DoubleChecked
Start a project
Menu
What we build

Your own AI environment, where company data stays inside.

Your people get the benefit of AI on contracts, client files and reports, without that material going to a public AI service.

Last updated

Start a projectSee what we have built

A private AI workspace is your own AI environment, set up inside systems your company controls, where staff ask questions of contracts, client files and reports. Company data does not go to public AI services. DoubleChecked contracts directly with Anthropic under a data processing agreement, and the workspace keeps an audit log of what was done.

What is a private AI workspace?

It is a place where your staff use AI on your own material. The model, the documents and the record of what was done all sit inside an environment your company controls. Staff ask questions, draft and summarize. The material they work with does not go to a public AI service.

Inside your environment

ContractsClient filesReportsThe AI modelThe record

Outside

Public AI servicesResellers

We contract directly with Anthropic under a data processing agreement.

Zero data retention is available through the API, not the chat interface.

Is it the right fit?

It fits when

  • Client or company material cannot be pasted into a public AI service.
  • Staff already use AI on their own and you want rules around it.
  • Answers must come from your own files, policies and precedents.
  • You need to show who used what, and when.
  • Many roles will use it, each with different access.

It is the wrong tool when

  • Only one or two people would use it, and none of the material is sensitive.
  • Your files are so scattered that nobody could say where they live.
  • You expect it to make decisions that a named person must make.
  • A standard business account from an AI vendor already meets your rules.

What can staff do in a workspace?

These are illustrations of work staff can do inside a private workspace, grouped by industry. They are examples of what is possible, not descriptions of client work. The real work is further down.

CPA and accounting firms

Example

Client letters drafted from the firm's own precedents

Today
A manager searches old letters for something close, copies it and rewrites it for the new client.
The build
The workspace drafts the letter from the firm's own past letters and the client's file. It shows which precedents it used.
A person still decides
The partner edits and signs the letter.
Example

"Where is this in our files" questions across the document store

Today
Staff browse folders and ask colleagues where a prior year memo or an engagement letter was saved.
The build
Staff ask in plain words. The workspace answers from the firm's document store and points to the file and page. It only shows files the person is allowed to see.
A person still decides
The staff member opens the file and confirms it is the right one.

Investment advisers and asset managers

Example

A long client file summarised before a meeting

Today
An adviser skims years of notes, emails and statements in the hour before a client review.
The build
The workspace summarises the file into one page: history, open items and recent changes. Each point links to the record it came from.
A person still decides
The adviser decides what to raise with the client.
Example

First drafts of investment commentary from the firm's research notes

Today
An analyst gathers the team's notes and writes the quarterly commentary from a blank page.
The build
The workspace drafts commentary from the firm's own research notes, in the firm's usual structure and tone. It marks any statement it cannot tie to a note.
A person still decides
The portfolio manager and compliance review it before anything is sent.

Law firms

Example

A deposition outline prepared from the case file

Today
An associate reads the file and builds an outline of topics, exhibits and questions over several days.
The build
The workspace reads the case file inside the firm's environment and drafts an outline by topic, with the exhibit and page behind each one.
A person still decides
The attorney decides the strategy and the questions.
Example

A draft policy compared against the firm's written procedures

Today
Someone reads the new draft next to the current manual and notes where the two disagree.
The build
The workspace compares the draft with the firm's written procedures and lists each conflict and gap, quoting both texts.
A person still decides
The responsible partner decides what the policy says.

Mortgage and lending

Example

A loan file summarised for a second reviewer

Today
A second reviewer opens a full loan file cold and reads it through to find what the first reviewer saw.
The build
The workspace summarises the file: the borrower's documents, the conditions and what is outstanding. Each line points to the page in the file.
A person still decides
The reviewer reads the source pages and makes the call.
Example

Replies to routine customer questions drafted from the firm's own policies

Today
Staff look up the policy and type the same kind of reply many times a week.
The build
The workspace drafts the reply from the firm's written policies and cites the policy it used. Questions the policies do not cover are marked for a person.
A person still decides
A staff member reads and sends each reply.

