The Strassmann Re-Correction
Why AI Value Remains a Management Challenge
The Strassmann Re-Correction
Why AI Value Remains a Management Challenge
In 1990, Paul Strassmann introduced a groundbreaking concept with his book The Business Value of Computers. He presented clear data demonstrating a lack of correlation between IT spending and profitability. Strassmann argued that computers don’t create value; rather, managers do. He posited that technology merely serves as a “force multiplier” for existing management quality.
Fast forward to 2026, and we’re no longer merely purchasing hardware; we’re deploying autonomous multi-agent systems. Yet, we’re encountering the same fundamental issue. AI permeates every aspect of our lives, yet it’s absent from productivity statistics. This is the Productivity Paradox 2.0, and resolving it necessitates revisiting Strassmann’s most crucial metric: Return on Management (ROM).
The Trap of the “Blunt End”
Strassmann’s central thesis centred on the distinction between the “sharp end” (revenue-generating production) and the “blunt end” (administrative management). He observed that a significant portion of IT spending was absorbed by the blunt end, creating a feedback loop of “information for information’s sake.”
In the Agentic Era, this trap has expanded. Companies are currently fixated on generating volume. They’re leveraging large language models (LLMs) to produce unprecedented amounts of code, emails and internal reports. However, if this output fails to directly enhance the output of the “sharp end,” it becomes economically detrimental.
Every agent you deploy incurs a “Management Cost.” This cost arises from the need to supervise, verify its output and coordinate hand-offs. If your AI implementation increases this coordination overhead faster than it generates market value, your Return on Management (ROM) decreases. In Strassmann’s view, you haven’t purchased a solution; you’ve acquired an expensive layer of chaos.
Return on Management (ROM) as the North Star
To determine the business value of AI, we must abandon the focus on “cost savings” and embrace Strassmann’s ROM formula.
By 2026, Management Value Added will be defined by Decision Velocity and Information Integrity.
The essence of a successful AI strategy lies not in automating “how many tasks” but in “how can we leverage agents to reduce the cost of reaching a correct decision?”
For instance, if an agentic workflow summarises a meeting that should have been an email, the ROM is zero. Conversely, if the same workflow identifies a supply chain bottleneck and proposes a pre-verified alternative before a human even notices the delay, the Management Value Added is substantial. The key is transforming information into actionable intelligence, not merely rearranging text.
Transitioning from Generative to Orchestrative
Strassmann famously cautioned against “paving the cow paths” — digitising a mess rather than redesigning the process. To avoid this with AI, we must shift from ad-hoc “chat” interfaces to structured Agentic Design Patterns.
1. Shift from “Chat” to “Script”
Value lies in the precise scripting of multi-step systems. Successful managers don’t simply ask models for help; they design orchestrated workflows where agents function as dependable components within a larger machine. This is the core of Vibe Engineering: the disciplined translation of human intent into programmatic execution.
2. The Minimalist Mandate
Strassmann discovered that the most successful companies were often the most frugal with IT, deploying it only where it offered a significant advantage. In 2026, this translates to practising Systemic Minimalism. Avoid deploying a swarm when a single well-prompted agent suffices. Every unnecessary agent increases the denominator of your Return on Machine (ROM).
3. Focus on the “Sharp End”
Channel your agentic resources towards customer-facing outcomes. Leverage AI to bridge the gap between customer needs and your company’s fulfilment. If your AI doesn’t interact with the product price or delivery, it’s likely just “blunt end” noise.
4. Audit for “Workslop”
Just as Strassmann audited for “unproductive information,” modern managers must audit for “AI slop.” If your team spends more time “fact-checking” agent output than they did previously on the actual work, your technology is hindering productivity.
Conclusion
Paul Strassmann’s insights were correct thirty-five years ago, and his logic remains the only solution to prevent the AI bubble from bursting today. When you apply a multi-agent system to a disorganised company, you simply get a disorganised company operating at machine speed.
The business value of AI isn’t a technical attribute of the model; it’s a result of the management quality behind it. To succeed in 2026, stop being a consumer of AI and start being an architect of Management Value.