Hiring is slow
Good people take months to find, and the work piles up while you look.
PilotNow taking a small group of service organisations
Your next hire, with the Key factor.
Hire an AI employee for the client work your team repeats. It learns your way of doing it, proves itself on your past work, then chases papers, prepares the work and hands it to your reviewer with every source attached. Nothing reaches a client until your people sign off.
No signup needed. Your request goes straight to our team.
- 1pilot run on your own work before it starts
- 3questions at most, asked before the work
- 5rules it never breaks
- 0salaries or seats
Works where your work already is
- Gmail
- Outlook
- WhatsApp Business
- Google Drive
- OneDrive
- Nextcloud, run by us
- Excel
- Word & PDF
- Secure upload links
- E-signatures
- 1,000+ more apps, with your approval
The problem
A service business grows only as fast as it can hire
Every new client means more hours, and more hours mean more people.
Training never ends
Each new person has to learn your methods, your clients and your standards, and some leave once they have.
Quality depends on who did it
The same service comes out differently from different hands, so senior people re-check everything.
What it does
A working day, with an AI team
A Coordinator runs the front desk and AI employees do the services. Every job passes two checkpoints: a ready check before the work starts, and your sign-off before anything goes out.
- Morning
Everyone gets a brief
Each person's own list, in order: work waiting for their review, drafts to approve, questions for them, work at risk of running late, renewals coming up.
- All day
Client mail is handled
The Coordinator reads every client email and WhatsApp message, files the papers to the right job, answers known clients when you've allowed it, and hands the rest to the right person.
- Checkpoint 1
The ready check
Every file is checked before it's used: that it opens, is this client's and covers the right period. A file that fails is never used, and if something it needs isn't known, it asks the right person once.
- On its calendar
Missing papers are chased
It asks the client for what's missing, with a secure upload link, and starts on what has come, unless you've marked a paper as needed first.
- As papers arrive
The work is prepared
Your method, in a sealed workspace with only that client's files. When the rest arrives it carries on from where it stopped.
- Checkpoint 2
Your people sign off
The reviewer is shown the few things that need their judgement, with the rest already checked. Only after approval does anything go to the client.
The Key factor
The highest standard of service work, measured
Every agency promises quality. The Key factor is quality you can check: three things every Keyrole employee is held to, each measured or recorded on your own work.
Right first time
How often your reviewers approve its work without sending it back. That's its Key score.
On schedule
Each service runs on its own calendar. It chases missing papers from the client and starts on what has come.
Traceable
Every figure points to the page or row it came from, and every step it takes is in the audit log.
How a Key score is made96Key score
approved first time · 48sent back · 2
48 approved first time÷50 decided jobs=96
- Made from your own reviewers' decisions
- Last 50 jobs, 180 days; shown after five
- Pilot runs are never counted
Built for review
Made for the person who signs off
Your reviewer is the final check, so their time goes on judgement. The AI decides what needs their eyes on this job, whatever kind of work it is; the platform makes sure nothing that failed a check is left off.
Check these, not everything
Each job opens with the few things that need your reviewer's judgement, each pointing to the cell or passage, and the same notes are in the file itself. What the platform has already confirmed is one line, so nobody re-reads a log.
Check theseRead with focus
Built the same way as the last approved one; two things are new.
- Output tax is 31% higher than last time and the sales register doesn't explain it.Summary › B3 · flagged in the file
- Assumed the August statement is final: tick it or send it back.Assumptions
- One invoice cites a page that doesn't show that amount.Sales › row 14 · flagged in the file
Checked by the platform: 212 of 212 figures traced to their source · calculations re-run, same result · 6 files, all this client's and this period
Example
Every figure has a source
Click a number to see the page or row it came from. Quotes are checked word for word against the document.
Assumptions, ticked
Anything it had to assume is listed, and your reviewer ticks each one before the job can be approved.
