Audiences from an objective
Write what the campaign is for and it proposes audiences you could actually build a list of. Each one is editable, and nothing is written down until you tick it.
AI agents write to the people on your lists from your own domain. Whatever replies lands on a board that tracks it, becomes a quote that prices itself, and ends as an invoice drafted in one click. Then the work turns to keeping them: every account watched, follow-ups written, feedback answered, and what each client could buy next — one client record, in your customers' languages, no hand-offs.
Five surfaces, one engine — agents that go and find the work, a board that tracks what comes back, a studio that prices it, a close that bills in one click, and an account desk that keeps every client and grows it.
Say what a campaign is for and the workspace proposes the audiences, drafts every email in the sequence from what your own knowledge base says, and sends as a named colleague through your own mailbox. Replies come back sorted by what each person actually asked for, scored, and one step from becoming a deal.
A drag-and-drop board over stages you define, each with its own win probability. Column totals stay per currency, the KPI strip keeps a live weighted forecast, and stale deals flag themselves before they go cold.
Pick services and the spec rules resolve the right rate — in the client's currency, on fresh FX. The live preview uses the same renderer as the final PDF, so what you see is literally what they sign.
Live preview = final PDF
Route the approval, send a branded email with the PDF attached, and share a live link your client can accept on the spot. An accepted quote becomes its draft invoice in one click with the tax already right, and every outcome feeds your win rate, cycle time and forecast.
Once they buy, the work becomes keeping them. Today lists the clients who need you — an order that is late, spend that is falling, a complaint to answer — and account programs write the follow-ups: a thank-you after delivery, an answer to feedback, a note when a client goes quiet, a renewal before a contract ends. Growth shows what each client could buy next, valued a year and ranked, with ideas AI spots that the rules cannot.
A model is good at reading unstructured things — a company's website, a Word file's layout, a year of closed deals, the reply somebody just sent. It is bad at arithmetic you can do exactly. So money, permissions and lifecycle are code; drafting, reading, sorting and summarising are AI. Every one of them proposes, and you dispose.
Write what the campaign is for and it proposes audiences you could actually build a list of. Each one is editable, and nothing is written down until you tick it.
Proposes the angles worth trying and drafts the sequence against your objective — grounded in what you told it about your business, never invented.
Describe what the agent is for and Generate with AI drafts it. Every field stays yours to edit, nothing is saved until you press Save, and a new agent gets its photograph when you first save it.
Reads what a reply is actually FOR — pricing, a call, the wrong person, not until Q3 — and can carry more than one. The intents are yours to edit.
Reads your website once, respecting robots.txt, so the agent quotes what you actually say rather than making something up.
Group into themes folds the questions your agents could not answer into a few themes, so you write one knowledge article instead of reading forty questions.
Reads a draft before it leaves the building and says what looks wrong — tone, an anomalous price, a scope that will not survive a lawyer.
When a client spreadsheet's columns defeat the automatic guess, Match with AI reads the headings and first rows and fills in the fields it recognises. Nothing it already matched is overwritten, and nothing goes in until you press Import.
One concrete step for this deal, drawn from its own history — not a generic nudge about following up.
Reads your next six months and says in a paragraph where the weight actually sits, and which month is carrying a deal it should not.
What a year of closed deals had in common. The report that changes what you do next quarter rather than describing last one.
Reads a client's public pages and proposes industry, size, revenue band, founding year, socials — each with its source and how sure it is.
Upload the .docx your clients already recognise and it comes back as a real editable template — tables, art, headers and RTL intact.
Rates from a daily web feed that refresh by themselves every day, every currency your records use added with its rate, and AI only for a currency the feed lacks — each stamped with where it came from and when, and a warning on any quote using a stale one.
On day one it proposes your services, units and starter content from a sentence about what you sell. All of it lands in editable fields.
Reads a client's reply after a delivery and says whether it is praise, a complaint or neither, and how serious — and asks a person when it is not sure. How sure it must be is learned from your corrections.
Adapts each program email to the client — what was just delivered, its segment, what it said — and checks every fact before anyone sees it. A translation, or an answer to a complaint, always waits for a person.
Reads your whole book of accounts and names the few clients and moves worth your week, each tied to the figures behind it — and checked again every time it is shown.
Three points to raise with a client about what else it could buy, built only from its own ideas and figures. A point that fails the check is dropped.
Reads an anonymous sheet of each client's facts and may add up to two more things it could buy — only from what you sell, never with a figure it was not given — valued and ranked like every other idea.
Reads your service names and sorts them into categories and the parts each includes, so Growth tells a real upgrade from a variant in any industry. Anything you saved yourself is left alone.
Nothing a model produces is written to a record without a person accepting it or a validator agreeing. Patchy coverage then degrades gracefully instead of filling your CRM with confident nonsense.
Whatever the task. Weighted credits are a promise nobody can check. A timeout, a provider error or an invalid answer refunds — you got nothing, you pay nothing.
Reading your own data needs no switch. Looking up a named customer on the public web shows every suggestion with its source, and nothing reaches the record until a person accepts it.
Credentials, bank digits, tax numbers and another workspace's data are stripped in code, not asked for in a prompt. Prompts and answers never enter the audit log.
Everything a team needs to find the work, win it and keep the client — from the first campaign to the next order.
Start from six seeded roles — Owner, Admin, Sales Manager, Agent, Finance / Analyst, Viewer — or design your own across more than fifty permission switches. Visibility scopes decide, separately for deals, quotations, clients, leads and reports, whose records each person sees: their own, their team's, or everyone's. Per-user exceptions cover the one colleague who is, well, an exception.
A deal lands on the board — typed in, imported from a spreadsheet, converted from a lead, or auto-created by its first quote.
Build the proposal in the studio. Pricing, currency, VAT and template resolve themselves; the checks watch your back.
Approval routes itself to the right manager. One click sends a branded email, the PDF and a live share link.
Your client accepts online. The deal closes, the forecast moves, and the reports already know why you won.
After delivery the account looks after itself: thank-yous and follow-ups drafted, feedback answered, and the next sale suggested before the client thinks of it.
Spin up your workspace in minutes. Free to start, no credit card — in ten languages out of the box.