Agent Team Readiness for a Planning Process

From WFM Labs

Agent team readiness for a planning process is the check a workforce function runs on one process, the short-term loop, the monthly cycle, the schedule check, before handing it to an agent team. It asks four questions: are the definitions the process uses in a ledger; is the process documented to L0–L3; does it run on a stated cycle with a calendar; and is there a register for the questions it will raise. The page's claim is that these four are the whole of "standardize, then automate" at the altitude of one process, that they cross-walk exactly to the wiki's function-level standardizations and handover tests, and that a process failing any of them can still be worked with an assistant but cannot be run by a team. This page gives the check, the cross-walk, the steps from assistant to team to ecosystem, and the common ways a function fools itself about readiness.

Why a per-process check

Standardize Before You Automate states the principle for a whole function assembled from several heritages: six things standardized, the stage ended by the first handover. The Agentic Handover Gate gives the five tests one process passes. Neither says what a planning team should look at on Monday morning when someone proposes handing the reforecast to agents. This page is that list. It is narrower than both, and it is meant to be run in an hour on one process, by the planner who owns it.

The reason it is short is that an agent executes what is written down and nothing else. A process with a definition the agent cannot cite, a step it cannot find in a table, a cycle it cannot read from a calendar, or a question it has nowhere to file will not fail loudly; it will run on whatever the specification's author assumed, and the person is left with the exception nobody specified, the ironies of automation seen from the other side.[1]

The four questions

# Question Passes when Fails when Cross-walk
1 Definitions. Is every metric the process reads or writes a record in the definitions ledger, with formula, source and owner? Every column in the process's inputs and outputs cites a record; two records with one name do not exist A column called AHT with no record; "everyone knows what handled means"; a report and a platform that disagree by a definition Standardization 5 (data) on Standardize Before You Automate; test 1 of the handover gate needs it
2 Process. Is the process documented to Process Decomposition (L0–L3) with an L2 step table whose tools column names the systems each step touches? An L0 identity with an owner; an L1 flow; an L2 table with owner, tools, duration and approval per step; L3 job aids named in column 11 The process lives in one analyst's spreadsheet; the "step table" is a checklist without owners or tools Standardization 2 (process); test 1 (documented) of The Agentic Handover Gate
3 Cycle. Does the process run on a stated clock with a calendar, and does the calendar say what is due when and what waits for what? A daily, weekly or monthly clock with business-day offsets, as Capacity Planning Cycle gives for the monthly one; the process's place on it is written The process runs "when the data arrives" or "when finance asks"; nobody can say what yesterday's run consisted of Standardization 6 (interfaces with named owners and cadences); the clocks of The Agent Team Model
4 Register. Is there a register with an intake discipline where the questions the process raises are opened, graded and filed? A register with permanent IDs, a hypothesis table and graded findings, per The Question Register and Knowledge Base; someone owns it Questions are answered in email from memory; the same question is answered twice in a month with different numbers Standardization 4 (an owned objective register) is the sibling; Executive Issue Register is the precedent

A process that passes all four is a candidate for a team. It then goes through the handover gate's remaining tests, instrumented, measured here, overseen first, reversible and logged, which this page does not restate. A process that passes questions 1 and 2 but not 3 and 4 is in the overseen band of Three Bands of Work at best: an assistant can help a planner run it, and the planner is the clock and the register.

From assistant to team to ecosystem

AI Leverage Maturity in WFM Teams gives a four-level ladder of a planning team's own AI practice: no AI, individual prompting, structured infrastructure, supervised autonomy. This page's three steps run alongside that ladder rather than reproducing it. The Assistant step is the ladder's individual prompting; the Team step reaches its supervised autonomy, passing through structured infrastructure on the way, since the ledgers and specifications are that infrastructure; the Interconnected-ecosystem step lies beyond the ladder's top, because the ladder describes one team's practice and the ecosystem describes several teams and their boundaries. The mapping column in the table makes the alignment explicit.

Step What it is What exists What it cannot do Leverage-ladder level
Assistant A planner uses a model in a chat window to speed up the steps of a process the planner still runs Nothing durable; questions 1 to 4 may all fail Run unattended; be checked; survive the planner leaving L1, individual prompting
Team The process runs on a clock by specialists over ledgers, under gates; the planner signs Questions 1 to 4 pass for this process; the roster of The Agent Team Model runs it Reach beyond the ledgers; exchange with other systems except by hand L2 then L3: structured infrastructure, then supervised autonomy
Interconnected ecosystem Adapter agents exchange graded, signed data with the platform of record, the ACD, HR, finance and the planning platform; several teams share ledgers Questions 1 to 4 pass for every process on the loop; the mapping files of Integration Agents exist for every boundary Act without a gate; the condition on which a gate is released is stated in evidence terms on Human Gates and Number Grades and is earned by catch rates, not declared Beyond L3; the ladder describes one team's practice

The step from assistant to team is the one that compounds, for the reason AI Leverage Maturity in WFM Teams gives: architecture compounds while models depreciate, and the team step is the point at which the method becomes an artifact rather than a habit. It is also the step at which the four questions stop being optional. An assistant tolerates a missing definition because the planner supplies it from memory every time; a team does not, because nobody is there to.

