AI Agent Program for a Resource Optimization Center

From WFM Labs

Part of the Planning Week chain · previous: Systems Administration for WFM Platforms · next: The Agent Team Ladder: Alpha to Production AI agent program for a resource optimization center is the program by which a workforce planning function introduces agent teams into its own recurring work, one book of business at a time, under a charter that names one owner, a first ninety days, its dependencies and two measures. The program is the delivery vehicle; the teams, their ledgers, clocks and gates, are described on AI Agent Teams for Workforce Management and are not restated here. The page's claim is that a function which charters agents as a program, with a ladder, a gate that stays on and two measures that are not headcount, gets evidence it can plan on, where a tool gets a demonstration. The page produces the agent program's charter row, template T7 (PG-007 on the charter page), and is worked in a session with Wiki:Packs/Agent Program Readiness (CP-OPS-005).

The subject is the planning function's own agents; agents that handle customer contacts are on Agentic AI Workforce Planning and Workforce Planning with AI Agents, and the routing layer between those agents and people is on AI Agent Orchestration for WFM.

Why a program and not a tool

A function that has read The Agent Team Model can build a first team in weeks. What it cannot do in weeks is know what the team is worth, because what a planner-hour covers today is the measurement A Roadmap Pattern for Agent Teams in WFM says almost no function has made. A tool is adopted; a program is measured. The program form exists so that the measurement is taken before the claim is made, the gate is held by a named seat rather than by habit, and the second book is entered by a checklist rather than by enthusiasm. The form is the ordinary program charter on Function and Program Charters, held in a portfolio per Program and Portfolio Management, with an owner as a seat: a post on the org chart with its scope and its sunset condition, never a name.

The charter

Field The agent program's entry, generically
Objective Run an agent team on one real book under a planner gate at every transition that changes a plan, daily clock first and monthly clock second, so the function learns what the team covers before it plans on the team covering it
Scope in One book; the planning-loop roster on the daily and monthly clocks; the intake door; the ledgers of Living Ledgers; file adapters to the short-term WFM engine and the long-term capacity engine; every number graded; the ladder on The Agent Team Ladder: Alpha to Production
Scope out More than one book until the clone checklist passes; replacing the planner at any rung; any action on the intraday automation platform. The scheduling and real-time rosters are in scope on their own cards and their own clocks, and they may open before the planning loop's exit evidence is complete, provided the acceptance is recorded — the order and its price are on A Roadmap Pattern for Agent Teams in WFM
Owner One seat, a planning director with an analytics-and-automation transformation track. The method (the gate, the grades, the ladder's exit evidence) is held by the seat that owns placement, not by the program
First ninety days Choose the book (D-13); populate its ledgers from the platforms' exports; run the daily clock with the gate on and publish the first catch rate
Dependencies The Definitions Register; the first-wave L2 tables for the steps the agents run (Process Shells for a Workforce Standard); a file interface to the short-term WFM engine or the standard instance's adapter (Technology Migration Plan for a Workforce Function)
Decision needed Which book carries the pilot (D-13). Carried and not voted: the planner gate stays on; the program is judged on detection and catch rate, never on headcount
Measures Days-to-detection of a level shift; catch rate per process per period with its injection rate and sample size; register questions answered as graded cards inside two business days; carried assumptions labeled, all of them
Risks Judged on hours freed before it is judged on detection; a book with two definitions of one metric; vocabulary adopted without practice change

The owner is one seat because a program with two owners has none; the method is held elsewhere because the seat that runs the pilot has an interest in its passing. The measures are two because they fail differently: a team can name a shift on the second day and still miss the errors an overseer injects, and a well-checked team can still be slow.

The claim, stated with humility

The program's founding claim, as sponsors usually put it, is that agent teams compress the planner-hours spent on written-down work by a large multiple and lift the function a maturity level. The program's contribution is to make that claim gradable rather than to repeat it: it enters the function's register at grade Asserted on the day the charter is accepted, with the measurement named beside it and a date. Until the measurement exists, every hours claim is [A] and every plan built on one is [A]. The planner-hour measurement itself, what it logs and why the ordering of the work follows from it, is on A Roadmap Pattern for Agent Teams in WFM and is not restated here; so is the field evidence on where generative assistance helps most.

