On a stormy night, three days after Christmas in 1879, the Tay Bridge, once hailed as an engineering marvel and the longest bridge in the world at the time, collapsed into icy waters near Dundee, Scotland. A passenger train was crossing at the moment the High Girders gave way, killing everyone on board.

Prior to building, the bridge's engineer, Sir Thomas Bouch, had consulted experts on wind loading and concluded that no explicit allowance for wind pressure was necessary, a judgment the subsequent inquiry would find catastrophically wrong, exposing flawed intuition, absent records, and overlooked risks.

The Tay Bridge disaster became a foundational case study for modern engineering ethics and standards. It directly led to the establishment of rigorous wind-loading regulations, shifting engineering from "expert intuition" to a world of auditable decision infrastructure.

We no longer build bridges based on what an individual engineer feels is sufficient. We build them on collective, verifiable records of environmental data, material science, and safety margins, tested, documented, and independent of any single judgment.

The Modern Parallel: Today's Procurement

Today, enterprise software procurement has not made that transition. Companies still spend millions on business-critical software without a structured system to capture why those decisions are being made. The reasoning, the trade-offs, the rejected alternatives, the risks consciously accepted. None of it survives the decision.

Take a team evaluating three vector databases for a new AI project. The stakes are tangible. The latency benchmarks, the embedding quality, the trade-off between Vendor A's security ceiling and Vendor B's cost: that is all real, rigorous work. The conclusion that Vendor A is the only one capable of handling edge cases in a secure deployment is hard-won intelligence.

And then it evaporates.

Into a dead Slack thread, an archived email, the fading memory of a lead engineer who eventually leaves. When renewal time comes, the context is gone. Institutional knowledge resets to zero. No flywheel, no learning. Just a cycle that repeats without ever getting smarter.

The Workflow Trap

In the last decade, enterprise software underwent a massive operational overhaul. Companies like Zip, Oro Labs, Coupa, and Ariba won by perfecting the how of procurement: the Intake-to-Procure workflow, approval routing, procurement orchestration etc.

The train moves efficiently across the bridge. But moving the train was never enough.

What is missing is the Decision Infrastructure; this is the equivalent of the wind-loading reports and material tests that justify the purchase in the first place. Modern procurement tools have perfected the process of moving a decision through an organization. None of them capture the reasoning behind the decision itself.

The HOW is solved. The WHY is not.

The consequences of this gap are already visible. Months after a software purchase, the same questions resurface. During renewals, audits, or company milestones.

• How did we decide on this?
• What were we trying to solve?
• Did it deliver?

In most companies, no one can answer with confidence.

This is not a failure of process. The process ran. The approval was routed. The purchase order was raised. The contract was signed. Every step of the procurement workflow completed exactly as designed.

What failed was never part of the design.

No system captured why this vendor over that one. No record exists of what was rejected and why. No artifact documents the risks that were consciously accepted, the assumptions made with and without evidence, or the outcomes that should have been committed to and tracked.

The decision happened. The reasoning behind it did not survive. This is an infrastructure gap. The same infrastructure gap that brought down the Tay Bridge.

Decision Infrastructure

Decision infrastructure is the set of systems, standards, and data that allow organizations to make repeatable, high-stakes decisions with clarity, accountability, and memory.

We are moving towards a world where AI agents will handle procurement autonomously: Evaluating vendors, issuing RFPs, running negotiations, and increasingly, making decisions. But an AI agent cannot reason from a marketing brochure or a generic review. It cannot operate on intuition.

AI agents need context, standards, and decision primitives: structured, verified data points on how companies solved similar problems, under what constraints, with what outcomes, what was rejected and why.

Software spend is now capital allocation. And as capital allocation scales, it demands increasingly structured and defensible reasoning behind each decision.

Decisions must be:

• Defensible (The Audit Trail),
• Durable (The Lifecycle Record)
• Re-usable (The Compounding Effect)

In the last decade, we built magnificent rails to move software purchasing decisions through organizations. We perfected the how. We made the train run on time.

We never built the wind-loading report. The storm is coming.

That is the gap. Merkville exists to close it.