The enterprise AI gold rush has a dirty secret. MIT Sloan Management Review researchers Melissa Webster and George Westerman spent months inside 21 large enterprises hunting for the transformations everyone keeps promising. What they found was striking: despite two years of hype, experimentation, and sizable cloud invoices, the sweeping, generative AI–powered business transformations many boards envisioned simply haven’t materialized.
Yet some companies are seeing real returns. And here’s the twist—they’re not swinging for the fences. They’re climbing a staircase.
The “Small t” Doctrine
Webster and Westerman’s central insight reframes the enterprise AI conversation. Forget Big-T Transformation—the kind of sweeping reinvention consultants love to pitch. The organizations generating measurable value are pursuing incremental “small t” transformations, a series of low-risk, high-learning steps that build capability and trust over time.
It’s a three-level architecture:
- Individual Productivity: Employees start with low-stakes tasks like inbox summarization, meeting transcription, calendar optimization, or briefing prep. Humans stay in the loop, wins are tracked, and trust in the technology builds.
- Workflow Integration: AI starts assisting specific roles and processes—developers using it to document code, marketers running AI-assisted campaigns, and customer service deploying smarter chatbots. Companies like Adobe, SAP, and Workday are operating here, seeing meaningful efficiency and quality gains.
- Product & Experience Transformation: The riskiest and most valuable level—where AI reshapes customer-facing products and experiences entirely. Most companies haven’t reached this stage yet. That’s not failure; that’s the point.
Compound Interest, Not a Lottery Ticket
What makes this framework counterintuitive in an era obsessed with disruption? AI value compounds. Each small transformation builds organizational muscle: data hygiene, governance frameworks, employee fluency, and management buy-in. These are the prerequisites for the next, harder transformation.
The companies chasing instant, sweeping “transformative AI” skips this compounding entirely—they try to collect interest on a deposit they never made. Conventional ROI playbooks systematically undervalue the emergent capabilities these systems generate, failing to capture significant portions of actual value.
Vanguard provides a vivid example. Its AI initiatives—contact center efficiency improvements, adviser support, and internal AI training through the Vanguard AI Academy—deliver an estimated $500 million in ROI. It wasn’t a single moonshot deployment; it was a deliberate, level-by-level capability build. Technology didn’t transform Vanguard. Vanguard transformed itself, with AI as the instrument.
The Gold Rush Problem
Webster’s warning is blunt:
“Building the right strategy doesn’t go hand in hand with the gold rush mentality going on right now with generative AI.”
The metaphor is apt. In an actual gold rush, the steady winners weren’t prospectors—they were the shovel sellers, the infrastructure builders, the assay office operators. In enterprise AI, the winners aren’t necessarily the boldest visionaries—they’re the ones quietly building the scaffolding that makes bold transformations possible later.
Risk management and capability building aren’t obstacles. They’re preconditions for lasting transformation.
Lessons for the C-Suite
AI budgets are growing faster than AI strategies. Pressure to show results intensifies every quarter. Webster and Westerman’s prescription is unglamorous but essential:
- Don’t mistake enthusiasm for strategy.
- Don’t try to hammer every problem with AI.
- Focus on areas where AI adds real value.
- Find internal champions and let early wins generate momentum.
This isn’t just about technology—it’s about organizational behavior. The question isn’t what AI can do. It’s whether the organization has built the conditions for AI to perform reliably, safely, and at scale.
The companies answering yes in 2029 are the ones starting their small-t transformations today.
Related: AI Transformation Is a Governance Problem (Not Tech) — 2026 Truth
