Execution Speed Was the Moat. Learning Speed Is.

For twenty years, execution speed was the competitive moat. Ship faster, iterate faster, deploy faster. The companies that won were the ones that compressed the distance between decision and release.

A Robonomics essay this week argues that moat has quietly moved. The new one is learning speed: how quickly an organization can absorb what AI makes newly possible and restructure around it. Most companies miss this because they are doing what Henry Ford warned about a century ago. They are asking AI for a faster horse.

The Faster Horse Problem

A faster horse is an AI that writes the same emails, summarizes the same meetings, and drafts the same decks, only quicker. The work does not change. The org chart does not change. The meetings do not change. You just feed the old machine a new engine.

This is where most enterprise AI programs sit right now. They have measured productivity gains on individual tasks and declared victory. They have not asked the harder question: if this technology had existed when we designed the company, would we have designed it this way?

The answer is almost always no. And that gap between “the company you would design today” and “the company you actually have” is where learning speed becomes decisive. The companies redesigning from scratch are not 20% faster. They are structurally different.

The Translation Cost Collapse

The Robonomics piece offers a framing I keep coming back to. AI collapses translation costs between departments.

Think about what a traditional org chart is actually for. Marketing speaks in campaigns. Engineering speaks in tickets. Finance speaks in GL codes. Legal speaks in clauses. Every handoff between these languages used to require a translator: a project manager, a BA, a director who could speak two dialects, a meeting where everyone aligned on what “done” meant. The hierarchy existed because translation was expensive and someone had to pay for it.

AI makes translation cheap. A well-instrumented agent can read an engineering PR, produce a marketing-ready release note, generate the legal review checklist, and reconcile the finance impact in one pass. The translator role does not disappear. It becomes ambient.

When translation costs fall, the communication layers those costs justified stop making sense. That is the structural shift. It is not “we have better tools.” It is “the reason this box existed on the org chart is gone.”

What Replaces the Boxes

Four structural moves show up in the companies actually redesigning:

Sequential work becomes simultaneous. Instead of marketing waiting on engineering waiting on legal, cross-functional squads of three to five people own a problem end-to-end. They have AI as the translator between disciplines they themselves do not fully speak.

Functions become capability atoms. The old silos were nouns: Marketing, Engineering, Finance. The new units are verbs: Acquire, Build, Price, Retain. Each atom has humans and agents working against the same goal, and the atom is small enough that nobody forgets what the goal is.

Quarterly launches become continuous deployment. Not just for code. For policies, pricing, positioning, and process. When you can test a change in a day instead of a quarter, the org structure that relied on quarterly rhythms becomes a tax.

Committee-driven roadmaps become systems-driven ones. The Robonomics example is a fintech that built automated intelligence to detect merchant cash-flow stress and auto-offer short-term financing. No product committee meeting. No quarterly review. The system reads the signal, the offer goes out, the humans supervise. That loop is the roadmap.

The Middle Management Question

Every essay on org redesign eventually runs into the same awkward question. What happens to middle management?

The honest answer, and the one the Robonomics piece makes explicitly, is that most middle management work was information routing. Taking updates from below, summarizing for above, taking priorities from above, translating for below. If AI collapses translation costs, most of that work is gone.

The managers who survive are the ones who were never really routing information in the first place. They were doing three other things: judgment on hard calls, coaching on craft, and navigating ambiguity when the system cannot. Those jobs are not going anywhere. The job title “manager of 12 people doing the same thing” is.

This is not a layoff memo disguised as an essay. It is a reframing. The middle management compression is not about cutting headcount. It is about recognizing that the role was always two jobs in one suit, and AI just unbundled them.

The Learning Speed Diagnostic

Here is a test you can run on your own organization this week.

Pick a decision your company made in the last quarter. Trace it backward. How many meetings, documents, and handoffs were required to make it? How much of that was substance, and how much was translation? If you pull out the translation, what is actually left?

For most companies I have worked with recently, the answer is uncomfortable. Sixty to eighty percent of the effort was translation. The substance fit on one page.

That ratio is the learning speed problem. Execution speed tells you how fast you can run the current loop. Learning speed tells you how fast you can change the loop itself. A company that spends eighty percent of its cycles on translation cannot change the loop at all. It is too busy maintaining it.

What to Do Monday Morning

Three moves, in order.

First, find the translation work in your org. Not the obvious kind. The meetings that exist because two departments do not speak the same language. The documents that exist because a decision had to be handed from one layer to another. Make a list. It will be longer than you expect.

Second, pick the smallest translation loop you can instrument. Not the biggest. The smallest. Replace the translation layer with an agent that both sides can trust. Measure the latency before and after. This is your proof point.

Third, and this is the one most companies skip, redesign the org chart around what the instrumented loop makes possible. If the agent now handles the translation between engineering and marketing for release notes, the person who used to own that handoff is free. Give them a new job that uses judgment, not routing. Do not let the saved capacity quietly refill with meetings.

The Faster Horse Is Still a Horse

The companies still treating AI as a productivity tool are running faster races on the same track. The companies treating it as an org redesign trigger are building a different track entirely.

Both approaches produce measurable gains. Only one of them produces a moat. The first one plateaus the moment the tools commoditize, which is already happening. The second one compounds, because every structural change you make teaches you how to make the next one faster. That is what learning speed actually means. Not “we learn quickly.” It means “the cost of the next restructure is lower than the last one, because we got better at restructuring.”

A companion piece, 99.5% AI Adoption at a $32B Company, documented what one company’s org-design-first approach looks like from the inside. The Robonomics essay zooms out and tells you why the approach works structurally. Read them together. The first is the case study. The second is the physics.

---

Source: Robonomics, “Org Design in the Age of AI” (April 2026). The piece argues that learning speed has replaced execution speed as the primary competitive advantage, and that AI’s collapse of inter-departmental translation costs is the structural force dissolving traditional hierarchy. This essay extends that framing with the translation-cost diagnostic as a concrete governance tool.

Victorino Group helps leaders redesign orgs around AI-era translation costs, not just AI-era tools. Let’s talk.