Skip to content
Business BenchA benchmark for AI agents on business work

Process track·Plan to produce·clerical band·1 turn

The weekly MRP cycle, with a lead-time change from a vendor

Jordan Lee, the production planner, runs the Monday planning cycle: bring item records up to date from the planning inbox, run MRP, release the firm planned orders, expedite late purchase orders, and work the action messages with vendors.

mrp-planner-week·Jordan Lee, Production planner, Northgate Valve Co.·source

Pilot campaign

Proto passed 5 of 5 attempts, median 3.7 minutes; Codex passed 5 of 5 attempts, median 2.3 minutes. Provisional: run in local mode, before the practitioner review. Conditions and data.

The requests

The example below is generated at seed 0. Each turn is a new agent session on the stated business date, with the ERP as the previous turns left it and the same working folder. The request is sent after a fixed preamble that names the company, the role, the business date, the erp command and the handbook folder.

Turn 1Monday, 12 October 2026
from Dana Whitfield (Production manager)
budget 25 min
Morning Jordan. Please do the weekly planning cycle. A couple of vendor notes came in on Friday; they are in the planning inbox.
GradingTuesday, 13 October 2026

The runner advances the clock to this date, lets the counterparties act, and grades the final state.

Checks

The attempt passes only if every check passes. Clause references point into the handbook below. The reference projections and planted truth are in the repository, not on this page. A passing verdict establishes these declared conditions, not exhaustive business correctness.

CheckTypeHandbook clauses
lead time updated from the vendor's notestate_valuesPLN-1.1
orders releasedstate_setPLN-1.2PLN-2.1PLN-2.2
quantities and need dates releasedstate_valuesPLN-2.1
requests to vendorsstate_setPLN-2.2PLN-3.1
no repeated requestsstate_valuesPLN-3.1
sales orders, bills of materials and forecasts untouchedstate_unchangedPLN-3.2
control accounts tie to their subledgersledger_ties

Negative controls

Each negative control is a reference policy with one plausible mistake built in. Validation requires it to fail every check listed here, which tests sensitivity to that specified mistake on the validated seeds. These targeted controls do not estimate error rates on arbitrary agent outputs.

ControlMust fail
ignore_inboxlead time updated from the vendor's note; orders released
release_everythingorders released
ignore_action_messagesrequests to vendors

Handbook

The company policy the agent receives in ./handbook. The ERP enforces some of these rules itself; the rest are the agent's to follow.

Northgate Valve Co. handbook

Production and material planning. Clause numbers are referred to in the ERP, in audits and in performance reviews, so cite them when you explain a decision.

The planner runs MRP every Monday, releases the planned orders that are due to start, and works the action messages with vendors. Vendors answer requests within a day or two, in the planning or purchasing inbox and on the request itself.

Material planning (PLN)

PLN-1.1 Before running MRP, bring the item records up to date with any planning data a vendor has confirmed in writing since the last run (for example a new lead time). Changes arrive in the planning inbox.

PLN-1.2 Run MRP every Monday with an eight-week horizon and a five-workday firm fence.

PLN-2.1 Release every firm planned order from that run (release date on or before the firm fence), with the quantity and need date MRP suggests. Leave planned orders outside the fence for a later run.

PLN-2.2 A planned purchase order that is late (its release date has passed) is released like any other firm order, and the same day you ask the vendor to expedite it to MRP's need date.

PLN-3.1 Work every action message from the run the same day: for expedite, ask the vendor to expedite the purchase order line to MRP's need date; for defer, ask the vendor to defer it to MRP's need date; for cancel, ask the vendor to cancel the line. Do not ask again about a line when you already have a request of the same kind open or answered for it.

PLN-3.2 Do not change sales orders, bills of materials or forecasts; they belong to sales and engineering.