Chasing an insurance claim
A dedicated agent runs a home insurance claim end to end. It drafts the emails, tracks progress and keeps the pressure on the insurer.
Examples
24 things participants have done with their own agent
A personal AI agent works from your own email, calendar and documents, so it can do real work: triaging the inbox, chasing a claim, drafting a proposal, tracking a pipeline or building a small tool. These are real examples from people on our programmes, anonymised and grouped by what they were for.
Life admin
Agents that take the tedious, paperwork-heavy parts of life off an executive's plate.
Checking insurance quotes
The agent compared quotes, caught two that underinsured the car and one with the wrong model, and drafted the correction email to the broker. The owner just pressed send.
Coordinating medical appointments
One agent keeps its owner on track with several doctors ahead of an operation. Another built a shared document so family could support a relative through a procedure, with files, a shared to-do list and drafted emails to the doctors.
Keeping important documents in one place
ID, visa and other key documents are stored in one place. The agent can also search years of email and messages to rebuild records such as an address history.
Tracking days spent in the UK
The agent counts days in the country from email and messages, and updates itself when a new flight confirmation arrives. Useful for tax residency.
Preparing for trips and meetings
Separate agents handle upcoming travel and committee meetings, so the papers, dates and follow-ups are ready before the day.
Nudging better sleep
With wearable data connected, the agent sends an evening nudge when sleep has been short.
Running a business
Agents that do the repetitive operating work of a small business or a senior role.
Building a pipeline tracker
From email, messages and proposals in a shared drive, the agent built a lightweight CRM. It worked out expected revenue and a weighted likelihood for each deal on its own.
Triaging email and approving in one tap
Each morning the agent sorts the inbox into categories such as publish, pitch or press release. The owner approves with a button press and never opens the mailbox.
Drafting proposals
From a short enquiry, the agent found a previous proposal, researched the company, suggested adjusted pricing and drafted the document.
Raising invoices
The agent raises invoices in the accounting package.
Running live dashboards
It combines messages and email into live business dashboards.
Setting up a new company
The agent ran trademark checks across UK and US registries, compared incorporation options, and rebuilt a branding deck around a new name and logo.
Reviewing board papers
An executive has the agent review board papers and spreadsheets and draft emails with the right attachments.
Running a team of sub-agents
One participant offloaded a whole workload. The agent created a sub-agent for each area, with one overseeing the rest.
Finding the one email that settles it
An agent searched years of email to confirm a participant had never agreed to something. Another found a single buried message in a long chain that settled a client question.
Building things
Agents that write software, documents and systems for people who do not code.
Organising a group event
For a group sports day, the agent built the format, a live scoring page, a rule sheet and a fact sheet. When a player was added late, it updated every document and the live scoring, then messaged the group itself.
Rebuilding a project tool
An agent built a working copy of its owner's project management tool and filled it with their actions and projects.
Syncing calendars across providers
An agent built its own tool to sync iCloud and Google calendars. Another participant copied the approach and added Outlook.
Cleaning up files and media
The agent went through thousands of files across many folders, flagged the duplicates and left a filing system ready for approval.
Filing screenshots and links
It files random screenshots and articles into a drive and reports daily. Forgotten good ideas come back as actionable steps.
Standardising research reviews
After one strong research spreadsheet, the participant made that format the standard for every future review.
Tracking restaurant deals
A dedicated lane tracks restaurant deals, discount codes and sentiment from social media.
Gathering teaching material
An educator used their agent to gather quotes for a student workshop on bias and AI.
A closer look: one executive's first 48 hours
The detail behind one of the examples above, from a senior executive in financial services.
Senior executive, financial services
Week 1 · 48h
Built a live personal dashboard from scratch
-
Dashboard
Built a personal morning-briefing dashboard from scratch, pulling live data from their personal Google Calendar, personal M365, Gmail, Apple Health and daily signals into a single view.
-
Debug & patch
Spotted that their personal M365 integration was only returning 9 of 50+ calendar events (owned events only). The agent traced the API issue and patched the underlying CLI tool to use
calendarView- EA-arranged meetings and invites now all show up. -
Timezone fix
Every meeting was showing in GMT instead of the local timezone. The agent traced it to naive datetime handling in the ingest layer and fixed it. All meetings now display correctly in local time.
-
GTD board
Built a full GTD workflow into the dashboard - Projects, Inbox, Next Actions, Waiting For, Meeting Prep, Done, Someday, Reference. Drag-and-drop, colour coding, one-tap push from email or from AI-suggested actions.
-
Design iteration
Iterated the UI through several design directions - light, dark, navy - landing on a navy analytics look with teal accents. Desktop and mobile have separate layouts optimised for each, from a single codebase.
-
Health tracking
Integrated Apple Health data via Health Auto Export, auto-refreshing every 8 hours, with 30-day trend charts accessible on tap.
-
Action
Sent an email from their personal account within seconds of being asked - end-to-end action, not a prompt-and-copy workflow.
Questions people ask
What can a personal AI agent actually do for a busy executive?
It takes on the repetitive, paperwork-heavy parts of the job. Participants use theirs to triage email, chase insurance claims, draft proposals, keep a pipeline tracker, prepare for meetings and trips, review board papers and build small tools they need. Each one runs on the participant's own private server and works from their own email, calendar and documents.
Do you need to be technical to use an AI agent?
No. Most of what is on this page was done by non-technical senior people by asking in plain English, usually from their phone. The agent writes any code itself, and the programme starts with the first useful thing, not the theory.
Is it safe to give an AI agent access to my email and calendar?
Each participant's agent runs on a private server kept separate from work systems, with its own accounts. You choose what it can see and do, and it asks before anything that goes outside your own workspace, such as sending an email.
How quickly do people see results?
Most participants have something working in the first few days: a morning briefing, an inbox triage routine or a dashboard. The six-week programme builds from there.
Every example is anonymised and published with the participant's agreement. Figures and details are as described by participants.