The Frontline Dispatch

Making Good: solve one, you get a demo. Solve both, operations.

Written by Thor Hunter | Oct 21, 2026, 12:00:00 PM

Making Good, part 2 of 4. Part one starts here.

Since the Capgemini announcement, two questions have come up in almost every conversation I've had. Both are fair, and both point at the same two failures.

Frontline AI breaks in one of two ways. The agent never reaches the worker, or it reaches them without knowing the organization. Solve one and you get a good demo. Solve both and you get operations.

How is this different from Copilot?

Copilot is good at the job it was built for: a person with a Microsoft 365 seat, a laptop and an inbox. Most frontline workers have none of those. The housekeeper on floor twelve has a phone in her apron and a few minutes between rooms. An assistant she never opens can't help her, however good its answers are.

That is a reach problem. Enterprise AI arrives through the tools the enterprise already uses at the desk, and the frontline was never on those tools. By industry estimates, the 2.7 billion people who work without a desk have received about 1% of enterprise software spend (Emergence Capital; Gartner). AI inherited that distribution.

Our HR platform already sends messages. Why add anything?

It does send them, one way, from the top, to everyone on the list. A notification is useful. It doesn't create participation.

Participation needs access rules that move as fast as the schedule. Who is on this shift tonight. Who covers the east wing when someone calls off. What each person is allowed to see. Those rules change every shift, and a frontline worker who opens the HR suite about once a month to check a paystub won't feel them.

That's why RedeApp starts with the communication layer. Across RedeApp networks, 44% of monthly users come back every day, against a 13% industry average (RedeApp network telemetry, 2026). Daily use is what everything else gets built on.

What does a frontline AI need underneath it?

A record of the frontline. That's the grounding half. An agent is only as trustworthy as what sits under it, and for a shift worker that means four things: who works here, what each person may see, what the organization has actually published, and what happened on the last shift.

Shelbe answers from that record. Every answer cites its source and stays inside the asker's permissions, which come from the same identity data that governs every message in the network. Anything consequential waits for a person.

Without the record, you get a chatbot guessing, politely, at answers the organization already wrote down.

Do we have to replace what we already run?

No. The HR suite still runs payroll and the scheduling system still owns the schedule. RedeApp's App Hub keeps workflow ownership in Workday, UKG, SAP or ServiceNow while RedeApp carries the engagement, and AutoSync keeps people data current from whichever HRIS an organization uses.

This is where Capgemini matters. Most large organizations already run an enterprise program around those systems. The alliance brings the frontline into that program instead of building a second one beside it.

The paystub can stay where it is.

Next Wednesday, part three: the four questions your counsel will ask before any of this goes live.