Redact is the privacy layer between your client data and AI. It detects and strips names, account numbers, and identifying details before a document reaches ChatGPT, Claude, or any model, then restores them in the finished result.
$0 today. Billing starts after 90 days, and only if we've shipped.
Redact by hand. Copy the data, delete the names and account numbers, paste, then re-contextualize what you stripped. Hours of repetitive work, and one rushed paste away from a problem.
Stick to the one approved AI. If your firm blessed a single tool, you're locked into one model for every job, including the jobs it's bad at.
Skip AI on client work. The safest option, and the one where you watch peers get faster without you.
Redact exists so none of those is the answer: minutes instead of hours, the best AI for each task, and client identities that never leave your side of the line.
The AI provider never sees the original data. You see everything: what was found, what was changed, and what needed your judgment.
Drop in a PDF, spreadsheet, CSV, or CRM export.
Redact identifies names, account numbers, emails, addresses, and other client PII.
Review the Redaction Report before anything goes to AI. Uncertain items are flagged, never silently passed.
Take the sanitized file to ChatGPT, Claude, or whatever model fits the job.
Client details return to the finished output. The identity map never leaves your control.
A synthetic client file, so you can see exactly what Redact catches before you trust it with a real one.
Margaret Chen · Account 4471-8892-01 · DOB 03/14/1961 · m.chen@example.com
Balance $1.24M across three accounts. Advisory fee 0.95%. Address: 412 Harbor Lane, Mystic CT.
Joint household with David Chen. Successor trustee: Alice Chen, per the Chen family trust.
A synthetic client file. Click Detect to see what Redact finds.
Every document gets a Redaction Report: what was found, by type, with counts, and a clear list of anything Redact wasn't sure about.
That last part matters. A privacy tool that never shows uncertainty is asking you to hope. Redact makes uncertainty visible and gives you the final say, so "I hope this caught everything" becomes "I can see what it found."
Nothing is silently decided. Anything Redact can't confidently classify is flagged for you, not passed through.
Models will keep changing. The layer between your client data and whichever model you choose is what stays.
Already approved for Copilot? Keep it. Redact prepares data safely for the other AI workflows your firm allows. It doesn't replace your compliance program, and it doesn't grant permission to use any AI service. That decision stays with your firm, as it should.
Analyze fees across hundreds of accounts with AI, without exposing a single client identity. The audit that took an afternoon to two days happens in minutes.
Export your CRM, sanitize it, let AI find the patterns, restore the names in the result.
Give AI a sanitized meeting history and get prep notes back, restored and client-ready.
Run holdings through the model that's best at analysis, with names and account numbers stripped.
Draft personalized communications with AI without hand-cleaning every record first.
Defaults are a real example from a real advisor: a fee audit across 500 to 600 accounts that took an afternoon to two days by hand.
You should scrutinize this harder than any other SaaS you buy. These are the commitments Redact is being built to, stated plainly so you can hold us to them. Full security documentation ships before early access opens.
Billing starts at $39/mo per user after 90 days, and only if we've shipped by then. Reserving locks the founding-member rate.
Ship guarantee: you are not charged for 90 days. If we haven't shipped by then, you are never charged at all. If we have shipped and it isn't for you, cancel before billing starts.
Nothing. Reserving costs $0. Billing starts at $39/mo per user after 90 days, and only if we've shipped by then. If we haven't shipped, you are never charged.
No. Documents are processed for your session and not retained after processing completes, and they are never used to train models. The full data-handling documentation ships before early access opens, and it will be specific enough to verify.
Any of them. ChatGPT, Claude, Gemini, Copilot, AI notetakers, and whatever ships next. Redact is the model-agnostic layer between your client data and the AI you choose, so you're never locked into one model because it's the only one that was approved.
You see it before AI does. Every document gets a Redaction Report showing exactly what was detected, and anything Redact can't confidently classify is flagged for your review rather than silently passed through. We will never claim 100% detection; we'll show you the work instead.
Because the manual version only works until it doesn't. One rushed paste, one field you missed, and you have a problem. Redact does what you're already doing, automatically and visibly, every time, and gives you back the hours.
No, and it doesn't try to. Redact helps you run safer AI workflows within your firm's program. Which AI services your firm permits remains your firm's decision.
Pricing is per user, so a small firm can start with one advisor and add seats. If you're thinking about a team rollout, reply to any email from us and we'll set it up together.
Cancel any time. There is no annual commitment and no cancellation fee. And since billing doesn't start for 90 days, you can walk away before paying anything.