How to build a command centre with AI: step by step
To build a command centre with Corey, tell it what you need to see to run your week, let it read the sources you already have - your files, your repo, your accounts - and have it render a single page from them. You keep deciding what belongs on it. Corey helps you build it and rebuilds it as the numbers move; it is a projection of your own files, not a second place to keep the truth.
Most dashboards fail the same way. Somebody builds the screen first and works out what it is for afterwards, so it fills up with numbers that are true and useless, and within a month nobody opens it. A command centre is worth having only if it changes what you do on a Monday morning.
Corey helps you build one. It does not host it for you, and the difference matters: the page lives in your folder, reads your files and your accounts, and belongs to you. Here is the order that works.
1. Say what a good Monday morning looks like
Start with the decision rather than the dashboard. Tell Corey, in plain language, what you need to know to run the week, and be honest about what you keep meaning to check and never do. Cash position. Whether anything is waiting on you. What shipped. What is overdue.
Corey turns that into a short list of what the page has to show. Then cut the list. This is the only step where being ruthless is free, and every item you leave on now is one you will scroll past for a year.
2. Point it at the sources you already have
Corey reads what exists rather than asking you to type things in: the files in your folder, your repo, the accounts you have connected.
This step has a useful side effect. When a number you asked for has no source behind it, Corey says so instead of filling the gap. That is the finding, not a failure - the thing you most want to watch is very often the thing nobody ever set up a way to measure. You then decide: wire it up, track it by hand for now, or admit you do not actually need it.
3. Have it render one page
One page. Not a suite, not a set of tabs. Corey builds it in your own folder and ranks it so the things needing a decision sit above the things that are merely interesting - a deadline above a chart, an approval above a trend line.
If it does not fit on one screen, it is not a command centre yet; it is a report, and you will read it the way you read reports.
4. Cut it down
Use it for a week, then take off everything you looked at and did not act on.
This is the step people skip, and it is why dashboards rot. A page of thirty tiles is not more informative than a page of six - it is less, because the six that matter are now hidden among the twenty-four that do not. Tell Corey what to remove and it reships.
5. Set it to rebuild
Ask Corey to refresh the page on a schedule, so it is current when you open it rather than current when you last remembered to update it.
Because the page is generated from your sources rather than typed into, it fails loudly. A source that stops reporting leaves a visible gap instead of an old number that still looks perfectly current, which is the failure mode that makes people stop trusting a dashboard without ever quite noticing why.
The one rule: it is a view, not a store
Nothing on the command centre should be the only copy of anything. It is a projection, rebuilt from your files and accounts, and you should be able to delete the whole thing and regenerate it without losing information.
The moment you start typing a number straight into the page, you have created a second source of truth, and two sources of the same number will eventually disagree. When they do, you will not know which one is wrong. Keep the truth in the files and let the page be what it is: the screen that tells you what to do next.