Sales OS
Open workspace
Power-ups

Drafting that reads the deal first

Five agents, each with its own history and templates. They pull the real context off your deal before generating a word — and hand you a first draft to edit, never a final answer to trust blindly.

5 agents20 AI featuresYour team's templatesExport to DOCX & XLSX
The difference

A chatbot starts from nothing. These start from your deal.

Pasting a thread into a general assistant means re-explaining the client, the scope and the pricing every single time, and hoping you pasted everything that mattered. Here the agent is already inside the record — it reads the linked thread, the meeting notes, the estimation figures and your own template before it writes the first line.

What it loads first
  • The linked email thread01
  • Meeting minutes and requirements02
  • Estimation figures already approved03
  • Your team's template for this document04
  • Everything drafted on this deal before05

Then it writes. You edit and send — the draft lands attached to the deal, not in a chat window you will never find again.

The roster

Five agents, five jobs

Each one is narrow on purpose. A narrow agent can be told exactly what to read — and judged on whether it read it.

Drafter

Replies, recaps, follow-ups

You have eleven emails to send and none of them are hard — they just take an hour. Drafter writes them from the actual thread and client history, and you edit and send.

Reads

  • The email thread
  • Client history
  • Meeting notes
  • Your earlier drafts

Hands you

An editable draft, attached to the deal

Research

Know the account before the call

Builds the company and contact picture from web-grounded models, then fills the client record with what it found — so prep is reading a brief, not opening nine tabs.

Reads

  • Company name & domain
  • Known contacts
  • Live web results

Hands you

A brief, plus filled-in client fields

Estimation

Scope into phases, tasks and effort

Turns what was agreed in scoping into phases, tasks and effort on your team's own estimation template, priced in your workspace currency.

Reads

  • Requirements
  • Scoping notes
  • Your estimation template

Hands you

A costed workbook, exportable to XLSX

Proposal

The document the client actually reads

Writes the client-facing narrative from the approved SOW and estimation, so the story on page two and the number on page six cannot drift apart.

Reads

  • Approved SOW
  • Estimation figures
  • Your proposal template

Hands you

A formatted proposal, exportable to DOCX

SOW

Statements of work

Produces the statement of work in a full editor — figures, tables, revision history — instead of a document you rebuild from the last client's.

Reads

  • Requirements
  • Estimation scope
  • Your SOW template

Hands you

A rich-text SOW with figures and tables

Under the hood

20 AI features, each tunable on its own

The five agents are what you press. Underneath, every AI-powered behaviour in the product is a named key in one registry — which is what lets your admin pick the model per feature and see what each one costs.

Drafting

9

Replies, recaps, SOWs, estimations, proposals, figures, and the three template importers.

Pipeline

6

Deal scoring, deal matching, enrichment, lead extraction, and the advisor's signal and recommendation passes.

Research

2

Company briefs and contact research, both web-grounded.

Automation

2

Task generation from context, and the mapping that pushes them to Odoo.

Meetings

1

Meeting summarisation into structured minutes.

No black box

Generation you can audit

Something that writes to your clients is only worth using if you can see what it ran on and what it cost. All four of these are structural, not reporting bolted on afterwards.

A model per feature

Every AI call names a feature key, and each key maps to a model chosen in Settings. Drafting and deal scoring can run on entirely different models without anyone touching code.

Cost recorded per call

Token counts and cost are written to a usage ledger as the call happens, so spend is attributable to a feature, a user and a moment — not discovered at the end of the month.

Structured where it matters

Features that produce data rather than prose — estimations, deal scoring, lead extraction — request JSON-schema output, so the result parses or fails loudly instead of quietly inventing a number.

Overrides, bounded

Admins decide which features you may re-point at your own model choice. Where the default matters, it stays locked.

PreviousWorkspaceThe eight screens you live in, and how a deal moves through them.NextAutopilotTen jobs that keep the record current while you sell.