What a chartered program can state early is narrower and better than an hours figure, and it is what the charter is judged on: that the team runs every step of the process's L2 table every day, and that the grade profile of the daily note moved from one share of [M] figures to another. Both are read from the ledgers.

A second claim needs the same care. The Short-Term Forecasting Loop with an Agent Team shows a control chart on daily handle time flagging a level shift on the third day after a go-live, where the same shift read from a monthly trend line is not yet visible; that page calls it an illustration of the mechanism rather than a measured result, and this page keeps that grading. The days-to-detection measure exists so that a deployed team's number can one day be stated with a grade of its own.

How the program lifts the function

The program is two moves, and the steps it moves between — assistant, team, ecosystem — are defined on Agent Team Readiness for a Planning Process rather than here. The first move, Level 2 to Level 3, runs the assistant step: the program's contribution at that step is not the assistant but the discipline around it, that its output carries a grade and that its definitions are the register's, which is what makes the step worth more than the time it saves. The second move, Level 3 to Level 4, is the automation layer applied to the function's own work: once a process passes the readiness check it runs on the team step, and the ledgers it writes supply the provenance a distribution is planned on, which is the Level 4 threshold. The release condition for gate-free execution is stated in evidence terms on Human Gates and Number Grades.

Its part in resizing

The program is the third of three levers on Resizing a Workforce Function After an Acquisition, third in order and not in importance. An acquired estate arrives with its own planning function, each part correct for the business it served; size follows the estate through method first (one standard), platform second (one configuration) and agent teams third (written-down work handed over), never through a headcount target set first. The program's contribution is read from evidence, the planner-hour log before and with the team running, and a headcount consequence in that log is an outcome the function's leaders decide on, with The Hardening Residual's warning that the remaining work per planner is harder than the average that left. The charter carries no headcount number, no seat count and no transition chart; where the room must say what happens to a post, roles evolve, per Role Evolution in the Resource Optimization Center, and new job classes open, per The Automation Analyst and The Agent Overseer.

The three rosters, by clock

The program brings the series' three rosters into a function on three clocks and never folds them together.

Roster Clock Enters the program at Gate that governs it Page
Planning loop Daily, weekly, monthly, plus the intake door First, on the daily clock; the monthly clock joins when a plan of record is signed Planner gate; plan signature; register review The Short-Term Forecasting Loop with an Agent Team · Long-Term Planning Agents and the Plan of Record
Scheduling Weekly and per bid After the planning loop's gate is running, on its own card and its own rung Publication gate Scheduling Agents
Real time Intraday Last, writing only; action proposals go to the action gate Action gate Real-Time Agents

The order is the roadmap pattern's: the planning loop first because its failure is contained and its planner is the corrector; scheduling and real time last because that is where an error reaches a worker or a customer fastest. The order is a sequence, not a queue. A roster may open while the roster before it is still earning its exit evidence, and A Roadmap Pattern for Agent Teams in WFM's worked example does exactly that: the function accepted that a scheduling checker's first weeks would run beside a loop whose catch rate was one period old, and recorded the acceptance. The program's rule is that the acceptance is written down with its price, never that the later roster waits.

Governance

  • The planner gate stays on, and it is a program commitment rather than a stage. The five run-time gates and the condition on which any of them is released are stated once on Human Gates and Number Grades; what the program adds is that release is never entered on a roadmap as a milestone with a date. Keeping a gate that asks the signer to produce something rather than to click is what keeps it from decaying into the over-reliance human-factors research documents when the human is left to monitor rather than to act.[1]
  • Grades are enforced where the work happens. An ungraded cell is rejected at load and a carried assumption is relabeled rather than re-presented as measured, per Living Ledgers; that a figure carried across a channel, platform or cohort change is [A] in the new state is the rule on Human Gates and Number Grades, and a migrating book meets it on every figure.
  • Every agent has an owner. Each agent carries an identity that resolves to an accountable person, a permission set derived from the processes it is authorized to run, and a measured reputation, per Agent Identity and Custody; the overseer of The Agent Overseer runs the catch-rate program.

