Applied The OFCI Playbook

The OFCI Command Center

Last reviewed

Advanced

What you'll learn

~60 min
  • Build an Express.js dashboard server integrating data from all procurement tools
  • Implement a Monday Morning Run that refreshes all data sources and generates a prioritized briefing
  • Create a portfolio overview with project cards, RAG status, and drill-down views
  • Build a Top 15 priority action queue that surfaces the most urgent items across all tools

What you’re building

It’s 7 AM on Monday. You open your laptop and need to answer one question: “What do I need to do this week?” Right now that means checking email, opening three spreadsheets, scanning Procore for submittals, looking at the delivery calendar, hoping you don’t miss something critical, and trying to remember that thing from Friday that you swore you’d remember on Monday. You never remember. Nobody remembers.

The OFCI Command Center kills that entire ritual. One URL. One dashboard. One button. You click “Monday Morning Run” and in 10 seconds you have:

  • Portfolio health across all projects (RAG scorecards)
  • Your Top 15 most urgent action items, ranked by an algorithm that’s better at prioritizing than your Monday-morning brain
  • Overdue submittals that need follow-up calls
  • Procurement windows closing this week
  • Deliveries scheduled in the next 14 days
  • Budget variances that need attention

This is the capstone. Every tool you’ve built in this module becomes a data source that feeds the Command Center. Seven tools, one dashboard, zero “oh shit I forgot about the SV9 switchgear.”

💬The Monday Morning Run

You know the feeling — it’s drop day and you’re hitting 203, 374, and 131 before anybody else gets the call. You’ve got your route, you know which stores to trust and which ones will lie to your face, and you’re not wasting time on stores that never have shit. Your Monday Morning Run is the procurement version. One click, full refresh, prioritized action queue. No driving to six stores. No “oh crap I forgot about the SV9 switchgear submittal.” Just clarity, first thing, every week. That’s the whole damn point of everything we’ve built.

By the end of this lesson you will have a web dashboard — a page you open in your browser, like any website — that aggregates data from all your procurement tools, provides a portfolio overview with project-level drill-downs, runs a “Monday Morning Run” to refresh all data and generate a prioritized action queue, and gives you a single operational cockpit for OFCI procurement management. It’s the thing you wish you’d had for the last five years.

ℹSoftware pattern: Unified operations dashboard with orchestration

Multiple tool outputs → data aggregation layer → prioritized action queue → web dashboard with drill-downs. This pattern works for any operations center: NOC dashboards, DevOps control panels, executive command centers, project management hubs. The key insight is that individual tools are useful, but a unified view with intelligent prioritization is transformative.

🔍What's different about this lesson (read this first)click to expand

Lessons 1-7 built CLI tools — you typed a command, got text output. This lesson builds a web dashboard — a local website running on your machine that you view in Chrome or Edge.

Here’s the plain-English version of the new tech involved:

  • Express.js is a tiny web server. It’s like a local version of a website — instead of going to google.com, you go to localhost:3000 (your own machine). When you type node server.js, you’re starting that local website.
  • EJS templates are HTML pages with blanks that get filled in with your data. Think of them like mail merge in Word — the template has placeholders, and the server fills them in with your actual project numbers before showing you the page.
  • Routes are just URLs. When you click a project card, the browser goes to /project/SV9, and the server knows to show you the SV9 detail page. It’s the same as clicking a link on any website.
  • The data directory is where the dashboard reads JSON files from. They hold the same data your other tools work with; sample files ship with the dashboard, and a --json export added to each earlier tool (a one-line follow-up prompt) feeds it real numbers. The dashboard doesn’t do any procurement logic itself — it just reads the results and displays them visually.

The AI CLI tool handles all the code. You describe what you need in plan mode, answer its questions, and get a working dashboard. The only difference from previous lessons is that instead of running node src/index.js and reading terminal output, you run node server.js and open a browser tab.


