chunks, not massive transformations.
6. User-centric focus. Real user needs drive decisions, not technology possibilities.
7. Quality internal platforms. Developer tools and infrastructure that actually work reliably.

 

Only 13% of companies have these foundations in place.

The trust paradox reveals the problem. While developers use AI heavily, 23% trust it “a little” and 7% “not at all.” We’re adopting tools we don’t fully trust because we lack the systems to validate their output.

 

The 70-20-10 Pattern

 

BCG’s analysis of AI leaders found they generate $3.70 for every dollar invested. Their approach contradicts conventional wisdom: 70% of investment goes to people and processes, 20% to technology and data, only 10% to algorithms.

MIT’s research adds another layer. Companies building custom AI solutions succeed 33% of the time. Purchasing vendor solutions succeeds 67% of the time. The pattern appears clear once you see it.

According to the transcript, Lumen reduced sales prep from 4 hours to 15 minutes after establishing robust processes first. The sequence matters: process excellence, then AI amplification.

 

Industry Patterns From the Data

 

The research shows fintech companies lead with 49% success rates. Software companies follow at 46%. Traditional industries struggle more consistently. 

Gartner predicts 30% of GenAI projects will be abandoned after proof-of-concept by end of 2025. McKinsey found fewer than 10% of vertical use cases survive pilot stage.

Only 32% of organizations report high data readiness. Just 21% have necessary GPU infrastructure. These gaps explain why so many initiatives fail.

 

Your Monday Morning Move

 

The evidence suggests three actions.

 

– [ ] First, audit those seven DORA capabilities. Most organizations will score poorly. Face that baseline honestly.

– [ ] Second, pick one workflow to standardize completely before adding AI. Sales qualification. Customer onboarding. Code review. Whatever you choose, fix the process first.

– [ ] Third, reallocate budget toward change management. The data shows organizations underinvest here by multiples.

 

The research makes one thing clear: excellence before AI leads to multiplication of strengths. Dysfunction before AI leads to amplified problems.

We know what works. The question is whether we’ll follow the evidence or repeat the failures.

 

If you have questions or suggestions, happy to discuss any time.

 

—

 

Sources:

https://blog.google/technology/developers/dora-report-2025/

https://fortune.com/2025/08/18/mit-report-95-percent-generative-ai-pilots-at-companies-failing-cfo/

https://www.bcg.com/press/24october2024-ai-adoption-in-2024-74-of-companies-struggle-to-achieve-and-scale-value