AI inside the system of record, not beside it.

Dispatch AI answers from your programs, vendors and financials, drafts the writes it cannot make alone, and reads the inbox so your team doesn't retype it. Suggest, approve, automate: in that order, always.

12

confirmable action types, all permission-gated

1

registry chokepoint that fails closed

0

prices or numbers the model invents

6

features in the Dispatch AI layer

One exchange, end to end

It answers with its sources, then asks before it writes.

An assistant that cannot show you where an answer came from is a liability in an operations business. Every specific claim Dispatch AI makes carries the records it read, and anything that would change your data stops at a card a permitted human has to press.

  • Answers cite the records they came from
  • Scope toggle: one program or the whole organization
  • Writes route through one registry that fails closed
JD

Which vendors for the Lisbon congress haven't re-confirmed for Day 2?

Two of nine. Atlas Coaches (26 runs, Day 2 transfers) and Quinta Catering (gala dinner, 840 covers) are still on the original confirmation from 14 May.

vendors · 9 records events · Day 2 contracts · 2 PDFs

Proposed action · needs your confirmation

Send re-confirmation requests to 2 vendors

Uses the Day 2 service lines already on the program. Nothing is sent until you press Approve.

ApproveDiscard requires vendors.edit

01Dispatch AI

Answers that cite their sources, and writes that wait for approval.

A floating assistant on every page, a full workspace when you need it. Ask across the organization; get an answer that cites its records. When Dispatch AI proposes a change, a permitted human confirms it: every time.

  • Scope toggle: one program or the whole organization
  • Reads attendees, flights, vendors and financials, with citations
  • Twelve confirmable action types, from recording actuals to marking payments
  • One registry chokepoint, permission-gated, fails closed
Answers cite the records they came from. Writes are proposed, never taken.

02AI Inbox

Forward the message and review what it extracted.

Every program gets its own inbound address. Vendor quotes, rate changes, rooming lists and manifests are classified and extracted: with the original mail alongside, and the sender's identity verified against your contacts.

  • Classifies manifests, rooming lists, vendor quotes and vendor changes
  • Known contacts verified; unknown senders flagged, never trusted
  • Cost and charge impact pre-filled on the draft change request
  • Nothing commits without review: the hard rule of the whole AI layer
Forwarded mail becomes a draft change request. It stays a draft until someone reviews it.

Suggestion first, approval second, automation last.

The order is the product's opinion, not a roadmap stage. Autonomy is something a workflow earns on your terms, and some things never earn it.

01

AI suggests

Extractions, clusters, drafts and detected actuals arrive as proposals with their evidence attached, never as silent changes.

02

A human approves

Anything touching a vendor, a guest or money is confirmed by someone with the permission to make that change by hand.

03

Trust is earned

Routine, low-stakes flows graduate to one-click bulk accept. Cost-affecting and guest-affecting changes stay human. Forever.

The model selects and describes, and never sets a price.

Every price on a Dispatch AI draft comes out of your own Content Library and price book. Where the data is too thin for an insight, the product says so rather than filling the space.

Dispatch AI questions

What can Dispatch AI answer?

Anything the record holds: “Which vendors handle catering?”, “How many attendees have dietary restrictions?”, “Summarize our active programs”, “What is our cancellation policy?”. Scoped to one program or the whole organization. Every specific claim is supported by your data, and answers cite the records they came from.

Can it take actions, and who approves them?

On Growth and up, Dispatch AI can draft real writes: record an actual, log an interaction, set a follow-up cadence, mark a payment. As confirmable cards. Approval is by the person who asked, or anyone holding the live edit permission, depending on your setting. The registry is permission-gated and fails closed, and every AI-sourced change is labeled in the activity log.

What does Dispatch AI know about us?

What you let it. The Knowledge Base indexes documents you upload plus the chat channels you explicitly opt in, alongside clients, deals, interactions and contracts in your tenancy. Memory, preferences and context learned from conversations, is inspectable, per-item forgettable, and can be switched off entirely.

Does it ever make up prices or numbers?

No. Prices come from your Content Library and price book: the model selects and describes, never costs. And where the data is insufficient for an insight, the product says so instead of filling the space: no seeded numbers, anywhere.

How is AI usage billed?

Metered in credits at one chokepoint, with a monthly allowance for the whole organization on every plan. You can buy packs, enable auto-refill, or cap overage. If credits run out, AI features pause after a five-day grace period: the rest of the product is untouched.

Bring a question your current systems cannot answer.

Bring your most difficult operational question to the demonstration. If the record can answer it, Dispatch AI will, with the source records cited.