Two studies came out in November that tell you everything about AI in 2025.
MIT built a simulation of 151 million US workers and found AI can technically automate 11.7% of jobs right now. Not someday—today. (https://iceberg.mit.edu/report.pdf)
McKinsey surveyed 1,993 companies and found only 6% get real business value from AI, despite 88% using it somewhere. (https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai)
That gap is the entire story.
What MIT Actually Found
MIT’s Project Iceberg used Oak Ridge National Laboratory’s supercomputer to simulate the entire US workforce. They wanted to know: where can AI actually replace human work with today’s technology?
The answer surprised them.
Everyone looks at tech jobs—software developers, data scientists. That’s only 2.2% of the workforce that AI can meaningfully affect.
The real exposure is invisible:
- Processing invoices
- Reviewing standard contracts
- Writing routine reports
- Answering customer emails
- Moving data from documents into systems
- Scheduling and coordination
All the boring cognitive work that happens in every company, every day.
MIT found this “hidden” work represents 11.7% of the US labor market—$1.2 trillion in wages. And here’s the kicker: it’s not concentrated in Silicon Valley. Ohio, Michigan, Tennessee—the cognitive work supporting traditional industries is just as exposed as coastal tech hubs.
The automation potential is already here. It’s not a future scenario.
What McKinsey Actually Found
Meanwhile, McKinsey looked at actual companies.
88% use AI in at least one function. Marketing teams use it for content. Service teams use chatbots. Developers use Copilot. It’s everywhere.
But when McKinsey asked “is this moving your bottom line?” the answer was uncomfortable.
Only 39% of companies report any EBIT impact at all. Most of those say it’s less than 5%.
The high performers—companies where AI actually delivers 5%+ EBIT impact—represent just 6% of all companies surveyed.
So what are those 6% doing differently?
The One Thing High Performers Do
McKinsey ran correlation analysis on 25 different factors. Budget size, tech stack, AI models, team structure, everything.
One thing mattered most: redesigning workflows.
55% of high performers fundamentally changed how work flows when they deployed AI.
Everyone else is doing what McKinsey calls “plug-in thinking”—they take their existing process and add an AI tool to it. Customer service stays the same, but now with a chatbot. Report writing stays the same, but now with ChatGPT.
High performers ask a different question: “If we rebuilt this process from scratch with AI, what would it look like?”
That’s not a technology question. It’s an organizational design question.
Why Most Companies Are Stuck
I see this pattern constantly. Companies know they need to “do AI.” So they:
- Run pilots in isolated teams
- Celebrate small wins
- Write up case studies
- Never actually change how work flows
The IT team deploys tools. The business units keep working the same way. Six months later, executives ask “where’s the ROI?”
The ROI is hiding behind your unchanged processes.
MIT proved the technical capability exists. Your invoice processing, contract reviews, report generation, and data entry can be automated with today’s tools.
But you can’t get there by adding AI to your current process. You have to redesign the process around what AI can do.
The Real Pattern
Look at the McKinsey high performers:
They think bigger. 3.6x more likely to aim for transformation, not just efficiency gains. They’re not trying to save 10% on costs—they’re trying to rebuild how the business operates.
Leaders actually own it. 48% of high performers say senior leaders clearly own AI initiatives. Everywhere else? 16%. This isn’t something you delegate to IT and hope for impact.
They scale aggressively. Instead of sequential pilots (prove marketing, then try sales, then try operations), they go broad fast. Multiple functions simultaneously.
They measure business outcomes, not AI metrics. Not “how many employees use ChatGPT” but “did this change our EBIT?”
What This Actually Means
If you lead AI anywhere—enterprise, mid-market, startup—here’s what these studies tell you:
The capability gap is closed. As I already stated a few weeks ago: The bottleneck isn’t the technology. Stop planning for “when AI gets good enough.” MIT’s research shows it’s good enough now for a huge chunk of cognitive work.
The execution gap is wide open. When 88% of companies use AI but only 6% get real value, execution is your competitive advantage. Most companies haven’t figured out how to actually change how they work.
Your window is narrow. Right now, organizational capability is rare. In 2-3 years, the playbooks will mature, consulting firms will productize the approaches, and everyone will know how to redesign workflows. Your advantage is acting while most competitors are still running pilots.
The value is in boring work. MIT’s data shows the big opportunity isn’t in flashy AI applications. It’s in the routine cognitive work that happens in every backoffice, every finance team, every operations group. If you’re only thinking about engineering use cases, you’re missing 80% of the opportunity, even though these engineering case might be easier to sell to a board.
What To Actually Do
Forget the pilot theater. Here’s the pattern that works:
Pick 3 processes with clear P&L impact. Not “let’s try AI in marketing.” Pick specific end-to-end processes. Order-to-cash. Hire-to-onboard. Issue-to-resolution.
Redesign them completely. Not “add AI here.” Ask “if this process didn’t exist and we built it today with AI available, what would it look like?” Kill steps that aren’t needed. Automate what’s automatable. Keep humans where they add unique value and sometimes just use “old” but reliable tech like RPA for example. Use AI at the right step and task in the whole workflow.
Measure EBIT contribution. Not “adoption rate” or “user satisfaction.” Did revenue go up? Did costs go down? Did throughput increase? Real business metrics.
Get executive ownership. This isn’t an IT project. The high performers have C-suite leaders who own the transformation, protect the budget, and drive adoption. If your AI strategy is owned by a director or group of people three levels down, you’re not in the 6%. (Trust me…)
The exposure MIT found is real. The value McKinsey’s high performers prove is real. The gap between them is temporary.
Most companies are wasting the capability by not changing how they work.
Don’t be most companies.
How is your organization approaching this? Are you redesigning work or just adding tools to existing processes?