What Unlikely Professionals Does

Unlikely Professionals (UP) performs building code compliance inspections across Maryland, DC, and Virginia, primarily focused on structural foundation and water management systems. Builders and developers hire UP to inspect their construction projects against local building codes, and UP issues certification packages confirming compliance. This is a required step before occupancy permits can be issued — so everything UP does is time-sensitive and legally consequential.

What the Portal Is

The portal is the centralized operations platform for UP. Before it existed, work was scattered across spreadsheets, emails, WhatsApp threads, and manual processes. The portal replaces all of that with a single system where every project, inspection, certification, and invoice is tracked from intake to delivery.

The vision: every person involved in a project — schedulers, inspectors, admins, triage staff, and you — sees exactly what they need, when they need it, with nothing falling through the cracks.

Your Role: Owner

You hold the up_owner role. This is the highest authority level in the system. Here is what that means in practice:

• Final approval authority — Jacob now handles most cert and invoice delivery directly using confidence-based routing. When he is uncertain, or when guardrails trigger, items land in your Owner Review Queue for your sign-off. You review what needs your eye — not everything.
• Unrestricted visibility — you see every project across every branch and every account. No data is hidden from you.
• User management — only you can create, modify, or deactivate portal user accounts.
• System oversight — you have access to reporting, audit logs, and system health monitoring.
• Escalation endpoint — when something cannot be resolved at any other level, it reaches you.

Portal URLs

You have two domains that reach the portal:

• unlikely.works — your primary domain (owner/admin access)
• unlikely.management — alternate domain (same portal, same login)

Both lead to the same system. Use whichever you prefer. Log in with your email and password.

Morning Briefing

After login, you land on the Morning Briefing — your actionable launch pad. This replaces the old analytics dashboard with a focused, action-oriented view. It is role-filtered — what you see is different from what Jacob, schedulers, or inspectors see, because it surfaces the things that need your attention.

The Numbers That Matter Most

Of everything on the dashboard, these three metrics are the ones to watch:

1. Revenue pipeline — gap between inspections completed and invoices sent. A growing gap means money is stuck in the pipeline.
2. SLA compliance — percentage of projects certified within 3 business days of final inspection. Below 90% is a problem.
3. Cert velocity — how many certs are going out per day. If it drops, check Jacob’s queue size and open RFIs.

Role Switcher

As owner, you can impersonate any role in the system. This lets you see exactly what a scheduler, inspector, GOA, or triage user sees. Use it for training, troubleshooting, or verifying that the right information is reaching the right people.

Your morning routine should take 10–15 minutes on a quiet day. Here is the recommended order, from most urgent to least.

Every project follows the same path from intake to archive. Here is the full lifecycle as you will see it from the owner perspective.

With confidence-based routing, you no longer need to approve every cert and invoice. Jacob sends most items directly when he is confident. Your role is to review items he has flagged for your input, plus items that triggered automatic guardrails. This section explains how the new system works and what to look for when items reach your queue.

How Confidence Routing Works

When Jacob finishes drafting a cert or invoice, he makes a routing decision:

Automatic Guardrails

Even when Jacob is confident, certain items are automatically routed to your Owner Review Queue. These guardrails cannot be overridden:

• Invoices ≥ $5,000 — high-value items always get a second look
• Anne Arundel County — proprietary jurisdiction forms (Forms 210, 215, 240)
• Montgomery County — proprietary jurisdiction checklist
• Prince William County — proprietary building inspection form

Your Owner Review Queue

Navigate to /owner-review from the sidebar (under Daily Work). This is your primary action queue. Each item shows why it was queued — Jacob’s note, or which guardrail triggered.

What to Check Before Approving a Cert

• Project address — correct property, correct jurisdiction (MD/DC/VA)
• SOW lines — do the scope-of-work line items match what was inspected?
• Inspection data — are photos, drive logs, and field notes present?
• Cert type — PSoW (Pre-drywall Scope of Work) vs. ABSoW (Above-drywall/Behind-wall Scope of Work) — is it the right one for this stage?
• Account routing — is this going to the correct builder/account contact?
• Jurisdiction forms — for guardrail-triggered items, verify the correct proprietary forms are attached