The showcase

Here’s what you’re looking at when you open localhost:3000:

  • Dashboard home: Portfolio overview with project cards showing RAG status, budget health, and key metrics. Think of it as the “how screwed am I?” view, but prettier.
  • Monday Morning Run: One-click refresh that reads all data sources and generates the weekly briefing + action queue. This is the button that replaces your entire Monday morning orientation routine.
  • Top 15 Queue: Prioritized list of the 15 most urgent action items across all tools and projects, ranked by impact and urgency. Not ranked by “whichever PM yelled at you last,” which is how most of us currently prioritize.
  • Project drill-down: Click a project card to see the detailed scorecard, submittals, lead times, budget, and deliveries. Everything you need to walk into a project meeting and not look like you’re winging it.
  • Responsive design: Works on laptop (primary) and tablet (site visits). Because sometimes you need to pull this up in a construction trailer and pretend you knew about the delivery conflict all along.
💡A big build: let the plan split it

The plan for a full web dashboard will be longer than a CLI script’s plan, so before you approve it, you can tell the agent:

Split this plan into chunks I can check one at a time. Build the first chunk,
run it, and stop so I can look before you start the next one.

That is the rule of one from Breaking Down Tasks (one feature per step, the next only after the last one works), applied to a multi-part build. When the build works, keep the facts later requests repeat in a one-screen brief the tool reads at the start of every session:

Write AGENTS.md, one screen long, holding the three projects (SV9, CH2, DEN3), the four
JSON files each project folder holds, the dark theme, how to start the
server (node server.js, port 3000), and what the Monday Morning Run does.

Codex, Antigravity and Copilot read AGENTS.md on their own. Claude Code reads it directly when the project has no CLAUDE.md (v2.1.277 and later); if it has one, add the line @AGENTS.md to it so it loads too. Before each commit, run the tool’s built-in review on what changed: /code-review in Claude Code, /review in Codex and Copilot, or /diff in Antigravity followed by a request to review that diff.


Before you start

If your agent is still open from the previous lesson, exit that session first (/exit or Ctrl+D). Create a clean folder for this lesson and open it:

Terminal window
mkdir -p ~/ofci-playbook/08-ofci-command-center && cd ~/ofci-playbook/08-ofci-command-center

In Windows PowerShell:

Terminal window
mkdir ~/ofci-playbook/08-ofci-command-center; cd ~/ofci-playbook/08-ofci-command-center

Or open the folder directly in your desktop agent app. Start your agent the way Workbench Setup showed.

Switch to plan mode

Switch to Plan mode inside your session (Shift+Tab, /plan, or the mode picker) so the agent outlines the Express routes, prioritizer math, and views before writing files.

The conversation

Say what you want

Ask for the unified OFCI web dashboard in plain English:

I run OFCI procurement across three projects. Help me build a local web dashboard called ofci-command-center with project RAG cards, a Top 15 action queue, and a Monday Morning Run refresh. Ask me what you need first.

Answer its questions

Guide the agent through the routes, scoring formula, and layout:

