The colleague who already knows your workspace.
She retrieves from your actual contracts, remembers what you've told her, and prepares your renewals before you ask — not a generic chatbot wrapped around your documents.
This has been running in ContractDesk all along.
Donna isn't new. If you've opened the AI panel, asked a question inside a contract, or drafted an agreement with guided help, you've already been talking to her — she just wasn't named publicly until this page. What follows isn't a roadmap. It's what she already does, today, for every ContractDesk workspace.
Ask her anything about your contracts.
She answers from what she actually retrieved out of your own documents, with the specific passages cited — never a guess dressed up as an answer, and never a claim she can't source.
Not connected to a real account or real contracts — a scripted example of what a real, cite-or-decline answer looks like.
what does the Meridian NDA say about liability?
She remembers, and you can always see what.
Most AI products can't honestly tell you what they've inferred about you. Donna's memory is a visible, editable list — every fact she's kept can be read, and deleted, by you, at any time.
Every fact is visible, and every fact can be deleted — try it.
- Prefers 30-day notice periods over 60-day, unless the counterparty is a strategic partner.
- Meridian Consulting's primary contact is their ops lead, not legal, for routine amendments.
- Flags auto-renewal clauses even when inside the acceptable-risk band — wants them called out anyway.
She prepares the work before you ask.
When something needs attention — a renewal, an obligation coming due — she gathers the context and lays out your options in the same six-part shape every time: what happened, why it matters, what she gathered, and what's actually your call.
The Meridian Consulting service agreement renews in 14 days.
Why it matters
This is your third-largest active vendor contract by spend, and the notice window closes in 5 days.
- Contract
- Meridian Consulting — Service Agreement
- Renewal date
- In 14 days
- Notice window
- Closes in 5 days
Your judgment
Renew as-is, renegotiate the payment terms first, or let it lapse — Donna has prepared a draft renewal notice either way, but the decision is yours.
A renewal notice draft is ready to review. Nothing has been sent.
Nothing happens without your yes.
This confirmation step is the product's most important ritual, not a formality. Every action she prepares — assigning, approving, drafting, remembering — stops here until you say go.
Donna has prepared a task: Assign the Meridian renewal review to Priya.
Every consequential action pauses here, every time — nothing executes without this explicit yes.
One colleague, everywhere you work.
Ask about a contract in the global panel, open it, keep asking follow-up questions inside it — she doesn't lose the thread. "It," "that clause," "the other party" still resolve correctly, because it's genuinely one conversation, not a new blank thread every time you navigate.
That continuity is what makes her feel like a colleague you're mid-conversation with, not a search box you reset every time you change screens.
Global panel
"Find the Meridian NDA."
You open the contract →
"What's its indemnity clause say?"
Still the same thread →
"Create a task to review that."
Illustrative sequence — not a live query
Built around a bright line, not a promise.
Every factual claim carries a citation back to the actual passage it came from. When nothing relevant is retrieved, she says so directly instead of answering from general knowledge.
Every consequential action pauses for an explicit, separate confirmation before it executes — no exceptions, no "trusted" fast path.
What she can retrieve is enforced structurally, at the data layer — not by a prompt instruction asking a model to behave. She can't read what you can't.
Routine is mechanically verifiable — she handles that. Judgment binds people, so it's always presented to you, never taken on your behalf.
She is not infallible, and we don't claim she is — like any colleague, she can decline, be corrected, and get things wrong. What's structural is that she'll tell you when she doesn't know, rather than guess.
Where she lives today.
An honest inventory, not a bigger promise than reality — this is exactly what you'll find when you sign in.
Ask about anything in your workspace, from anywhere in the app.
A scoped conversation about the document you have open, in its own margin.
The same colleague, the same memory, helping structure a new contract from scratch.
Where she surfaces the work she has already prepared, waiting on your judgment.
What she doesn't do — yet.
Being clear about the boundary is part of trusting what's inside it. None of this is a hedge — it's the honest edge of what's built today.
- Answer exact, structured questions with a precise filtered list ("every vendor contract over ₹50L renewing this quarter") — today her retrieval is grounded in meaning, not exact numeric filtering. That is on the roadmap, not live.
- Take your voice, or speak back to you — every interaction today is text.
- Watch your whole portfolio continuously, around the clock. Two specific background processes (renewals and obligations) run on an hourly check — real, but narrower than "always watching."
- Execute anything on her own judgment. Every action she prepares pauses for your explicit confirmation — always.
- Answer differently depending on your role. The same Donna, the same voice, for everyone in your workspace today.
Frequently asked questions
No — she's the AI colleague already built into ContractDesk, not an add-on or a separate app. If you've used the AI panel, the assistant inside a contract, or the drafting flow, you've already been talking to her.
No. Every consequential action — assigning a task, starting an approval, drafting a document, remembering a fact — pauses for an explicit confirmation before anything happens. She prepares and stages; you decide.
Yes — every remembered fact is visible in your memory settings, and you can delete any of them, individually, at any time. This is not a hidden model of you; it is a list you can read and edit.
No. She retrieves from your own workspace's real contracts, filtered by what you're actually permitted to read — she can't answer from another organisation's data, and she can't answer from general knowledge dressed up as your document. When she hasn't retrieved anything relevant, she says so.
Not continuously. Two background processes — for renewals and obligations — run on an hourly check today. That is real, ongoing work, but it is not an always-on monitoring dashboard, and we won't describe it as one.
Not today. Every interaction with Donna is text, in the app.
See her working in your own workspace.
The fastest way to judge Donna honestly is to watch her answer a question about your own contract — not a demo script.
No free trial — every review is a paid, lawyer-signed-off outcome. The Contract Health Check above is free and instant, no account required.