How functions fool themselves

  • The definitions exist in a document nobody cites. Question 1 asks whether the process's columns cite records, not whether a glossary exists; a measurement without an operational definition has no meaning, however well the glossary is written.[2] The test is mechanical: reject a column without a record and see what breaks.
  • The L2 table was written for the audit, not the work. A table whose steps do not match what the planner does on Tuesday will produce an agent that does the audit version. The test is to run the table by hand for one cycle and log every deviation; the deviations are the real table.
  • The cycle is a calendar of meetings. A meeting is where a decision is taken; a cycle is what produces the pack the meeting reads. Question 3 asks for the second.
  • The register is an issue log. An issue log holds things being resolved; a question register holds things being answered, with a hypothesis table. A function with only the first will find its "why" questions answered in email again within a month.
  • Readiness is claimed for the function, not the process. The four questions are per process. A function with a definitions ledger for the monthly cycle and none for the intraday reforecast is ready for one team and not the other.

The task-based view of automation makes the general point: what can be automated is what has been codified, and codifiability is a property of how much effort has been spent writing the task down, not of the task's nature.[3] The four questions are a measure of that effort for one process.

Worked example

The series example's planning function runs the check on the short-term voice reforecast in February 2026, before the March go-live. Question 1: the demand feed's columns cite records for offered, handled, abandoned and handle time, but "contacts per transaction" has two records in circulation, CPT-01 and CPT-02; a narrow fail, cleared when the owner retires CPT-02 the following week. Question 2: L0 and L1 exist; the L2 table's tools column is blank on three steps, filled in an afternoon. Pass. Question 3: daily at 07:00, note by 09:00, platform load by 10:00, on a calendar. Pass. Question 4: an issue log and no question register. Fail; a register opens with Q-001 on 23 February. The process is declared ready on Monday 2 March, the day the migration goes live, which is why the daily clock on The Short-Term Forecasting Loop with an Agent Team was running on day 3. The monthly cycle passes all four; the intraday reforecast fails question 2 outright and stays with an assistant until April.

What would change this

The four questions are a distillation, and a function that handed a process to a team with three of them passing and ran it for a year without the fourth's failure mode appearing would have grounds to drop that question. The page expects that to happen least for question 1 and most for question 4, because a register can be opened after the first unexplained miss whereas a definition cannot be reconstructed after the first bad plan.

How this connects

This page is the per-process form of Standardize Before You Automate and the entry check before The Agentic Handover Gate. Its four questions name the four artifacts the series builds on: the definitions ledger of Living Ledgers, the L2 table of Process Decomposition (L0–L3), the clocks of The Agent Team Model and Capacity Planning Cycle, and the register of The Question Register and Knowledge Base. Its three steps are the leverage ladder of AI Leverage Maturity in WFM Teams applied to one process, and A Roadmap Pattern for Agent Teams in WFM sequences which processes are checked first.

The check is the entry condition for the first rung of The Agent Team Ladder: Alpha to Production and is re-run on every book a team is cloned onto, because a clone carries scaffolding and never the previous book's definitions; the program that runs the check as part of its first ninety days is AI Agent Program for a Resource Optimization Center.

Maturity Model Position

Passing the four questions for a process is Level 3 work on the WFM Labs Maturity Model™: definitions, documentation, a clock, a register. Running the process with a team is the Level 4 threshold. The check is deliberately usable at Level 2 as a diagnostic, where it will usually show that the function's most important process fails questions 1 and 4.

See Also

References

  1. ↑ Bainbridge, L. (1983). "Ironies of Automation". Automatica 19(6), 775–779. doi:10.1016/0005-1098(83)90046-8.
  2. ↑ Deming, W. E. (1986). Out of the Crisis. MIT Center for Advanced Engineering Study. ISBN 0-911379-01-0.
  3. ↑ Autor, D. H., Levy, F., & Murnane, R. J. (2003). "The Skill Content of Recent Technological Change: An Empirical Exploration". Quarterly Journal of Economics 118(4), 1279–1333. doi:10.1162/003355303322552801.
Retrieved from "https://wiki.wfmlabs.org/w/index.php?title=Agent_Team_Readiness_for_a_Planning_Process&oldid=4763"