If it asksYou might say
What routes and views should the server expose?Serve / for the dashboard, /project/:code for project drill-downs, /briefing for the Monday run, and /api/refresh to reload.
How should the Top 15 action queue score and rank items?Multiply (base score + days overdue or past order-by) by equipment criticality: 1.5x generators/switchgear, 1.2x chillers/UPS, 1.0x CRAHs/PDUs.
What base scores should each source category use?Lead-time black 100, submittal Tier 3 90, lead-time red 80, delivery conflict 70, Tier 2 60, unconfirmed delivery 50.
What does clicking Monday Morning Run do?Re-reads all project JSON files from data/, recalculates scores, saves a snapshot in data/snapshots/, and redirects to /briefing.
What styling and sample data should it include?Create sample JSON files under data/sv9, data/ch2, and data/den3, and use a dark theme with amber accent #D97706.
ℹReference notes to hand the agent (or check against)click to expand
  • Data folders: data/sv9/, data/ch2/, and data/den3/, each holding budget.json, submittals.json, leadtimes.json, and deliveries.json.
  • Top 15 Prioritizer formula:
    • Base scores: Lead time black = 100, Submittal Tier 3 overdue = 90, Lead time red = 80, Delivery conflict = 70, Submittal Tier 2 overdue = 60, Delivery unconfirmed (<14 days) = 50, Budget variance (>10%) = 40, Budget variance (5-10%) = 20.
    • Days modifier: + days_overdue for submittals; + abs(days_past_order_by) for lead times (use absolute value so past-due items score higher, not lower).
    • Criticality multiplier: generator/switchgear = 1.5x, chiller/UPS = 1.2x, CRAH/PDU = 1.0x.
    • Final score: final_score = (base_score + days_modifier) * criticality_multiplier, sorted descending, top 15 across all projects.
  • Dashboard views:
    • / (dashboard.ejs): Portfolio header with last refresh timestamp and Monday Morning Run button, project RAG cards, Top 15 Action Queue with slide-out detail panel, and Upcoming Week timeline.
    • /project/:code (project.ejs): Full 4-dimension RAG scorecard, budget variance table, submittal breakdown, lead-time watchboard, and upcoming deliveries for one project.
    • /briefing (briefing.ejs): Monday Morning Run summary and key changes since the last snapshot in data/snapshots/YYYY-MM-DD.json.
  • Theme and packages: Dark theme (#0a0a0f background, #1a1a2e cards, #e0e0e0 text, #D97706 amber accent, RAG colors #ef4444 / #f59e0b / #22c55e). Use express, ejs, and chalk@4.

Read the plan

Before approving the agent’s plan, verify four things:

  1. All three projects aggregated: aggregator.js loads JSON files from data/sv9/, data/ch2/, and data/den3/ into one combined portfolio model.
  2. Exact Top 15 scoring formula: prioritizer.js computes (base_score + days_modifier) * criticality_multiplier using Math.abs on days past the order-by date so overdue POs rank highest.
  3. Live disk reload on Monday Morning Run: Triggering /api/refresh clears any in-memory cache, re-reads all JSON files from disk, writes data/snapshots/YYYY-MM-DD.json, and renders /briefing.
  4. Working drill-down routes: Each project card links to /project/:code, rendering all four dimensions (schedule, budget, submittals, deliveries) for that site.

If the plan subtracts negative days_remaining instead of adding Math.abs(days_past_order_by), correct it before approving:

Use the absolute value of days past the order-by date so past-due lead-time items score higher in the Top 15 queue.

Approve the plan once all four checks pass, and ask the agent to save a Git checkpoint when the server starts cleanly.

💡The capstone experience

This lesson ties everything together. The data files in the data/ directory simulate what you’d get from running the tools you built in Lessons 1-7. In a real deployment, you’d point the aggregator at the actual output directories of those tools. For now, the sample JSON files give you a working dashboard to demonstrate the concept.


What you get

Fire it up

Ask your agent to install what is needed, start the dashboard server, and open it in your browser:

Install what the dashboard needs, start the local server, and open http://localhost:3000 in my browser.

If you prefer running the commands in a terminal yourself:

Terminal window
cd ofci-command-center
npm install
node server.js

Open http://localhost:3000 in your browser. You should see the OFCI Command Center with project cards, the Top 15 action queue, and the Monday Morning Run button. If this is the first time you’ve built something with a browser interface through a plan-mode conversation — yeah, that reaction is normal. It’s a whole dashboard, built step by step.

If something is off

ProblemFollow-up prompt
Server crashes on startupThe dashboard server fails to start. Make sure all dependencies are installed, the template views folder and static public folder are wired up properly, and all view templates exist before starting the server.
Templates show raw variablesThe dashboard pages are displaying raw template placeholders instead of project values. Make sure each page receives the loaded projects, action queue, and refresh timestamp when rendering.
Project cards don’t link to detail viewClicking a project card doesn't open that project's detail page. Make sure each project card links to /project/SV9 (using the project's code) and loads all four dimensions for that project.
Monday Morning Run doesn’t refresh dataClicking Monday Morning Run doesn't update the numbers. Make sure /api/refresh reloads all project JSON files from disk instead of a cached variable, recalculates the Top 15 scores, saves a snapshot, and redirects to /briefing.