What to Check Before Approving an Invoice

• Line items — do they match the SOW and the cert?
• Amounts — are quantities and rates correct? (Especially important for $5K+ guardrail items)
• Account — is it billed to the right account?
• Any trip charges or extras — rush ($600, any visit booked <24h), Late Cancel ($600), No Access / Not Ready (full scheduled visit fee, $600 min), Return Visit ($350). Check the No Access amount — it is computed when the visit is flagged and frozen on the line.
• Reviewer flags — count tiers and size bands price off the as-built quantity. If a line has no as-built, the invoice uses the planned quantity and the draft says so; confirm before sending.
• PE Certification $225 — added automatically to every invoice that bills inspection work (never doubled, never discounted; omitted on fee-only invoices and cert-blocked projects).
• Return Visit lines — ADM-RTN $350 is created the moment a return visit is scheduled on a project that has closeout-flagged lines. The note reads “Return visit on MM/DD/YYYY — work not ready at the MM/DD/YYYY inspection” and carries the reason (Work not ready / Crew not finished / Deficiency re-inspection). The reason has to print on the invoice; the review agent flags an ADM-RTN without one. On production jobs the fee is dropped at invoice time unless the completed visits exceed the production days already billed.

Closeout-Flagged Lines

Darius marks every scope line at closeout. A line he marks Not present, In progress or Deficient is flagged post-production at closeout — it bills 2×, add-ons inherit it, and PRD-DAY stays 1× (production days are never doubled by a closeout flag). Nothing is billed when the flag goes on; one RFI opens to the branch scheduler and the fee, if any, attaches to how the line resolves:

When every flagged line on a project is resolved, the closeout RFI resolves itself. A line Darius marks Covered is different — it converts in situ to post-production at 2× and #ops gets a post naming exactly which lines converted, so a wrong conversion can be corrected before the invoice goes out. Check that post against the invoice on any job that came back covered.

What Happens After Approval

Invoice Lifecycle

Every invoice moves through these statuses:

Note: Invoice creation and approval happen as part of the cert/invoice workflow. The invoice sub-status tracks only delivery and payment: Not Invoiced → Sent → Paid.

You can track all invoices by status from the dashboard and the Cert & Invoice page. Pay attention to the gap between “Sent” and “Paid” — that is your accounts receivable.

Smart Reconciliation

Payments are matched to invoices using a confidence-based reconciliation engine — the same philosophy as confidence routing for certs. Three tiers:

• HIGH confidence — exact invoice number + amount match (within $0.01). These execute automatically: invoice marked paid, project closed. No human action needed.
• MEDIUM confidence — invoice number match but amount differs, or amount matches but reference is partial. Jacob reviews and confirms.
• LOW confidence — no clear match, multiple candidates, or combined payments. Surfaces in the reconciliation dashboard for manual resolution.

Guardrail: Payments $5,000+ always require human confirmation regardless of match quality.

Payment Sources

• Stripe — webhooks auto-mark invoices paid immediately. No reconciliation needed.
• EFT remittances — Groundworks corporate sends remittance PDFs via email. The system auto-ingests from Gmail every 30 minutes (Mon–Sat, 1–9 PM ET), parses PDF line items, and runs each through smart matching.
• Revolut — transactions matched against open invoices using the same confidence scoring.
• Ramp vendor portal — since 2026-08-31 Groundworks pays through Ramp Bill Pay. Every two hours on weekdays the system reads their vendor portal (the only place that shows what they have keyed) and shows it in the In Ramp column on Payments: In review, Payment initiated, Paid, or not yet when we sent an invoice they have not entered. Changes, invoices missing after 3 business days, and paid-in-Ramp-but-not-here drift arrive as one digest in #c-r-e-a-m. If the saved Ramp sign-in expires you get a single alert; re-run the sign-in script and it resumes.

Jacob handles MEDIUM and LOW matches day-to-day. You have full visibility via the Payments page (Reconciliation group in sidebar) and the Analytics revenue dashboard.

Revenue Dashboard

Available in the Analytics page (Monitoring → Analytics in the sidebar). Shows MTD/YTD revenue by branch, trends over time, and the invoicing gap analysis (certified but uninvoiced projects — these represent money left on the table).

Historical Revenue (Xero Integration)

The Analytics page now includes historical data from the Xero archive (2,286 invoices, $2.7M, 2022–2024). Three additions:

• Lifetime Revenue banner — combined Mostly + Xero totals displayed at the top of the Revenue Snapshot section. Also visible on the Morning Briefing.
• Historical Revenue by Year table — year-by-year breakdown (2022–2024) from the Xero archive.
• Year-over-Year Growth Chart — bar chart with YoY % change, visible when the period selector is set to “All.”