Operations-heavy businesses

Example

Procedure questions answered from the operations manual

Today
Dispatchers and office staff interrupt a supervisor to ask how an exception is handled.
The build
Staff ask the workspace in plain words. It answers from the company's own manuals and notes and names the section it used.
A person still decides
The supervisor owns the manual and corrects it when an answer is wrong.
Example

Contracts and vendor agreements summarised for renewal

Today
Someone digs out each agreement before renewal to find the term, notice period and what was promised.
The build
The workspace summarises each agreement on request and lists the dates and obligations, with the page for each. Staff can ask follow-up questions of the same text.
A person still decides
The owner of the relationship decides whether to renew.

Have material that must stay inside?

Describe it. We will say what a workspace would take.

Start a project

Built and running

Built and running

An investment adviser registered with the SEC, about 35 staff

Before
Non-public client data could not leave the firm, and there was no governance an examiner could review.
What it does
A private AI workspace for client records runs on dedicated hardware inside the firm. The cloud and network architecture were rebuilt, a SOC 2 program was run, and governance was written for examinations.
Result
Zero non-public records sent to public AI services after the build.

Read the story

Anonymized at client request. Reference calls available under NDA.

How does a build work?

Step 1ScopeOne working session. A plan and a price.
Step 2BuildOn your own documents, shown early and often.
Step 3LaunchYour people trained on their real work.
Step 4RunHosted, watched, improved.

The plan sets the timeline. For larger builds there is an optional scoping sprint: two weeks and a working prototype on your own documents.

  1. Agree which material goes in and who may see which part of it.
  2. Set up the environment, sign-in and roles inside systems you control.
  3. Connect the document stores the workspace will answer from.
  4. Test it on real questions from your own staff and fix what is weak.
  5. Write the rules for use and complete the named security review.
  6. Train each role on its real work, then switch on with the audit log running.

What do you get at handover?

  • A working workspace in your environment

    Set up inside systems your company controls.

  • The record of actions

    Audit logging of what was done.

  • Written rules for use

    Who may use what, and what the workspace may do.

  • Training

    For each role, on its real work.

  • A named security review

    Done before launch.

  • Someone to call

    DoubleChecked runs it: hosted, watched, improved.

How is our data handled?

Company data stays inside your environment and does not go to public AI services. DoubleChecked contracts directly with Anthropic under commercial terms and a data processing agreement, with no reseller in between. One limit to know: zero data retention is available through the API, not the chat interface. A named security review of each build happens before launch. Read how we handle security

Who can get inSingle sign-on and roles
Where it runsA private environment you control
What it may doChecks before it confirms an action
Who signs offA person at your company
What it didAn audit log

Settings we deploy

  • Single sign-on
  • Role design
  • Retention settings
  • Audit logging
  • Spend limits
  • Network allowlisting

What does it connect to?

  • Document management
  • File shares
  • Email and calendar
  • CRM
  • Spreadsheets
  • Line of business systems

Questions buyers ask

How long does it take?

The plan sets the timeline. After one working session you have a plan and a fixed price, and you see working software early and often during the build.

What will it cost?

Every project gets a fixed price after one working session, and you know the number before the build starts. A working session is 45 minutes on one idea from your company.

Where does our data go?

Company data stays inside your environment and does not go to public AI services. Zero data retention is available through the API, not the chat interface. A named security review of each build happens before launch.

Who checks the output?

A person at your company reviews and signs off on outputs that matter. Audit logging keeps a record of what the workspace did, and the software checks before it confirms an action.

What happens when it is wrong?

Staff see the source behind each answer and correct it. The correction is recorded in the audit log, and the rules and sources are adjusted so the same error is less likely.

What do you need from us to start?

One working session, 45 minutes on one idea from your company. DoubleChecked signs an NDA before that first session and replies within one business day.

Why not just use a public chat service with a policy?

A policy relies on every person following it every time. A private workspace keeps the material inside your environment and adds sign-in, roles and an audit log, so use can be reviewed.

Investment adviser

A private AI workspace for client records, with zero records sent to public AI services.

Redaction for AI and court filings

Keep client identifiers out of AI tools and court filings.

How to keep MNPI out of public AI tools

An insight on controlling MNPI by policy.

Security

How we handle data, access and review.

AI enablement

Training that gets your people using AI well.

Solutions

The rest of what we offer.

Tell us what you want built.

A working session is 45 minutes on one idea from your company: what to build, what it needs, and what it would cost.

Start a project
  • A fixed price before we build
  • Your data stays in your environment
  • A record of every action
  • Reference calls under NDA
  • NDA before the first session
  • Since 2019
Start a project