The resin order no longer starts in a senior person's inbox.

Flyte knew what to buy because two senior people carried the supplier, stock and production judgement. Kaizen captured that judgement and built a daily routine that prepares the orders for approval.

Original job: Raw-material ordering
Practitioners: Two senior people
Go-live: 9 July 2026
Evidence: Go-live record and client measurement

01 / Original job

The calculation happened before the purchase order.

The morning started with supplier prices, current stock, the production schedule and a spreadsheet. Someone senior worked out what the business needed, then entered the purchase order into the system.

A second senior person could cover the job, but only by carrying the same knowledge. The process worked because they knew the grades, suppliers, lot sizes and exceptions. None of that judgement belonged to the business as a written operating record.

02 / Captured judgement

The rules were the work.

Kaizen worked through a real ordering cycle with the people who held it. The record covered supplier choice, current pricing, stock, safety stock, production demand, minimum quantities and the reasons to stop the normal path.

The team did not automate every decision. It separated repeatable calculations from the exceptions that still need someone who understands the business.

03 / Running job

Every morning starts with a checked list.

At 7:30 each morning, the routine checks 335 materials and shows which need action. It prepares draft purchase orders in their own section so Flyte can review and approve them.

The first live-data test checked 333 materials. It identified three to order and nineteen to watch. The people who knew the job checked the recommendations before the routine went live.

Approved operating shape

The business owns the routine. People own the exceptions.

The record below explains the public operating boundary. It is not Flyte's confidential delivery document.

Input: Supplier pricing, stock and production demand
Written rules: Order timing, lot sizes, safety stock and exclusions
Runs now: Daily material check and draft purchase-order preparation
Human decision: Review exceptions and approve every purchase order
Acceptance evidence: Live-data validation followed by production go-live
Ownership: Flyte keeps the rules, running routine and approval path

Recorded evidence

What supports the result.

2 July 2026
Live-data test completed across 333 materials.
9 July 2026
The routine went live.
After go-live
Flyte estimated about twenty hours of senior time returned each week.

Systems in the job

The systems follow the work.

Salesforce holds the operating records. Anthropic Claude applies the written ordering logic through purpose-built MCP tools. The system choice supports the job. It does not replace the approval decision.

A similar job

What still needs a senior person every morning?

Describe the repeated work. You do not need to know which system or model should handle it.

Talk to Mara