Skills Intelligence Tells You Who Could Move. It Doesn't Move Them.

Start by giving the skills-based strategy its due, because it earned it. The shift from a rigid job architecture to skills-as-currency was real progress. A live inventory of capabilities across the workforce, a match score against an open requisition, a recommended learning path that turns a gap into a route — these are hard problems, and the category solved a meaningful part of them. For a talent leader trying to cut external hiring, shorten time-to-fill, and retain people by showing them somewhere to go, that visibility genuinely matters.

So this is not a piece about how the map is wrong. The map is right. The signals are useful. The intelligence, in many cases, is excellent. The argument is narrower and sharper than "the technology disappoints": the map's value proposition ends at the edge of knowing — and the thing you actually care about, the person landing in the new role, begins on the far side of that edge.

Watch what happens the moment a recommendation appears. The system surfaces a strong internal match and says, in effect, consider redeploying this person. Everything after that word "consider" is a manual, cross-system project:

• Open or reshape the role — in the requisition system, with its own approvals.
• Reassign the current work — backfill or redistribute what the person is leaving behind.
• Enroll the reskilling path — in the learning system, tracked separately.
• Adjust compensation — in the comp system, against bands and thresholds, with sign-off.
• Re-provision access and tools — new entitlements, new systems, a device for the new function.
• Change reporting lines and the manager's task list — in the org record, in facilities, in a dozen small handoffs.

Each of those lives in a different system, is owned by a different function, and is moved by a different human filling out a different form. The skills platform recommended in seconds; the redeployment now takes weeks and a project plan. The intelligence is real-time. The outcome runs at the speed of coordination between people who don't share a runtime.

This is the gap that caps internal-mobility rates in practice. Leaders can prove, with data, exactly who could move — and still watch the number stay stubbornly low, because the friction was never in the knowing. It was in the moving.

Here is the part worth being precise about, because it is structural, not a defect a better model will one day patch. The skills graph sits on top of a system of record for people data. Its job — its architecture — is to know and to recommend. "Unified," in that world, means unified data and a unified view within the HR domain: one place to see capabilities, history, and matches. That is real and useful. But the moment a recommendation must become an action in the comp system, the access layer, the learning platform, and the org chart, it crosses out of the domain that produced it — as a handoff. The record was unified. The work was not.

And credit where it is due to the most capable version of this pattern. The closest thing to execution today is the workforce cloud that natively spans HR and IT and can provision applications and devices as someone is onboarded. That is genuinely more than a record update — it is real automation, and a pure system of record cannot claim it. But provisioning inside one suite's own two domains is still records-plus-point-automations bolted to the record. A redeployment does not respect suite boundaries: it reaches into learning, compensation, access governance, facilities, payroll, and a manager's queue at once. Automating the slice that happens to sit inside the suite, and handing off the rest, is not the same as running the whole process as one governed thing. The line isn't "they can't automate anything." The line is that the automation stops at the suite's own fences, and the people process doesn't.

Entroid closes the gap by moving the target. On a Composable Process Fabric, talent mobility is not a recommendation that kicks off a project. It is modelled, executed, and governed as one thing — a Deterministic Workflow composed from five primitives, running in a single runtime:

• Intelligence Orchestration produces the signal — the match, the adjacency, the reskilling route — the same intelligence the category is good at.
• The Deterministic Workflow carries the move from end to end, with governance enforced inline: entitlements, approvals, segregation of duties, and comp thresholds are checked at the moment of the action, not reconstructed in a downstream review.
• Atomic Agents execute the steps, with human-in-the-loop as a first-class control wherever a person should decide rather than a system.
• Connectors — the only primitive that touches external systems — do the actual work against the platforms that already hold it: re-provisioning access, enrolling the learning path, writing the comp change, updating the reporting line.
• The Growth and Workforce modules sit on that same fabric, so the mobility and reskilling process is native to it, not a separate app stitched in.

The result is a change of verb. "We can see the skills" becomes "the person is redeployed" — reassigned, enrolled, re-provisioned, comp adjusted — as a sequence of governed actions with one immutable per-action audit spanning HR, IT, finance, and facilities. The recommendation and its execution finally share a runtime and a governance model.

Be clear about what this is not. It is not magic that erases the other systems, and it is not a claim of zero integration. ES runs over the existing estate through governed Connectors — the learning platform, the comp system, and the access layer are still there. The difference is that the integration is a governed primitive inside one runtime, not a handoff between owners who each do their part in their own tool and hope the whole thing reconciles.

The strategic point is not "our AI is smarter." The intelligence can be identical. What changes is that the recommendation and the redeployment live on the same fabric under the same governance, so the cost and latency of actually moving someone collapse — and the cost and latency of moving people is precisely what has been capping the return on every skills investment.

Consider, purely illustratively, a support engineer whose capability profile flags her as a strong adjacency for an open solutions-architect role. In the map-only world, that insight becomes a memo and a to-do list: someone opens the req, someone drafts the comp change and chases sign-off, someone requests the new access, someone enrolls her in the bridge learning path, and the calendar does the rest. In the fabric world, that same match enters a governed workflow: the entitlement and segregation-of-duties checks run inline at the gate, the comp adjustment clears its threshold with the right approval, Connectors provision the access and enroll the path, the reporting line updates — and every one of those actions lands in a single per-action audit. This is an architectural illustration of how the design behaves, not a reported deployment or a promised result — but it is what "one governed workflow" actually means in motion.

For the risk-minded reader at the table — the CISO worried about who now has access to what, the CRO who has to defend the org's controls — that audit is not a footnote. In the manual world, the evidence of a redeployment is scattered across every system that did its part, which is exactly why proving control over access and compensation changes is so painful. When the move is one governed workflow, the evidence is one record, per action, immutable. Governance stops being the tax you pay after the move and becomes the rail the move runs on.