The first week has one stage a day, in the order chapter 6 gave them. You bring two source systems, one decision your planners make every week and three people for about two hours a day.
Stage one of chapter 6, on the calendar: a connection, 23 tables and one project by the Monday afternoon.
This is stage one of the six you saw in chapter 6, done for real. Registration is a connection and a choice of tables. Juno records the table and column names, their types, their keys and a sample, and the rows are read in place when a question or an agent needs them.
The week asks you for two source systems, and the showcase estate brought its two ERPs. The other three joined the same project the same day, which is why the Wednesday’s answer could read three systems as one.
Juno proposes, your people confirm, and one published version carries the meanings into every answer and every agent.
This is the review you saw on the mapping screen in chapter 6, and it answers the question chapter 5 left open. The agent treated SAP and Oracle lines as one thing because on this Tuesday a person confirmed that both mean “purchase order line”.
Two hours and ten minutes covered 412 columns because the work is a review of proposals, each with its confidence and its reason. The 33 adjustments mark the columns whose meaning their names hide, and each one is kept as evidence for the next table.
The first answer arrives in seconds, and its working is what lets a planner trust it.
The question is the one the desk asks every Monday, and the one chapter 1 opened with. The value in the first answer is the working beside it: the terms, the systems, the mapping version and the query, so L. Tan checks the number and sees where it came from.
On Monday 21 September, after the first week, the same question found the 38 lines worth $412,000 that chapter 5 followed. On this Wednesday it found that week’s lines, 31 worth $348,000, and a planner read three systems as one for the first time.
Three taught terms and 50 scored questions turn accuracy into a number your team can watch.
| Taught term | What it means | Who added it | Where it applies |
|---|---|---|---|
| NCNR | Non-cancellable and non-returnable | D. Okafor, buyer | Purchase order lines and supplier terms |
| Lead time fence | The horizon inside which the plan holds still because supply is committed | L. Tan, supply planner | Planned orders and schedule lines |
| Strat supplier | An alias for the strategic supplier class | P. Sharma, master-data steward | Suppliers and supplier classes |
Showcase estate. P. Sharma owns the vocabulary file and reviews all three entries; each carries the person who added it and the date.
These are the three terms of chapter 6, added by the people who use them. Three words changed how a dozen questions were understood, because the words a desk uses rarely match the column names in an ERP. The scored set turns accuracy from an impression into three numbers, and it re-runs after every later change.
A sentence becomes a planning table, the table is simulated on last quarter, and a named manager sets the level.
This is chapter 5 in miniature: the same sentence, the same five rows, the same limits and the same 13:30 release, on the day it happened. The nine settings of chapter 4 are made by the people chapter 4 named: the planner writes the rule, the owner sets the limits, and the manager, J. Ridley, sets the level.
A2 is the level of chapter 3: the agent decides and drafts, and a named person approves each action. The simulation is what lets J. Ridley sign, because last quarter is replayed case by case beside the decisions the desk remembers making.
At 13:30 on Friday 18 September the showcase estate had one agent running: the PO cancellation agent at A2, on rule version 3, with D. Okafor approving each action. Version 1 was the Planner Agent’s draft, version 2 followed L. Tan’s edit, and version 3 is what J. Ridley released.
On Monday 21 September at 06:00 the same agent ran on the same version and put 29 drafts worth $187,300 in the queue: the Monday chapter 5 followed.
One decision that repeats every week is worth more in this week than a wide catalogue of tables.
Your main ERP and one more: a second ERP, a planning suite, a warehouse system or a lake.
A decision your planners or buyers already make every week, with a number attached to it.
A data lead, a domain owner who knows the fields, and the planner or buyer who decides.
Two more people join at set points. The data lead opens the connections on the Monday and IT sets up access and security behind them; on the Friday the desk manager sets the level and reads the answer to the test write with IT. In the showcase estate the data lead was A. Novak, and the two who joined were J. Ridley and S. Iyer.
The decision matters more than the data volume. A week spent on one decision that repeats every Monday gives you an agent, a scored question set and a number you can hold up. Chapter 8 lists twenty decisions the market is deploying now; one of them is the right Monday for the week.
Four versioned outputs, each owned by a named person on your side.
From five systems, read where they live, grouped into one project.
One published mapping version that ChatLima and every agent read.
Your own questions with the answers you expect, scored and kept for re-runs.
One policy in plain language, simulated on last quarter, running at the level you set.
All four are yours and all four are versioned. The mapping carries its confirmations and corrections, the question set its expected answers, and the rule its owner, its simulation and its level. Keeping the week to one decision is what makes each of them small enough to check and complete enough to run.
Every approval, edit and hold is recorded and used, and the level moves on that record.
D. Okafor reads drafts faster as the weeks pass, because every edit is recorded and used, the signal chapter 4 described. By day 30 the queue carried mostly lines the desk agreed with.
After six weeks the record is 212 cancellations drafted, 203 approved as written and 9 edited. That is the evidence chapter 3 asked for: the recorded cases pass, J. Ridley approves by name, the escalation route is live, and lines under $25,000 move to A3. Lines of $25,000 and above keep arriving as drafts, because the threshold stays where it is.
The week, the demo and the bake-off each have a scope you can read before you book.
Two of your systems, one weekly decision, five days.
Book the week →One hour on the showcase estate: the approval queue, the level review, the audit record and the manager report.
See the demo →Fifty of your own questions, scored for accuracy, speed and access correctness over two weeks on two of your systems.
Read the offer →Offer. The scope of each is agreed in writing before day one.
We bring the connectors, the ontology, the use-case design for your decision and someone who has run the week before. You keep everything built in it, and the second week starts on top of what the first one published.
A read connection to two systems with a service account, the names of the tables behind your weekly decision, and three people booked for two hours a day. We bring the rest.
The ERP that holds the decision and one system that holds the other half of the picture: a planning suite, a warehouse system, a supplier portal or a lake. The join across the two is where the answer lives.
The week runs on read access until the Friday. Write-back is switched on per application after a confirmed test write, at the moment you choose, and every write keeps the acknowledgement your ERP returns.
A second decision runs the same five stages on top of the tables and the mapping already published, so it takes less time. The first agent keeps running, and its decisions are recorded from the Friday afternoon.
Which of the twenty use cases is your Monday?
Pick the chain, read the use case, and bring it to the first week.
The conversation on its own: the question, the answer, the working, and the two-week bake-off on fifty of your questions.
Each one builds on the one before it.
Reads your rows in place.
Connect and move ↗One record per real thing.
Master and govern ↗Apps for every decision.
Operate and decide ↗Asks and acts, at your level.
Ask and act ↗You bring two source systems, one weekly decision and three people for two hours a day; you leave with registered data, a confirmed mapping, answers in ChatLima and one agent running at the level you chose.
