Documentation

Scope Change Intelligence

Point it at a client email or Teams thread, and it reads every change request against your contracted project — priced, classified, and ready to draft onto a change order.

What It Does

When a client sends "add two more displays, drop the ceiling mics, and swap the DSP", someone has to turn those sentences into priced change-order lines. Scope Change Intelligence does that reading for you. It compares each request against the contracted version of the project — the frozen baseline plus every approved change order — and computes what each change is actually worth: equipment price at your frozen contract rates, install hours, programming hours, and configuration hours.

The AI reads and classifies; the numbers always come from your own contract math — the same pricing the change-order editor and the customer document use. And the agent never commits anything: everything it drafts lands on a draft change order that you review, submit, and approve exactly as you would one you built by hand.

The Intelligence Platform panel at the Review step: an amber banner noting the thread was analyzed before, a Low-risk changes pile holding two changes with an Accept all button, and a review card for a new Crestron CP4N control processor asking which room it belongs in

The review board: safe changes collect in the low-risk pile, everything else gets its own card with the computed price and hours.

The screenshots through this page follow one real run: a client emailed a project asking for another control processor, a display quantity change, and an item that turned out not to be on the contract at all. Here's how that lands, step by step.

Before You Start

  • The project needs an approved contract — the workflow prices changes against the contracted baseline, so a project still in estimating has nothing to compare against.
  • Your email or Microsoft Teams account must be connected, and the thread you want to analyze must be in your own mailbox — thread content stays private to the person who starts the run.

Starting a Run

  1. Open the project and click AVIP ASSISTANT in the header to open the panel.
  2. On the Scope change tile, pick your source: Email or Teams.
  3. Search your threads by subject, or leave the box blank to list them all. Pick the thread that carries the change requests.
  4. The run moves through Locate thread → Analyzing → Review. Analysis usually takes under a minute; changes appear on the board as they are found.
The project header with the AVIP ASSISTANT button on the right, next to the project's action buttons

Step 1 — AVIP ASSISTANT lives in the project header.

The AVIP Assistant workflow tiles, with the Scope change tile offering Email and Teams source buttons

Step 2 — the Scope change tile, with Email and Teams as sources.

The thread picker listing three client email threads from the same sender; two carry amber notes reading 'Analyzed Jul 31 — drafted CO-004' and 'Analyzed Jul 31 — drafted CO-005'

Step 3 — pick the thread. Threads a run already read carry the amber note, with the change order they drafted.

Already-analyzed threads say so

A thread another run already read carries an amber note in the picker — "Analyzed Jul 31 — drafted CO-002". If you select one anyway, the board tells you whether the thread has changed since: if it hasn't, applying findings again would draft a second change order for the same requests.

Reading the Board

Every change the thread asks for becomes a card. The headline figure is the scope price — the change in equipment sell before tax. Below it, the hours the change moves: install, programming, and configuration, each shown only when the change actually moves them. Credits show negative, in both money and hours.

  • Low-risk changes — quantity and like-for-like changes the agent judged safe collect in one pile with an Accept all button. One optional reason covers the whole pile.
  • Review cards — anything needing judgment gets its own card with the agent's reasoning and one or more options, the recommended one marked.

Cards also cite your history: if the project previously approved the same kind of change — or dismissed this exact product — the card says so. That's your own past decisions informing the recommendation, which is why decisions ask for a reason.

Acting on a Finding

  • Apply — drafts the change onto the run's change order, exactly as priced. You'll be asked for a short reason; pick a chip or write your own. The next run reads these, so a stated reason is what makes the history worth anything.
  • Dismiss — records that the change is not being made, with a reason. Dismissals are precedent too.
  • Instruct — click the chat chip on a card and type the change in your own words: "make it 6 units", "use the Biamp instead". The agent prices your version and shows it before anything is drafted. If nothing can be drafted from the request as it stands, it tells you what's missing instead of offering a dead button.
  • New items ask for a room — a product the contract doesn't have yet needs a home before a line can be drafted. Pick the room on the card and the option rebuilds with real numbers.
The Crestron CP4N review card after picking Huddle Room: a scope price of +$2,325.00, with the recommended 'Add the requested item' option and its Apply button

The new-item card after picking a room — the option rebuilds with the real contract price, ready to apply.

The 'Draft this change' dialog asking for a reason, with quick-pick reason chips and a free-text box above the Draft and Cancel buttons

Every apply asks why. Pick a chip or write your own — the next run reads these.

What Gets Drafted — and What Doesn't

Each run drafts onto one draft change order, created on first apply. When every finding is decided, the run completes and offers Open change orders. The draft is internal-only until you tick Customer charge, and nothing is submitted, sent, or approved by the agent — the change-order page remains the only place that happens.

The board at Run complete: all three findings decided — the display quantity change marked ACCEPTED, an off-contract product marked DISMISSED, and the CP4N marked APPLIED at +$2,325.00 — with an Open change orders button

Every finding decided — accepted, dismissed, or applied — and the run completes.

The drafted change order in the editor with DRAFT and INTERNAL-ONLY badges: an Add line for the Crestron CP4N in Huddle Room at +$2,325.00, a Modify line for the Samsung displays at $0.00, and a running total showing Increase +$2,325.00

The result: an ordinary draft change order, with each applied finding as a line. From here it's your normal review-and-submit flow.

The agent drafts; you commit

Applying a finding creates draft change-order lines and nothing else. Submitting for signature and approving stay manual steps, with the same review you'd give any change order.

When the Thread Moves On

If the client replies after a run froze its reading, the board shows "The thread has moved on" with a Re-read the thread button. Re-reading supersedes only the findings nobody has decided yet — anything already drafted stays on the change order.

Questions Mid-Run

The chat rail belongs to the run while one is open: type into Ask about this run… to ask about what was found ("why is the DSP swap flagged?", "what does this do to margin?"). Answers arrive as chat bubbles without touching any finding.

Closing, Discarding, History

  • Close returns to the workflow tiles; the run keeps going and reappears when you come back.
  • Discard abandons the run. Lines already drafted stay on the draft change order — discard the change order itself from its own page if you don't want them.
  • Every decision is recorded with who made it, how, and why. Runs and findings are visible to your whole team; the thread's content stays private to whoever started the run.

Credits

Each analysis step and instruction uses AI credits from your organization's balance. See AI Credits for usage and billing.