Your documents, answered.
Private chat and search over your own documents, for one team. Ask in plain words and get an answer with its sources, worked out on your own machine or in your own cloud account.
WHAT WE BUILD
Search your documents, prepare work in your tools, or train on your past work when it pays. Three sizes, all on your hardware or in your own cloud account.
01 / DOCUMENT ACCESS
Ask in plain words. It searches only the documents your role may read and attaches the sources.
Which of our past proposals promised weekend site access? I need the exact wording before Thursday.
Three past proposals mention it. Here is a summary, with each proposal attached so you can check the exact wording.
VISIBLE TO / PROPOSALS TEAM ONLY
Two proposals offer weekend access with five days’ notice. One offers it for emergency repairs only. The wording differs, so check it before you reuse it.
The question and the answer are kept in the audit log.
Open each source and check the wording. Maya decides what goes into the new proposal.
Illustration with a fictional firm, not a live client conversation. Who can ask, what it may read and whether any work may use a hosted service are agreed with you in Scope.
02 / THREE SIZES OF BUILD
Build takes four to ten weeks, depending on the size. Each one is a fixed fee, agreed before we start, with the acceptance tests written down in Scope.
Private chat and search over your own documents, for one team. Ask in plain words and get an answer with its sources, worked out on your own machine or in your own cloud account.
Connected to your tools and procedures. It drafts the letter, the reply or the entry, and a person approves before anything is sent or spent.
The Company Agent, plus a model trained on your past work so it drafts the way your firm does. Before and after scores show whether the training earned its place.
Your own system handles the routine work. The hardest part can still go to a hosted service you approve, so the bill drops but does not go to zero. Owning has costs too: hardware, power and Care.
03 / APPROVAL
It can read, draft and prepare. Sending, filing and spending wait for a person, whatever the setting.
READ
Each person gets answers from the documents their role may read, and no more.
DRAFT
The reply, the entry or the summary is drafted with its sources, then waits for a person.
APPROVE
Nothing is sent, filed or spent until a named person approves it. The approval goes into the audit log.
What it may read and draft is agreed in Scope and written into the acceptance tests. It cannot approve its own work.
04 / ACCEPTANCE
Every size is built and tested on your real tasks. It starts with Discovery: two weeks, for a fixed fee agreed upfront, with a roadmap you keep either way.
05 / OWNERSHIP
When all of your AI is rented, someone else sets the price and the limits, and can change both. Owning the routine work takes that back.
RUNNING COST
Routine questions stop running up someone else’s meter. Hardware, power and Care still cost money, and Discovery puts them on your own numbers.
YOUR DATA
Your documents stay on your machine or in your own cloud account, under keys in your name. If any work may use a hosted service, you decide what it can see.
CAPACITY
Your hardware sets the capacity for local work. Discovery sizes it against your tasks and includes its costs in the savings worksheet.
Care keeps it current: monitoring, security updates, and newer open models tested and swapped in as they get better. Cancel any month.