Skip to content
Fritz Gerald ZephirinMenu

Social

Redesign the Workflow or Skip the Tool

3 min read

There is one sentence in our purchasing process that has saved more money than every negotiation tactic combined: show me the step this removes.

The rule exists because of a pattern every operator will recognize. A team requests an AI tool. The demo is impressive. The pilot is “successful” — meaning people used it and said nice things. Six months later, the workflow takes exactly as long as before, costs slightly more, and the tool’s strongest defender can’t point to a number that moved. Multiply this by an economy and you get the statistics this site keeps returning to: 95 percent of pilots with no P&L impact, nine in ten executives reporting nothing.

The mechanism of the failure is the oldest one in technology economics. Factories bolted electric motors where the steam engine had been and gained nothing for forty years, because the layout — the thing that actually determined throughput — never changed. A workflow is a layout. If the AI drafts the email but the same three people still approve it, you’ve added a license fee to an unchanged pipeline. If the bot answers the ticket but a human still reviews every answer because the queue routes everything to review, you’ve bought a motor and kept the shaft. Work flows at the speed of its narrowest step, and a tool that doesn’t move the narrow step moves nothing — visible effort, invisible throughput.

So the rule, in full: no tool above a trivial cost threshold is purchased unless the purchase doc contains a before-and-after workflow diagram in which at least one step is deleted or one handoff is redrawn. Not “made faster” — deleted or redrawn. “This makes step four easier” is the tell of a bolt-on; “there is no step four anymore” is a redesign. The diagram takes an hour to draw and is the cheapest diligence in business: if nobody can draw the after, the organization is not buying a tool, it’s buying an alibi.

Three supporting disciplines make the rule work in practice:

The process owner runs the pilot, not the enthusiast. Enthusiasts produce successful pilots by definition — enthusiasm is the variable being measured. The person who owns the workflow’s output owns the trial, and a skeptic’s sign-off is worth ten advocates’.

Baseline first. Two weeks of honest before-measurement — volume, cycle time, quality sample — is the price of admission. Without a before, every after is a vibe.

The redesign is scheduled, not promised. “We’ll change the process later” is how the line shaft survived four decades. The workflow change ships with the tool — same project, same deadline — or the tool waits. Later never comes; we stopped pretending it does.

The honest boundary of the rule: it has a floor. Spellcheck-class tools — cheap, individual, no workflow footprint — don’t need diagrams; applying the rule to a $20 subscription is theater. The rule activates where cost or criticality makes failure expensive: anything touching customers, anything with per-seat pricing at scale, anything with agent-grade permissions. And occasionally the rule’s logic runs in reverse, which is its most useful trick: sometimes the diagram session reveals the step could be deleted without buying anything — the after-picture shows the before-picture was the problem all along. When that happens, the purchase dies and the workflow improves anyway: the rule paying for itself in the most satisfying possible currency, zero dollars, spent well.

What the rule really enforces is a change in the question. “Should we buy this AI tool?” invites a technology evaluation, and the technology is usually fine. “What does our work look like after?” forces an organizational decision — the only kind that the evidence says actually produces returns. Tools don’t transform companies. Redesigns do. The tool is just the occasion.

Sources & further reading