Companies that recognize they're stuck at Stage 2 of the Harness Map but try to cross the cliff by buying more Stage 2 tools. Why this fails, why it's the most common failure mode we see in mid-market AI investment, and what crossing the cliff actually requires.
A specific failure mode shows up in mid-market AI investment with enough regularity to deserve a name. The Stage 2.5 trap, after the Harness Map cliff it tries to climb without crossing.
The shape of it: the leadership team has read the Harness Map. They recognize the firm is at Stage 2 — a handful of point AI tools, some individual time savings, no workflow that's been redesigned around AI. The diagnosis is right. The next move is wrong. Instead of crossing the cliff to Stage 3 (workflow AI), they buy more Stage 2 tools. Six months later they have twice as many subscriptions, a higher monthly bill, and the same operating model.
This post is about why the trap happens, what it looks like in operating terms, and what crossing the cliff actually requires.
Buying a tool is a single procurement decision. Sign the SOW, run the deployment, train the users, watch the time-savings dashboard. The decision is reversible (cancel the subscription) and the upside is legible (hours saved per week). On a quarterly cadence, the we're doing more AI line in the executive update reads as a series of wins.
Crossing the cliff is a different category of decision. It requires committing a senior operator's time for a quarter — usually the COO, the VP of Operations, or someone with explicit authority over the workflow being redesigned. It requires writing a redesigned process before any tool is bought. It requires a measurement framework that didn't exist before. It requires the team running the workflow to accept that their day-to-day will change.
The procurement decision is simple. The redesign decision is hard. Most companies, faced with the choice, do the simple thing repeatedly and call it progress.
Two patterns recur:
The tool stack as a substitute for the redesign. The team subscribes to a writing assistant, a meeting summarizer, a research tool, a data-prep tool, a slide-generation tool, a customer-research tool. Each saves time in its own pocket. None of them touches the workflow that connects the pockets. The team feels more productive day-to-day. The throughput of the overall workflow is unchanged because the bottleneck was never in any of the pockets — it was in the handoff between them.
The tool stack as compliance theater. The executive sponsor needs to show AI activity on the quarterly board update. The procurement of additional tools provides the activity. The number of tools rises. The dollar spend rises. The narrative on the slide reads well. None of the tools are integrated into a measured workflow improvement. The investment is real; the leverage isn't.
In both patterns, the team is genuinely working in good faith with the resources available. The failure isn't of effort. It's of strategy.
Crossing from Stage 2 to Stage 3 is not a procurement event. It's an organizational event. The minimum components:
An owner with authority. Someone with the political and operational authority to redesign the workflow end-to-end. Not the IT manager, not the VP of innovation, not a consultant on retainer. The person who runs the workflow now and who will be accountable for the redesigned version.
A dedicated calendar window. Crossing the cliff doesn't happen in the margins of someone's existing job. The owner needs a quarter where the redesign is their first-priority work. If the firm can't afford to pull the owner off other duties for a quarter, the firm cannot cross the cliff this year. That isn't a moral failing; it's a constraint that needs to be named.
A redesigned process, in writing, before tools. The redesign comes first. The tools come second. A workflow redesign that starts with the question what AI tool should we buy is putting the cart before the horse.
A measurement framework. Define the metrics that will tell the team whether the new workflow is better than the old workflow, and instrument them before the redesign goes live. Most Stage 3 failures fail because they were never measured against a real baseline.
These four components are not optional. Stage 2.5 is the state of having one or two of them; Stage 3 is the state of having all four.
Three moves, in order:
Stop buying tools. Whatever Stage 2 tools you have are enough. Adding more will not close the gap. The next dollar should go to redesign capacity, not new subscriptions.
Name an owner for one workflow. Pick one — not three, not five. One workflow. Assign a named owner with a quarterly mandate to redesign it. Communicate the mandate to the team that runs the workflow today.
Define what done looks like. Before the redesign begins, write down two or three outcome metrics that will determine whether the redesign worked. Baseline them at current state. Then start.
These three moves don't require an external consultant, and they don't require a tool budget. They require leadership commitment. The companies that take these three moves cross the cliff; the companies that don't, accumulate Stage 2 subscriptions and a year later wonder why their AI strategy hasn't produced operational change.
If you want to walk where you actually sit on the Harness Map and what crossing would look like for one of your workflows, the Readiness assessment is a self-administered version of the diagnostic we run. The discovery call is the next step if the diagnosis comes back ambiguous.