Ecoscape Solutions · Orange NSW

Command Centre plan

For Dan Curran. Prepared by Ben Sims, Lawn Tips, 12 November 2026. Built on your blueprint (24.9), the technical handover brief, the engine framework and the SharePoint notes.

Your design holds up. This adds the build order and what we learned running ours.

What you've already settled

Your documents answer most of the big questions, so this plan doesn't redo them:

AroFlo owns jobs, quotes, costing, timesheets, claims and invoices. Xero owns the ledger. Employment Hero owns staff and tickets.
Microsoft 365 is the operations layer: the Master List is the command board, Teams is where staff work, Power BI does the forecasting.
One 7am digest per PM, a nightly check that flags only exceptions, and a 6-point handover gate before a job goes active.
SharePoint goes to the 4-site layout from the HPIT guide, with one flat folder per job and a separate archive library for the 50,000 old files.
Claude does the reading and drafting work Power Automate can't: tenders, specs, dockets and email threads.

That last point changes what we build for you. Your staff will live in Teams, so the daily screen should be in Teams too. The web dashboard we branded for you in October is still there if you want an office TV or a directors' view, but it's optional.

Your five questions

From "What we need to review" in your blueprint. Straight answers, including where ours is different to yours.

1. Getting AroFlo jobs in without parsing emails

How ours does it: it never reads notification emails. Every feed is a scheduled pull straight from that system's API.

Do the same with AroFlo. Every hour, ask AroFlo's API for jobs changed since the last run and write them into the Master List, matched on AroFlo Job ID so a re-run updates the row instead of adding a duplicate. A scheduled pull is sturdier than a webhook at your volume: a missed webhook is gone for good, a missed pull is caught the next hour.

CheckThe HTTP action in Power Automate is a premium connector. Either the flow owner gets a premium licence, or the hourly pull runs as a Claude routine instead. AroFlo may also need to switch API access on for you.

2. How staff see their daily tasks

How ours does it: a private web page that works on a phone, laptop or the office TV. The top shows only what needs doing today, then the numbers, then a shared to-do list and a box to ask questions.

For you, keep it where you planned: the "My Active Actions" List view and the 7am card in Teams. Your PMs are already in Teams, and a second place to check is a second place to ignore.

3. A morning digest that isn't spam

How ours does it: one scheduled run per job, each posting one card. The morning report is a single message.

What we learned the hard way:

One message per person per day. Everything else waits for it.
Exceptions only. A healthy job says nothing.
Every number says how old it is. If the AroFlo pull failed overnight, the card says so at the top instead of quietly showing yesterday's figures as today's.
Start with three rules, not ten. Add a rule only after running it by hand and seeing it catch something real.

4. Microsoft Lists or Dataverse

How ours does it: neither. It runs on a small Cloudflare database.

For 50 to 100 jobs a year, Lists is plenty and costs nothing extra. Use two lists: Jobs (one row per AroFlo job) and Claim Cycles (one row per monthly claim), linked by a lookup on AroFlo Job ID. A one-week job has one cycle, a six-month contract has six, and both look the same. Index Job Number, Stage, PM Responsible and Forecast Invoice Date.

5. The Power BI schema and formulas

How ours does it: we don't use Power BI, so there are no formulas to hand over.

Claude can write them for you from the column list in your handover brief, which is already enough: the 30/60/90 day forecast from Forecast Invoice Date and Forecast Amount, billed against forecast, and days in each stage. Build it once the List has a month of real rows, so you're checking the formulas against real figures.

Where Claude fits

As the builder

You describe a flow in plain words and Claude writes the steps, the formulas, the Adaptive Card JSON and the Power BI measures, then tells you how to test each one. This is how most of ours was built. Nobody here writes code by hand.

As the reader

Scheduled jobs on your own Claude account for work that needs reading and judgement: pulling scope and pricing out of tender PDFs, checking tender specs against your TDS sheets, matching delivery dockets to AroFlo POs, and summarising long RFI threads.

The rule we run by: Power Automate does anything that must happen the same way every time. Claude does the reading and drafting. A person approves anything that goes to a client or a supplier.

Build order

Each step runs clean for two weeks before the next one starts. Adding the next thing before the last one is trusted means nobody trusts any of it. That's the biggest lesson from ours.

SharePoint restructure

Underway with HPIT. Static assets (fleet, product specs, mix designs, SOPs) move out of Project Delivery. The old 50,000 files go to a cloud-only archive library. Active jobs get one flat folder each.

Done when: the active library is under 10,000 files, with Job Number, Client, Category and Status indexed and filled in automatically from the folder.

Master List and AroFlo intake

The Jobs and Claim Cycles lists, fed by the hourly AroFlo pull. Turn off the email-parsing flows once the pull is proven.

Done when: every won job shows up within the hour with no duplicates, for two weeks with no hand fixes.

7am digest and nightly exceptions

Three rules to start: no client contact for 72 hours on an active job, SWMS not approved at mobilisation, and a claim due within 14 days that isn't submitted.

Done when: PMs work from the card, and every flag it raises turns out to be real.

6-point handover gate

The Adaptive Card you've already designed, blocking Pending Verification from moving to Active Delivery until all six boxes are ticked.

Done when: no job goes active with a box unticked, and nobody has to override it.

Power BI cash forecast

30/60/90 day forecast, billed against forecast, and stage velocity, refreshed twice a day.

Done when: last month's forecast lands within an agreed margin of what was actually billed.

Claude reading jobs

Start with one: whichever of tenders, dockets or RFIs costs the most office hours now. Add the second only once the first is saving real time.

Done when: a person checks its output and changes little.

What it costs to run

Item	What you'd pay
Lists, Teams, SharePoint, standard Power Automate	Already in your Microsoft 365 plan.
Power Automate premium	Only if the AroFlo pull runs in Power Automate. One licence for the person who owns that flow. Not needed if Claude does the pull.
Power BI	Check which licence your plan includes. Sharing dashboards usually needs Pro for each person viewing them.
Claude	A paid plan for whoever builds and runs the Claude jobs.
Dataverse	Not needed.

Questions for you

Does your Microsoft 365 plan include Power Automate premium, or should the AroFlo pull run from Claude?
Has AroFlo switched API access on for your account?
Who owns the build day to day: you, or HPIT?
For step 5, which hurts most right now: tenders, dockets or RFIs?