---
# /blocks/personal-assistant/ - the company brain as everyone's personal
# assistant (James, 22 October 2026), the twelfth use case, in the format-2
# structure that /blocks/compliance/ set. Drawn by UseCasePage from the fields
# below (`format: 2`); there is no Markdown body.
# Origin: the brain's design note, personal-assistant-use-case.md (vault
# commit 6e83f16), and the /good-morning skill it builds on. Copy reviewed by
# James on 22 Oct before the build.
# Standing gates, checked 22 Oct:
# - Assistants are "built to connect" (the site's cleared wording); only
#   claude.ai has been tested, so nothing here says ChatGPT connects.
# - Design only, no build status, like every use case. The one status line is
#   the FAQ, which points at /how-we-run-ours/.
# - Left out: voice and a desktop chat box (connector use by voice is likely
#   but untested, vault 31c31d9; never name Gemini), emailed per-person
#   briefings, "set up for 2-50 people" (the site's ICP is 50 to 500), and
#   any comparison with other vendors.
# The outputs are illustrative composites (no client, no real figure), and the
# page says so.
title: 'Personal assistant'
slug: 'personal-assistant'
headline: "Your own assistant, *with your company's memory*."
object: 'What one person needs to know about the company today: who does what, what matters now, the clients and the decisions.'
pain: 'Half my week goes on finding out what a colleague already knows.'
summary: "**Bring your own assistant; we give it your company's memory** - each person keeps the AI assistant they already use, and it answers from what your company knows, with its sources, and only what that person is allowed to see."
description: 'Every person''s own AI assistant, connected to your company brain: answers with sources, briefings against your priorities, and the rules for who may know what.'
metaTitle: 'A company-aware personal assistant, on your company brain'
related:
  - 'account'
  - 'capacity'
  - 'compliance'
order: 12
format: 2

signs:
  - 'We ask the same three people where everything is.'
  - 'Everyone has an AI assistant, and none of them knows who our clients are.'
  - 'Decisions are made in meetings and found again by asking whoever was there.'
  - 'A new starter takes a month to learn who owns what.'
  - 'Preparing for a client meeting means five systems and an old email thread.'

beforeAfter:
  - moment: 'The start of the day'
    today: 'You work out what matters from whatever shouts loudest in the inbox.'
    after: 'You ask for your day and get the few things that matter, each tied to a company priority you own, with who to talk to and where it came from.'
  - moment: 'Before a client meeting'
    today: 'Five systems and an old thread, ten minutes before the call.'
    after: 'You ask it to prep you: who is coming, your history with them, the open next steps and what to get out of it, with sources.'
  - moment: 'Someone needs an answer'
    today: 'You ask around, and the answer depends on who you ask.'
    after: 'You ask your assistant. It answers from the company''s record, cites the note it came from, and only draws on what you are allowed to see.'
  - moment: 'After a meeting'
    today: 'What was agreed lives in one person''s notes.'
    after: 'You say "remember this for the team". It proposes what to file and where, and nothing is filed until you confirm.'

outputs:
  - kind: 'morning briefing'
    title: 'Good morning: three things today'
    fields:
      - ['first', 'Reply to the renewal question from your largest account. It serves the retention priority you own. Talk to the account lead.']
      - ['your meetings', 'Two today. The 2pm is with a prospect you last spoke to in June, and the open next step was a revised proposal.']
      - ['against the priorities', 'Nothing today moves the hiring priority you own. Its review is due on Friday.']
      - ['sources', 'The priorities note, your profile, the pipeline record and the meeting notes, each linked.']
    permission: 'recommend'
    status: 'read before the first meeting; nothing filed'
  - kind: 'answer'
    title: '"What did we decide about the pricing floor?"'
    fields:
      - ['answer', 'The leadership team set a pricing floor in September. Anything below it needs the sales director''s approval.']
      - ['owner', 'The sales director, who reviews it each quarter.']
      - ['what you cannot see', 'The discounts agreed client by client. Your role does not include them, and the answer says so rather than guessing.']
      - ['sources', 'The meeting note and the pricing policy, each linked.']
    permission: 'observe'
    status: 'answered with its sources, within your access'
  - kind: 'filing'
    title: 'Remember this for the team: the client wants monthly reporting'
    fields:
      - ['what it heard', 'On today''s call, the client asked to move from quarterly to monthly reporting from January.']
      - ['where it would go', 'The client''s record, under agreed changes, with you as the source.']
      - ['who will see it', 'The account team and finance. Not the wider company.']
      - ['waiting on', 'Your yes. Nothing is filed until you confirm.']
    permission: 'draft'
    status: 'filed as you, after you confirmed'

steps:
  notices: 'What you ask, in the assistant you already use, and what has moved in your areas since you last looked.'
  connects: 'Your question to the company record: people and roles, priorities, clients, projects and decisions, each with an owner.'
  judges: 'What matters to you today against the priorities you own, and which colleague owns or knows the thing you are asking about.'
  checks: 'Every answer against who you are and what you are allowed to see, before it reaches you.'
  acts: 'It answers, briefs and drafts. Anything it files or proposes waits for your yes, and is recorded as you.'
systems: 'It works through the AI assistant each person already uses, connected to your company brain by one connector built on the open standard AI assistants use, alongside their own calendar and mail. Personal memory stays in each person''s own assistant. The company brain holds only what is shared under your rules.'