🔧

When Things Go Wrong

Use the Symptom → Evidence → Request pattern: describe what you see, paste the error, then ask for a fix.

Symptom
The Top 15 queue shows the same project's items exclusively
Evidence
All 15 action items are from SV9, even though CH2 and DEN3 also have urgent items
What to ask the AI
"The prioritizer is likely loading data from only one project's directory. Make sure aggregator.js reads data from ALL project subdirectories in the data/ folder, not just the first one found. The prioritizer should score items from all projects in a single array, then sort and take the top 15. If one project genuinely has all 15 most urgent items, add a diversity bonus to ensure at least the top item from each project appears."
Symptom
The dashboard looks broken on mobile / tablet
Evidence
Project cards overlap, text is cut off, and the action queue is unreadable on iPad
What to ask the AI
"Project cards overlap and text gets cut off on tablet and mobile screens. Make the layout responsive: show three project columns on desktop, two on tablet, and a single full-width column on mobile so cards and action queue items never overflow."
Symptom
The Monday Morning Run snapshot saves but doesn't show changes
Evidence
I click Monday Morning Run, it says 'Refresh complete', but the briefing view shows the same data as before
What to ask the AI
"The refresh route needs to clear the in-memory cache before re-reading files. If you're caching the aggregated data in a module-level variable, set it to null before re-reading. Also make sure the briefing view reads from the freshly-saved snapshot, not from the stale cache. Add a 'last refreshed' timestamp that updates visibly in the UI."

How it works (the 2-minute explanation)

Four moving pieces, and none of them are complicated on their own. The power is in the wiring.

  1. Express server serves an EJS-templated web dashboard on port 3000. That’s it — it’s a local website. Routes handle the main dashboard, project detail views, and the Monday Morning Run briefing. When you click things, the server shows you different pages with different data. If you’ve ever used a website, you already understand this part.
  2. Aggregator reads JSON data files from each project’s subdirectory and merges them into a unified data model. Think of it as the thing that takes output from seven different tools and makes one big picture out of it. Without this, you’re the aggregator — and you’re doing it in your head every Monday morning while also trying to remember if you fed the dog.
  3. Prioritizer is the part that makes this thing actually useful. It scores every actionable item across all tools and projects using a weighted algorithm that considers urgency, impact, and equipment criticality. Then it ranks them and gives you the top 15. Your brain is great at procurement decisions. It’s terrible at remembering which of 100 items across three projects needs attention first. Let the algorithm handle the sorting so your brain can handle the decisions.
  4. Monday Morning Run re-reads all data sources, recalculates everything, saves a snapshot for historical comparison, and generates the weekly briefing. One click. Full refresh. Prioritized action queue. That’s the whole thing.
🔍From seven tools to one platformclick to expand

Look at what you’ve built across this module:

  1. RFP Machine → generates procurement documents
  2. Vendor Scorecard → evaluates bids with weighted scoring
  3. Submittal Tracker → manages submittal lifecycles
  4. Lead Time Watchboard → monitors procurement windows
  5. Budget Hound → tracks financial health
  6. Delivery War Room → coordinates logistics
  7. Executive Briefing → synthesizes for leadership
  8. Command Center → unifies everything into one view

Each tool solves one problem well. The Command Center connects them into an integrated procurement operating system. This is the difference between having seven useful scripts and having a platform. And you built it all from prompts.


Customize it

You’ve got a working command center. Now make it yours. These are the extensions that turn “cool demo” into “this is how I actually run my portfolio.”

Add a Slack or Teams notification

Send the Monday briefing to your team chat using a Slack incoming-webhook payload or an Adaptive Card for a Teams Workflows webhook (Teams no longer accepts the old Office 365 connector webhooks):

Add a notify option to Monday Morning Run (?notify=true on /api/refresh) that posts the portfolio RAG summary, Top 5 urgent action items, and dashboard link to a configured Slack or Teams Workflows webhook URL.