The program is costed the way Automation Economics and ROI Decision Frameworks asks any automation to be costed, on full cost accounting rather than the primary reduction alone: the scaffolding, the overseer's time and the exception queue on the cost side, measured coverage on the benefit side.

Worked example

The series' example function, roughly 40–50 planners [E] across about a dozen books, runs its planning week Mon 20 Apr – Thu 23 Apr 2026, and the charter is written around work that is already running rather than around a start date. Its planning-loop team has been on the migrating corporate client book since Mon 2 Mar 2026 with the planner gate exercised on every run; by Fri 3 Apr a first catch rate was published — 9 of 10 known errors detected, from 6 injected and 4 found by deep verification of a 5 percent sample [C from the overseer's log] — and the monthly clock signed a plan of record on Fri 27 Mar. A scheduling checker has run since Thu 26 Mar under a publication gate and a real-time issuer since Wed 8 Apr under an action gate, both earning catch rates of their own.

On Day 4 morning the room therefore charters a program that exists, not one that starts. It accepts seven program charters, the agent program's among them (PG-007); it records the founding hours claim in the register at grade Asserted with the planner-hour log named as the measurement that would move it; and at D-13 it takes the next rung for each of the three teams and the second book the planning loop clones onto, choosing one with its own migration window so the clone's carried benchmarks are tested rather than assumed. The risk is recorded beside the choice: a book crossing a platform change carries every figure at [A] in the new state. The first ninety days run to Tue 30 Jun 2026. The dates are the example's and are illustrative.

The artifact this page produces

Program charter (T7), the agent program's row. The program number is assigned with the other charters on Function and Program Charters, where the agent program is PG-007. One filled example row:

ID Program Objective Scope in Scope out Owner (seat) First 90 days Dependencies Decision needed Measures Risks Lanes Arc
PG-007 Agent teams for the planning function's own work run the three rosters on one real book with a planner gate at every transition that changes a plan, and clone onto a second one book on three clocks; the intake door; the ledgers; file adapters; the ladder cards more than two books; replacing the planner at any rung; any action on the intraday automation platform the planning director seat with the analytics-and-automation track; method held by the placement seat set the next rung per roster; run the clone checklist on the second book; publish a catch rate per roster the definitions register; first-wave L2 tables; a file interface D-13 (the second book, and the next rung per roster) days-to-detection; catch rate (proportion, injection rate, n); cards inside two business days judged on headcount; a book with two definitions of one metric; vocabulary without practice L-001, L-002 Arc 1

Produced in a working session with Wiki:Packs/Agent Program Readiness (CP-OPS-005); the filled set is part of blueprint v0.1.

What would change this

The claim is that the program form, with a ladder, a gate that stays on and two non-headcount measures, produces evidence a function can plan on where tool adoption does not. A function that adopted an agent team as a tool, without a charter or a ladder, and produced a catch rate and a planner-hour log of the same quality inside the same period would show the form to be overhead; functions that measured hours first and still kept their gates and catch rates would weaken the measures rule. Neither observation exists at the time of writing.

How this connects

Maturity Model Position

The program is chartered at Level 2 or Level 3 on the WFM Labs Maturity Model™ and is the vehicle by which a planning function's own work reaches Level 4: assistants and the standardization they force at Level 3; the automation layer, a team on a clock over ledgers under gates, at Level 4. It stops before Level 5 by design; the evidence its gates and measures produce is what a function would use to decide whether to go there. Four scales on this wiki use the word "level"; Planning Week for a Workforce Function states which is which. This page uses the maturity Levels 1 to 5.

See Also

References

  1. ↑ Parasuraman, R., & Riley, V. (1997). "Humans and Automation: Use, Misuse, Disuse, Abuse". Human Factors 39(2), 230–253. doi:10.1518/001872097778543886.
Retrieved from "https://wiki.wfmlabs.org/w/index.php?title=AI_Agent_Program_for_a_Resource_Optimization_Center&oldid=4739"