dpaperwork

Solution · Canon

Ask about a client six months later. It answers like it was there.

What you agreed with a client, why you changed the retainer, how they like to be updated. It lives in a few heads and old email threads. When a new hire joins, or you have a busy week, every conversation starts from zero.

The context walks out of the room

Generic AI starts every chat blank, so its drafts read like they were written by a stranger. dpaperwork keeps a memory of your company and recalls the relevant parts for every draft, answer and Skill run. The proposal reflects the last decision. The report reflects what actually shipped.

It does the legwork. You make the call.

Nothing reaches a client without your yes, and every draft shows what it changed.

1 · Notices

Brightline proposal needed for the new quarter

Spotted by the Monitor, without you looking.

2 · Drafts

Drafted using their preference for async updates and the quarterly retainer you agreed in September.

Grounded in the record’s real history.

3 · You decide

Approve, edit, or skip

ApproveEditSkip

4 · Done

Sent from your own address

Sent

Four kinds of memory, recalled automatically

Company context is your profile and is always included. The activity log is written for you: deals closed, cycles finished, documents sent, Skills run. Decisions are the calls you make: when a newer one covers the same topic, it offers to replace the old one and you confirm. Definitions are your own meanings for terms and measures. Only the entries that matter to the question are brought in.

Four kinds of memory

Your context, an automatic activity log, the decisions you have recorded, and your own definitions.

Recalled by meaning

It searches by meaning, not keywords, so only what is relevant to the question comes in.

Every output starts grounded

Drafts, chat answers and Skill runs all read the memory first. Nothing starts blank.

Same week, different ending

A new hire asks what was agreed and nobody remembers.

The answer comes from the logged decision, with its date.

A proposal ignores what you agreed last quarter.

The draft uses the current decision and the client’s known preferences.

A report is written from scratch from scattered notes.

It is drafted from the activity log, and you edit it.

What’s doing the work

Company context: your profile, always included in full, and it grows as you work
Activity log: an automatic trail of deals closed, cycles finished, documents sent and Skills run
Decisions: explicit calls you have made. It offers to replace an older one on the same topic and waits for your confirm
Definitions: what your own words and measures mean, like "at-risk" or "billable utilisation", edited in place, with an option to keep one out of the AI
Weekly summaries are drafted for your review before the memory treats them as settled
Correct the AI once in a chat and Canon proposes it as a Candidate. Accept it and everyone uses it from then on
Every chat answer carries a receipt: what it was sourced from, and what it assumed
Skills: your Skills read as context. Chat can offer to run one, and it only runs on your approval
Every draft, chat answer and Skill run reads Canon first
The memory shapes drafts but never acts. Client-facing output still waits for your accept
Built onCanonChatDocsSkills

Questions

Get early access.

A company memory your drafts start from, and your approval before anything goes out.