Add historical trend charts

Add a /trends page that loads all snapshots from data/snapshots/ and plots visual trend lines for budget variance, submittal completion rate, and lead-time window health.

Add a custom action queue

Add pin, snooze, and dismiss actions to the Top 15 queue, persisted in user-prefs.json: pinned items stay on top, snoozed items hide for N days, and dismissed items clear until the next Monday run.

Try it yourself

  1. Open your agent in a clean folder and switch to plan mode.
  2. Describe the command center, answer the agent’s questions, and approve the plan.
  3. Ask your agent to install what is needed and start the server (or run npm install && node server.js in your terminal), then open the dashboard in your browser.
  4. Click the Monday Morning Run button and review the briefing. This is the moment it all clicks.
  5. Click each project card and explore the detail views. Everything you used to hold in your head is now on a screen.
  6. Scan the Top 15 queue — imagine this is your actual Monday morning. What would you do first? The algorithm already answered that for you, but it’s worth checking whether your gut agrees.
  7. Pick one customization and add it. The Slack notification is the one most people want first, because your PM is going to lose their mind when they see this.

This is the tool you open at 7 AM every Monday. If it saves you 30 minutes of “what do I need to focus on?” orientation time, that’s 26 hours per year. If it prevents one missed delivery or one forgotten submittal follow-up, it’s paid for itself ten times over. And you built it with your agent on a subscription you already pay for.


Key takeaways

  • Unified views beat scattered tools, and it’s not close — seven individual CLI tools are powerful, but a single dashboard that aggregates them all shows you connections and conflicts that no individual tool reveals. That delivery conflict on DEN3 that intersects with the budget variance on SV9? You’d never catch that in separate spreadsheets.
  • The Monday Morning Run is the killer feature — one click, full refresh, prioritized action queue. It turns “let me check everything and hope I don’t miss something” into “here’s what matters, ranked, with context.” Your Monday mornings just got 30 minutes shorter and significantly less stressful.
  • Prioritization by algorithm beats prioritization by memory — your brain is good at procurement decisions. It’s not good at remembering which of 100 items across three projects needs attention first. That’s not a character flaw, that’s just how brains work. The scoring algorithm handles the sorting so you can focus on the actual work.
  • You built all of this from prompts — eight lessons, eight tools, one integrated platform. The IT department would’ve needed a budget request, three meetings, and a vendor evaluation just to START talking about this. You built it in a weekend.

KNOWLEDGE CHECK

You open the Command Center on Monday morning. The Top 15 queue shows item #1 as '🔴 SV9 — Issue PO for switchgear (order-by date was 14 days ago)' and item #2 as '🔴 CH2 — Follow up on Eaton submittal (22 days overdue, Tier 3).' Both are critical. How does the prioritizer determine that the switchgear PO is #1?


Module complete

You just built eight tools and a command center from prompts. Eight. Every one runs locally on your machine. Every one solves a real problem you face every week. Here’s the full lineup:

LessonToolWhat It Does
1RFP MachineGenerates complete RFP documents from rough equipment specs
2Vendor ScorecardEvaluates vendor bids with weighted multi-criteria scoring
3Submittal TrackerTracks submittal lifecycles across projects with overdue alerts
4Lead Time WatchboardMonitors procurement windows and calculates order-by dates
5Budget HoundAnalyzes budget variances with waterfall charts and narratives
6Delivery War RoomManages delivery logistics with conflict detection
7Executive BriefingGenerates portfolio scorecards and risk heat maps for VP review
8OFCI Command CenterUnified dashboard with Monday Morning Run and Top 15 queue

The gap between “construction procurement manager” and “construction procurement manager who builds custom AI-powered tools” is eight prompts wide. You just crossed it. The IT department didn’t need to be involved. No software vendor gave you a demo. No budget request sat in someone’s inbox for six weeks. You just… built the thing you needed.

Now go open something good. Might I suggest the Provenance 14 — absolute overload of dark fruit, like Gushers fruit snacks overload. You’ve earned it.

intellimetrics · Arlington, VA
© 2026 intellimetrics.net
Search lessons