There’s a story from my earlier corporate years that has stayed with me. Not because it was dramatic, but because of what it reveals about how organizations fail to think.
Back then, global compute and storage costs were brutally high. Every gigabyte mattered. IT was viewed strictly as a cost center. Expensive, necessary, and tightly governed. And so, a mandate came down from on high:
Everyone needed to reduce the size of their email inboxes and folders.
To enforce it, the company implemented a hard storage cap. If you hit the limit, your email simply stopped sending and receiving until you purged old messages.
Across the organization, people dutifully deleted years of archived emails: project histories, negotiations, customer conversations, decisions made and undone, design debates, strategy discussions, leadership missteps, and breakthroughs.
All of it gone.
At the time, it felt like housekeeping. In hindsight, it was the mass destruction of an unrecognized asset.
A decade later, leading digital and data transformations, I realized the magnitude of that loss. That deleted email corpus would have been an extraordinary training dataset for generative AI. It represented institutional memory, behavioral patterns, decision heuristics, operational nuance, and the tacit knowledge that no document management system ever truly captures.
We didn’t just reduce storage costs. We erased the raw material of future intelligence.
The Strategic Error Behind the Error#
Looking back, the mistake wasn’t the storage cap itself. It was the mindset.
IT was treated as a cost to minimize, not a capability to unlock. When technology is framed as a cost center, the instinct is to shrink, prune, and constrain. You don’t ask, “What new value might this create?” You ask, “What can we cut?”
Data was not yet understood as an asset. At the time, storage was a financial cost; data was a liability risk. Nobody imagined a future where the everyday byproducts of work would power entirely new classes of intelligence.
Efficiency trumped foresight. The mandate optimized for today’s budget, not tomorrow’s opportunity. It solved a short-term operational problem at the expense of long-term strategic potential.
The Irony: The very emails we were told to delete could have trained models that would eventually automate, enhance, or replace half the work described in those same emails.
We Weren’t Wrong. We Were Early.#
It’s easy to look back and shake our heads. But the truth is: no one knew. Generative AI wasn’t in the lexicon. Storage cost economics were still restrictive. Data privacy frameworks were designed to reduce exposure, not enable intelligence.
But today, we do know.
Today, leaders understand that data is the differentiator. Storage is cheap. Compute is elastic. Models improve by learning from the specific organization they serve.
The Real Lesson: Don’t Make Tomorrow’s AI Dumber By Accident#
If the 2000s were the era of deletion, the 2020s must be the era of deliberate retention. Not reckless hoarding, but retention with purpose.
1. Treat data as a strategic asset, not a storage liability.
If you believe AI will shape competitive advantage, then you must protect the training fuel. That means implementing governance that values preservation as much as it values compliance.
2. Don’t optimize for today’s costs at tomorrow’s expense.
Generative AI rewards the organizations that understand themselves best. What looks like “operational exhaust” today might be the most valuable dataset in your company 10 years from now.
3. Build institutional memory that outlives turnover.
People leave. Email gets archived. Documents get lost. AI, properly designed, can preserve narrative continuity across years, roles, and leadership cycles.
4. Beware the hidden opportunity cost of small decisions.
The storage cap was, on paper, a tiny policy change. In impact, it was a decade-long strategic downgrade. This is true of many seemingly minor IT decisions: data retention settings, API access rules, log storage policies, file expiration defaults. Each one constrains future intelligence.
The Small Things Matter#
The lesson from that deleted email history is not to keep everything forever. It’s that the mundane often becomes meaningful in hindsight.
Leaders today must ask: What are we deleting now that future AI will wish we hadn’t?
- Are we pruning log files?
- Are we discarding support chats?
- Are we de-duplicating customer histories?
- Are we shrinking telemetry windows?
- Are we archiving away cultural signatures?
If generative AI is to become the cognitive companion to business (predicting, advising, drafting, optimizing), then the richness of what it can do will come from the richness of what we choose to keep.
The Strategic Mandate#
As organizations rethink their data and AI strategies, this principle must become non-negotiable:
Build for the future where AI needs to learn from us. Not the past where storage was expensive.
If you ever needed a parable to illustrate the cost of getting it wrong, remember the terabytes of email knowledge we once deleted just to save on disk space.
A decade later, the bill for that decision finally arrived. And it came in the currency of lost intelligence.