In 1979, a new tool called VisiCalc landed on finance desks. For the first time, analysts could model scenarios without waiting for IT, and adoption of the solution was explosive. Within a few years, the future of work was clear, and entire industries were running on spreadsheets. Unfortunately, however, this empowerment came with a cost: in 2012, for example, JPMorgan’s infamous “London Whale” losses of $6 billion were due in part to manual copy-paste errors in an Excel spreadsheet.
The tension between user freedom and organizational coherence defined the spreadsheet era. It also created the first “shadow IT,” that blackmarket of personal tools that undermine governance in the interest of individual agency.
We’re now at the front edge of a similar wave of AI innovation. Just as spreadsheets democratized data but spawned costly chaos, AI agents will democratize action. Unless leaders plan ahead, the shadow cast by AI will be far larger.
The Psychology of Ownership: From Excel Cells to the Future of Work
Spreadsheets were never just a neutral tool to be used; they quickly became personal vaults of intellectual property within organizations. Anyone who’s ever tried to help a company move from “spreadsheet hell” to a centralized system knows the vitriol (and workarounds) that result as folks lose “their” pivot tables.
This psychology of ownership is already emerging around AI agents. As we saw in this summer’s backlash to OpenAI’s GPT-5 rollout, changes that impact AI models result in uproar and resistance. Users build AI prompts, workflows, and trust around specific agent behaviors; when these behaviors change, the user response can be over-sized. The disruption is experienced viscerally, not intellectually.
The parallel with spreadsheets is clear, but the consequences will be more complex. While spreadsheets fragmented data and logic, agents fragment decision-making and workflow execution. Imagine thousands of digital colleagues driven by multi-agent orchestration, and the scale of risk grows exponentially.
The Future of Work: Avoiding Excel Déjà Vu
Gartner once estimated that shadow IT consumed 30–40% of tech spending, although it is often invisible to CIOs until something breaks. Shadow AI will be much harder to spot and much costlier when it fails
What can leaders learn from the spreadsheet era? Three lessons stand out:
- Balance empowerment with structure. Employees need freedom to experiment. But guardrails must exist from the start, not as a retrofit.
- Plan for integration early. Once agent sprawl takes root, it will be expensive and politically fraught to unwind. Build governance while adoption is still forming.
- Frame knowledge as collective. Position agents as company assets, not personal property. Without this framing, every future consolidation or change to the model will feel like a violation.
These lessons aren’t abstract. They’re the difference between AI accelerating your workforce and or fracturing it.
Leader Takeaways: Do These Now
- Invest in governance as design, not policing. Guardrails should be baked into the build process. Make it easy to stay compliant rather than punitive after the fact.
- Define ownership early. Who is accountable for an agent’s lifecycle, including its updates, monitoring, and retirement? If the answer is “no one,” the cracks are already forming.
- Audit emerging workarounds. Employees are likely already building agents in corners of the business you don’t see. Find them. The best should be promoted to enterprise tools with proper oversight. The rest should be controlled to minimize chaos.
- Reframe the narrative. Celebrate personal initiative, but make it clear: these are not private copilots. They’re shared infrastructure that strengthens the whole system.
Conclusion: The Future of Work Requires Foresight
Spreadsheets gave individuals unprecedented power, but the unintended consequence was organizational fragmentation that still haunts companies today.
In the future of work, AI agents will democratize not just data, but action. The stakes are higher, the risks are sharper, and the potential fallout will be faster. The choice is stark: repeat the mistakes of the spreadsheet era, or build a future where agents elevate both individuals and the enterprise.