City
The tech landscape is undergoing a massive shift toward AI, yet we’re already seeing the ghost of cloud sprawl past. Just as costs ballooned and promised efficiencies stalled a decade ago, structural inefficiencies are resurfacing in the AI era.
To avoid repeating history, the goal isn’t to assign blame for past mistakes, but it is to recalibrate the incentive structures that drove them. Here are the key lessons from the cloud era, reframed to help organizations navigate the AI revolution.
Moving Beyond the “Speed at Any Cost” Incentive
During the initial cloud rush, many IT departments were incentivized solely on the speed of migration. This led to the lift-and-shift phenomeno, moving legacy technical debt into the cloud without optimizing it. The result? Companies ended up paying more than necessary because they were running unoptimized architectures on someone else’s hardware.
In AI, the pressure to do something with LLMs is even higher. If we incentivize teams simply to launch a pilot project by next quarter, we risk creating AI debt.
Incentives should shift from deployment speed to architectural efficiency. Success shouldn’t just be it’s live, but it should be it’s live, it’s governed, and the unit cost per inference is sustainable.
Bridging the Technical Literacy Gap at the Board Level
For years, cloud strategy was often treated as an outsourced utility rather than a core business competency. Because many boards lacked deep technical literacy, they were incentivized to trust optimistic and high-level vendor promises over the cautionary warnings of their own architects.
AI is not a commodity you buy, but it is a strategic differentiator you build.
We must incentivize dissent. Organizations thrive when leaders are encouraged to ask the hard questions about data lineage, model drift, and long-term ROI, rather than just following the most polished vendor roadmap.
Solving the No Post-Mortem Culture
In many corporate environments, admitting that a cloud migration didn’t deliver the expected 20% savings is seen as a career risk. Consequently, no news is good news, and inefficient systems stay hidden until the bill becomes too large to ignore. Nearly 30% of cloud spend is still wasted on idle resources.
AI projects have an even higher failure rate. If failure is punished, the real reasons for those failures will be swept under the rug.
Create learning incentives. Reward teams for conducting honest post-mortems. If a project fails, the value should be captured in the data and lessons learned so the next project doesn’t hit the same wall.
Take Ownership
To move fast, many companies outsourced their entire cloud strategy to third-party consultants. While this provided immediate scale, it left the company without the internal muscle to manage or optimize those environments once the consultants left.
AI is a force multiplier for your specific business data. If you outsource the strategy, and not just the implementation, you are outsourcing your future competitive advantage.
Incentivize the development of internal talent. The most successful companies in the next decade will be those who incentivized their own engineers to become AI-fluent.
The Bottom Line
The system that rewards short-term optics over long-term optimization, has hidden costs. As we move into AI, the stakes are higher, while an unoptimized cloud bill is a problem, an unoptimized AI strategy can be even worse. By shifting our incentives from hype and speed to governance, efficiency, and internal mastery, we can ensure that the lessons of the cloud era aren’t just remembered, but acted upon.