runsOn:
  - part: '01'
    text: 'People and roles, priorities, clients, projects and decisions, written down, each with an owner and a date to review it by.'
  - part: '03'
    text: 'Your current priorities and who owns each one, so a briefing ranks your day against them.'
  - part: '05'
    text: 'The drives, notes and meeting records where your priorities, clients and decisions are already written.'
  - part: '06'
    text: 'Who is asking and what they may see, checked on every call, including what one colleague''s assistant may surface about another''s work.'
  - part: '07'
    text: 'Filing a note or proposing a task, only after the person says yes, and recorded as them.'
  - part: '08'
    text: 'The assistant each person already uses, and the panel where filings and answers can be reviewed.'
  - part: '09'
    text: 'Answers tested against questions your team already knows the answers to, before everyone relies on them.'
askIt: 'Who knows most about our largest client, and what did we last agree with them?'
compounds: 'The people, owners and priorities it keeps are what key-person risk reads for cover, what management reporting ranks against and what compliance names as each policy''s owner. It also reads what every other use case leaves, so any of it is one question away for anyone allowed to see it.'
area: 'business'
leaves: 'who does what, what matters now and the decisions behind it'
shares: ['capacity', 'reporting', 'compliance']
staysHuman: 'Anything filed for the team, sent to a client or added as a task waits for the person''s yes, and each person writes their own profile. Nothing is inferred about individuals, and it is not a way to watch people.'

start:
  scope: 'One team, and the questions its people ask each other most.'
  proof: 'It runs on your own priorities and records first, and the team checks its answers to questions they already know before relying on it.'
  connects: 'The assistants your people already use, through one connector, and the drives and notes where your priorities, clients and decisions are written. Connectors are built for you, as part of the work.'
  owns: 'You own the result: the record, the rules about who may know what, and the way it runs, handed over to your team or run for you.'

faq:
  - q: 'Do we have to switch assistants?'
    a: 'No. Each person keeps the one they use. Claude, ChatGPT and others are built to connect through the open standard AI assistants use, so the company brain outlives any one choice of tool.'
  - q: 'Will it know my personal things?'
    a: 'No. Personal memory stays in your own assistant. The company brain holds only what your company has chosen to share, under rules you can read.'
  - q: 'Who can see what?'
    a: 'Only what each person is allowed to see. The rules say what is stored, what is kept out and what one colleague''s assistant may surface about another''s work. Every question is recorded with who asked it and which sources the answer drew on.'
  - q: 'Is this a way to monitor staff?'
    a: 'No. Each person writes their own profile, nothing is inferred about individuals, and it answers questions about the work, not about how people spend their day.'
  - q: 'Is this available today?'
    a: 'Not as something you can switch on. The core of it, asking our own company brain from the assistant we already use, runs in ours today, and it is built for yours as part of the work. The question audit comes first and shows what it would and would not do for you before anything is built.'

backtest: 'Pick three questions someone asked a colleague last week: who owns a client, what was decided about a price, where a policy lives. How long did each answer take, and how sure was the person who gave it?'

# The hero's loop (BlockHeroLoop): a slice of the brain around a question,
# and three stories condensed from the outputs above.
loop:
  hub: 'question'
  nodes: ['person', 'priority', 'client', 'project', 'decision', 'meeting', 'policy', 'owner']
  links:
    - ['question', 'person']
    - ['question', 'decision']
    - ['question', 'client']
    - ['question', 'priority']
    - ['person', 'priority']
    - ['person', 'owner']
    - ['decision', 'meeting']
    - ['decision', 'policy']
    - ['client', 'project']
    - ['client', 'meeting']
    - ['policy', 'owner']
  stories:
    - signal: 'good morning'
      touches: ['person', 'priority']
      steps:
        - 'a person asks for their day'
        - 'the priorities they own'
        - 'what moves them today'
        - 'only what they may see'
      permission: 'recommend'
      outcome: 'three things, each with why'
    - signal: 'what did we decide?'
      touches: ['decision', 'meeting']
      steps:
        - 'a question in plain words'
        - 'the meeting where it was agreed'
        - 'the current decision and its owner'
        - 'within this person''s access'
      permission: 'observe'
      outcome: 'answered, with its sources'
    - signal: 'remember this'
      touches: ['client', 'meeting']
      steps:
        - 'something agreed on a call'
        - 'the client it belongs to'
        - 'where it should be filed'
        - 'who should be able to see it'
      permission: 'draft'
      outcome: 'filed as the person, after a yes'
noindex: false # launched 2026-10-22
updated: 2026-10-22
---