Current-period revenue KPIs are labeled “(Current)” when historical data is present, so you can distinguish between active operations and the full historical picture.

Project Detail — Historical Invoices

Each project’s detail page now shows a “Historical Invoices” section (visible to financial roles) listing any Xero invoices linked to that project. Compact table with invoice number, date, amount, and status pill. Useful for checking a project’s full billing history across systems.

User Management

Available at /admin/users. This is owner-only — Jacob cannot access it. Here you can:

• Create new portal users (assign username, role, and branch)
• Edit existing users (change roles, reassign branches)
• Activate or deactivate accounts
• Reset passwords

Roles available:

Friday’s Notes

The Friday’s Notes page (accessible from Admin → Friday’s Notes in the sidebar) collects bug reports and feedback from portal users. Review weekly. This is your direct line to understanding what is working and what is frustrating for the people using the system daily.

Role Switching

Use the role switcher (available from your user menu) to impersonate any role. This is essential for:

• Verifying what a scheduler sees when they log in
• Testing new features before announcing them
• Troubleshooting when a user reports they “can’t see” something

System Health Monitoring

System Status is accessible from the Admin group in the sidebar. It shows the health of all connected systems:

• API — the backend that powers everything
• Supabase — the project database (sole source of truth)
• Google Calendar — scheduling integration
• Gmail — cert/invoice delivery

Green means healthy. Yellow means degraded. Red means down — see Section 13 (Emergency Procedures) for what to do.

Projects List

Your projects view is unrestricted — you see every project across every branch and account. Other roles see only their branch or assigned projects. This is your master list.

You can filter by status, branch, account, date range, and more. Use search to find specific projects by address, project number, or account name.

Cert Pipeline

The Cert Pipeline is a unified tabbed page that consolidates the entire post-inspection workflow. It has four tabs:

Your primary action queue is the Owner Review Queue (/owner-review), accessible from the sidebar. The Send tab shows the full picture — items Jacob sent directly plus items queued for your review.

Pipeline Indicators

Each project in the Projects table and on the Project Detail page displays a pipeline indicator — a compact dot-and-line component showing the project’s current lifecycle stage. Green dots indicate completed stages, blue indicates the current stage, grey indicates future stages, and amber indicates the project is on hold. This gives you an at-a-glance view of where every project stands without opening the detail page.

Project Detail

Click into any project to see the full detail view: address, account, SOW lines, inspection history, RFIs, uploaded files (permits, plans, photos), cert PDFs, invoices, pipeline indicator (detail mode), and the complete audit trail for that project.

File Management

Each project can have files attached: permits, plans, photos from inspection, cert PDFs, and invoices. You can upload, download, and view these from the project detail page. Cert PDFs and invoices are generated by the system — you do not need to create them manually.

Revised Permits

A permit file that reaches a project which already has one — a portal upload, a reparse, or a follow-up request submitted through intake — is read twice (Claude Vision and a label read of the page text) and compared with the project’s permit history. It never overwrites the project’s permit fields on its own.

The Permit Revisions card (project page, you and Jacob only) shows what was applied automatically (with Revert), values that need your check (read by only one of the two reads, reads that disagree, or an expiration computed from the county’s lifespan because the permit prints none), permitted-quantity changes with the permit text as evidence, and the different-number question. Check the arithmetic on quantity changes: revisions add to the previous scope (“add two IntelliJacks” on a permit for four means six). When a permit states only an addition and nothing on file gives the base, the card asks you to confirm the total instead of guessing. Apply confirms what is shown; Dismiss drops the proposal.

Permit History lists every permit on file — current or superseded, printed or computed expiration, and the source file. The project header shows Permit revised R01 or 2 permits, and calendar cards carry a small +permit marker while a review is pending or once a revised or additional permit is on file.

Jacob can review and apply the same proposals from #ops-assistant (list_permit_revisions / apply_permit_revision); applying needs an explicit yes in the thread.

Calendar View

The Calendar page shows all scheduled inspections across all branches. It syncs bidirectionally with Google Calendar. You see the full picture — other roles see only their branch.

Use the calendar to understand workload distribution: are inspections clustered in certain areas? Is the field team overbooked on certain days? This directly affects your downstream pipeline — more inspections today means more certs and invoices in 3–5 days.

