Executive summary
I've seen companies pour fortunes into AI that nobody uses — and others transform their operation with a fraction of that investment. The difference was never the technology. It was leadership. 88% of organizations use AI, but only 6% actually get value from it. 74% never make it past the pilot. 30%+ abandon the project after proof of concept. And yet, the ones who do get it right: +15.8% revenue, +22.6% productivity. This research gives you the data, the mistakes that explain most of the failures, what sets the 6% apart, and a framework for deciding with judgment.
Context and state of the sector
It's 8 a.m. on a Monday and your operations director is reviewing the same report as always — that scene, which I described in the article on leadership and digital transformation, is the paralysis I see in most of the companies I work with.
88% use AI. 6% get value from it. Adoption went from 55% (2023) to 88% (2025) (McKinsey). But only 6% generate consistent value. The other 82% have decorative AI.
74% never make it past the pilot. The "valley of death" of enterprise AI (BCG, 2024). "We tried it, it worked, but we never integrated it."
30%+ of GenAI projects are abandoned post-POC. The ones that survive: +15.8% revenue, +22.6% productivity (Gartner, 2025). AI leaders: 1.5x more revenue growth than average (BCG).
Productivity has quadrupled. From 7% to 27% in AI-exposed industries, 2022-2024 (PwC).
Spending is accelerating. $1.48T (2025) → $2.52T (2026), +44% year over year. 33% of enterprise apps will include agentic AI by 2028 (Gartner).
Latin America: bigger gap, bigger opportunity. Only 23% generate value with AI. Only 6% report significant impact. But AI could generate $1.1-1.7T/year in the region (WEF + McKinsey, 2026).
Macro impact is no longer speculative. Global GDP +7% (~$7T) per Goldman Sachs. $15.7T by 2030 per PwC. GenAI: $2.6-4.4T/year per McKinsey.
Documented cases and trends
Bancolombia — a decision from the president's office, not IT. AI in risk, service and operations. A mandate from senior leadership with protected budget and metrics reported to the board. Result: one of the most innovative banks in Latin America.
McKinsey's 6% — what they have in common. (1) AI embedded in core processes, not side projects. (2) Investment in training and cultural change, not just tech. (3) Impact metrics = business metrics, not adoption metrics.
The agentic enterprise. AI adoption: 50% (2022) → 72% (2025). Agentic AI: 35% adoption + 44% planning. 51% of companies in North America are experimenting with AI agents (BCG + MIT Sloan, 2025). The next wave is already here.
$1 in tech = $3 in change management. McKinsey (2025). Companies that budget only for the license find out by month three that nobody's using it.
Common implementation mistakes
1. Handing it off to IT. It becomes a technical project with metrics nobody in leadership understands. Successful AI is a business project with a technology component.
2. Starting with the tool. "We need AI" isn't a strategy. For what? Cut costs? Scale? Decide faster? The answer defines everything.
3. Not budgeting for the change. $1 in tech = $3 in training, redesign and change management. License only = an abandoned tool.
4. Measuring adoption, not impact. "200 active users" isn't value. Are they generating more revenue? Cutting costs? Improving satisfaction? If you can't answer that, you don't know if it worked.
5. Doing nothing and thinking it's free. The cost of the status quo doesn't show up on an invoice. It shows up in clients who leave, talent who quits and competitors who get ahead.
Decision framework
| Question | Why it matters |
|---|---|
| Who's leading — IT or senior leadership? | IT = ceiling. Leadership = mandate. |
| A real business problem, or a feeling that you "should do something"? | Without a clear problem, there's no correct solution. |
| Does the budget include change management? | If not, get ready for an unused tool. |
| Success = adoption or business impact? | Define business metrics BEFORE you start. |
| Does your leadership team understand what AI can and can't do? | If not, expectations are misaligned. |
| Willing to change processes? | Automating something broken at machine speed isn't transformation. |
Want to answer those questions with data about your own company?
I offer a free AI Diagnostic: in about 15 minutes, we look together at where the real opportunity is. No commitment, no jargon, with a focus on business impact.
See the AI Diagnostic →References
- McKinsey / QuantumBlack — State of AI (2025)
- BCG — Where's the Value in AI? (2024)
- BCG + MIT Sloan — The Emerging Agentic Enterprise (2025)
- Gartner — GenAI Impact and Abandonment (2025)
- Gartner — AI Spending Forecast (2026)
- PwC — Global AI Jobs Barometer (2025)
- Goldman Sachs — AI and GDP (2023)
- McKinsey — Economic Potential of GenAI (2023)
- WEF + McKinsey — Latin America in the Intelligent Age (2026)
- Deloitte AI Institute — State of AI in Enterprise (2026)
Next step
Let's talk about your company →
In 15 minutes we'll assess which processes are consuming the most of your leadership time — and which have the greatest potential.