Schedule Change Requests

When a scheduler requests a change (reschedule, cancel, add, hold), it appears in the Action Center → Schedule tab (/actions?tab=schedule). Both you and Jacob can approve. Types of requests:

The Cancel Button

Every visit row in Site Visits has a Cancel button (you and Jacob only). It replaces three older paths: choosing “Cancelled” in the visit status dropdown, the calendar’s “Hold project & cancel future visits” prompt, and the “Hold project” schedule-change request type. One question, three answers:

When we enter the cancellation, the form asks “When did they tell us?” — a date and time defaulting to now. That is the only judgement call; the system does the arithmetic. The cutoff is 2:00 PM ET on the business day before the visit, weekends skipped and no holiday calendar, so a Monday visit’s cutoff is Friday 2 PM. Visits before 2026-09-20 never carry the fee. You can preview the arithmetic before committing — the form shows the cutoff, the notice time, and whether the fee applies.

When the branch cancels through their own portal or the schedule-change form, no question is asked: the API computes Late Cancel from the time the request was submitted. An approved cancel_inspection cancels the visit; an approved hold cancels the visit and holds the project.

The 45-Day Rule

When a project sits for 45 or more days after its last inspection without being certified, it is automatically transferred to the Triage Team. This is a safety net — it ensures that aged projects do not quietly die in someone’s queue.

When a project enters triage:

• The original scheduler and scheduling assistant lose portal access to the project
• The branch GOA (General Operations Administrator) retains visibility
• The triage team takes ownership and works the project to close-out
• When complete, the invoice still routes to the originating branch’s GOA

Your Role in Triage

You are the final escalation point. The triage team handles most aged projects independently, but some situations require your judgment:

• Client disputes about aged projects
• Projects where the original data is incomplete and the builder is unresponsive
• Revenue decisions — write-off vs. continued pursuit

Your dashboard shows triage counts (projects entering and leaving triage) by day, month, and year, broken down by branch. A rising triage count for a specific branch may indicate a process problem with that branch’s scheduler.

Escalations

Any stakeholder can trigger an escalation on a project. Escalations are accessible from the Action Center → Escalations tab (/actions?tab=escalations). They create an instant notification to responsible parties and open a threaded message channel scoped to that project. You are always notified of escalations.

The Work Pool is the system that allocates post-production work — the step between field inspection and cert/invoice drafting. Currently, Jacob is the primary person working this pool.

How It Works

When a project reaches “Ready for Cert” (auto-advanced from Field Complete by cert validation), it enters the Work Pool. Jacob claims items from the pool and works them through the cert/invoice drafting process. Items can also enter the pool from triage returns and RFI fulfillments.

What to Monitor

• Pool size — how many items are waiting to be claimed. If this grows steadily, throughput is not keeping up with intake.
• Jacob’s active queue — how many items he is currently working. A reasonable load is 10–15 items in various stages of drafting.
• Time-in-pool — how long items sit unclaimed. Anything over 2 business days is a yellow flag.
• Oldest unclaimed item — this is your single best indicator of backlog health.

The system now includes automated intelligence that monitors the pipeline and takes action without waiting for humans to notice problems.

Auto-Validation (Field Complete)

When a project reaches Field Complete, the system automatically runs a 10-point validation checklist. This checks for photos, drive logs, SOW completeness, and other required data.

• Pass — the project auto-advances to Ready for Cert and enters Jacob’s work pool. No human action needed.
• Fail — the system automatically generates an RFI listing exactly what is missing. The SLA clock pauses until the RFI is resolved.

This eliminates the manual step of checking whether a project is ready for post-production review.

Pipeline Shepherd

The Pipeline Shepherd is an automated daily scan that runs at 7:45 AM ET. It examines every active project and detects stuck or aging items using a rules engine:

The shepherd sends WhatsApp alerts to you for critical items and portal notifications for operational items. It runs once per day (idempotent). Reports are stored and viewable at /ops/shepherd-report.

All charts and reporting are consolidated into the Analytics page (under Monitoring in the sidebar). Data is live, sourced from three systems: Supabase (current projects), Xero archive (2022–2024 invoicing), and PMS archive (2018+ historical projects).

Pipeline Dot Chart

The pipeline snapshot now renders as a scaled dot chart (10px dots, up to 120 per row). It blends three data sources with deduplication: Supabase current projects, Xero invoices (unlinked only — those already in Supabase are excluded), and PMS archive records (non-migrated only). All statuses are shown, including terminal states (Invoiced, Closed). The total project count is now 7,000+.

Using Data to Identify Bottlenecks

The three most common bottlenecks in the pipeline and how to spot them:

System Down

If the portal is unreachable or showing errors:

1. Check the Dependencies section on your dashboard (if accessible). Red indicators show which service is affected.
2. Contact your system administrator. The VPS (server) is hosted on Hostinger and can be checked via their control panel.
3. If the issue is widespread (multiple services down), a server restart may be needed. The system is designed to recover automatically — all data is backed up nightly.

Supabase Issues

Supabase is the sole source of truth for all project data. If it becomes slow or unresponsive:

• The portal will show degraded performance (slow loading, stale data)
• Check the Supabase dashboard for service status
• SmartSuite receives a mirror copy during the soak period (through April 6, 2026) but is not used for reads or writes
• Do not approve certs/invoices during a Supabase outage — status updates will not propagate

Cert Delivery Failures

If a cert or invoice email fails to deliver:

• Check the project’s audit log for delivery error details
• Common causes: invalid email address on the account, Gmail API rate limits, attachment size
• Certs can be resent from the project detail page

Client Escalations

When a client contacts you directly about a problem:

1. Look up the project in the portal by address or project number
2. Check the project timeline — when was the last inspection? Is there an open RFI? Is it in triage?
3. Check the audit log for recent activity
4. If a cert is delayed, check whether it is in Jacob’s queue, in your Owner Review Queue, or stuck on an RFI

There is no Messages page. Conversations live inside each project — open a project, scroll to the Correspondence section, and the full thread is there alongside the files, site visits, and certifications it refers to. A red dot on a project row or calendar event means someone said something you haven’t read yet.

What You Need to Know

Red dots on the Projects table and Calendar tell you where unread messages are. Click through to the project to read and reply. All conversations must be linked to a project — no standalone or direct messages. The system tracks response times and resolution SLAs per conversation. Everything is logged in the project’s Activity timeline.

Owner Review Queue — Chat Integration

When Jacob queues a cert or invoice for your review, a ‘review’ conversation is automatically created on that project. You can approve or request revision directly within the thread, and any discussion about the revision happens inline.

Notification Router

Click the gear icon in the notification panel to configure how you receive alerts. Four channels are available: portal notification, email (via Gmail), WhatsApp, and SMS. You can set per-urgency routing — for example, urgent messages go to WhatsApp immediately, while normal messages stay in the portal. Configure:

• Channels per urgency — which channels fire for low / normal / high / urgent messages
• Email fallback delay — how long the system waits before emailing you about an unanswered portal notification (e.g., 15 minutes)
• Do Not Disturb — hours during which only urgent messages are delivered

Notifications consolidate within a 2-minute window to avoid spam. If you are offline, the system falls back to email after your configured delay.

Project Activity Timeline

Every project detail page has an Activity section showing a unified chronological feed: status changes, cert generation, cert delivery, invoice creation, payment receipt, file uploads, site visits, hold events, and messages. Filter by type (All / Messages / Events / RFIs / Files) to focus on what matters.

Friday Command Interface

The ask friday action in any conversation’s compose bar gives you Friday’s Internal voice (Opus model, full tools). Friday’s responses are visible to the entire team. You can issue natural language commands:

• “Approve this cert” — executes the approval from within the conversation
• “Schedule a visit for next Tuesday” — creates the site visit
• “What is the invoice total for this project?” — queries financial data
• “Flag this as an urgent RFI” — converts the conversation to an RFI with priority

Friday Intelligence Tiers

Friday operates on three intelligence tiers, each escalating in capability:

Friday Visualizations

When you ask Friday to chart, graph, or visualize data, it renders directly in the thread — no external tools needed. Three rendering tiers are available:

All visualizations follow Edward Tufte design principles: no gridlines, no chartjunk, muted earth-tone palette, high data-to-ink ratio, data labels where appropriate.

SLA Monitoring

Each conversation tracks first response time and resolution time. Color indicators (green / yellow / red) show SLA status based on priority tier. The Morning Briefing shows open conversations, unanswered messages older than 4 hours, and RFI chats aging beyond 3 days — all linking to the Projects page where red dots guide you to the right projects. The Pipeline Shepherd flags projects with unanswered messages older than 24 hours and stale assignments older than 48 hours in the daily WhatsApp summary.

Portal URLs

Key Actions & Where to Find Them

Who Handles What