Review depth follows risk
What its own checks find sets a job's risk, never its own confidence. High-risk work goes through every sign-off step, and approving it needs a reason.
Work in your own files
Files open in desktop Excel and Word, from your OneDrive or Nextcloud run by us. Comments in the file become review notes it answers.
The full record, when you need it
Every file it read, every step and every calculation, kept for an audit or a question, and out of the way otherwise.
How it learns
It gets better at your work, not just at work
What it knows belongs to your organisation, not to one AI employee. Hire another, or replace one, and nothing is lost.
- Knows what your people know. Your organisation, each role, each service, each client and each contact, kept as notes you can read and correct.
- Reads a client's history first. Past files and threads, with your say-so, so it starts already briefed.
- Trust is earned. A high Key score lets you loosen the review of low-risk steps, one service at a time. You decide, and the Zero Standard never loosens.
What it knows · Acme StudioExample
Client
Invoices go to accounts@, never to the founder. Reports by the 5th.
from approved workContact
Priya prefers WhatsApp and confirms figures herself.
told by your teamService
Show retainer hours used against hours bought.
a reviewer's correctionSuggested
Pays in two parts most months?
waiting for a person to confirmHiring
Hire in four steps, with a pilot run
Describe the role and teach it the services your team repeats, the way you do them. Nothing is generic: every skill is yours.
Every skill is your own: taught from a job you have already done, or explained step by step, and checked against your past work before it works for a client.
Accounting & tax
For exampleMonth-end books · reconciliations · returns · year-end packs
Legal
For exampleContract first review · matter intake summary · monthly matter update
HR & payroll
For exampleMonthly payroll run · statutory payroll reports · new joiner pack
Agencies & consulting
For exampleMonthly client report · proposal from your template · weekly status update
IT & managed services
For exampleMonthly service report · licence and asset review · renewal reminders
Insurance & property
For exampleClaims pack · renewal review · tenant statement
Services organisations teach their AI employees. Plus a Coordinator for your front desk, hired in one step.
- Step 1
Describe the role
Name it, choose its clients, and set who reviews and approves its work.
- Step 2
Teach a service
From a job you've already done, or step by step. It asks only what it still needs to know.
- Step 3 · checkpoint
The pilot run
It does a real past job and is compared with your team's result. It goes live only when you say so.
- Step 4
It starts work
From your inbox, WhatsApp and drives, on each service's own calendar. Its Key score starts counting.
For your clients
Your clients get a page of their own
No login and no app to install. A private link shows each client what you still need from them, how their work is going, and what's ready.
- Upload what's asked for, straight to the right job.
- Download, sign or confirm the work you've approved.
- "Do not contact" is respected everywhere, even for messages already approved.
A
Acme Studiofrom your organisation
ExampleWe still need
Bank statement · OctoberUpload
Card statement · OctoberUpload
In progress
October report
Ready for you
September reportSign
The Zero Standard
Five rules it never breaks
Each one is enforced by the product, not by a policy document, and each one you can check yourself.
0
sends without approval
Nothing reaches a client until a person approves the job.
Each job shows who approved it, and when.
0
figures without a source
Every number carries the page or row it came from, or is flagged.
Click any figure to see its source.
0
files from other clients
Each job runs in a sealed workspace with only its own client's files and no internet.
The audit log lists every file it opened.
0
mistakes repeated
A reviewer's correction goes into its memory, so it's made once.
Its memory shows what it learned, from whom.
0
steps off the record
Every step goes into a tamper-evident audit log.
Verify any signed document at /s/check.
Pricing
Set up once. Then pay for the work it does.
No salary, no seats and no annual lock-in. A one-time setup fee, then only the work it does.
Prepaid credits
Buy credits in advance and each job draws on them. Packs start at USD 10, or ₹500 in India, with tax added at checkout.
- See what each job cost you
- Nothing is spent while it waits for work
- Top up whenever you like
A limit you set
A monthly spending limit for your organisation, so a busy month never surprises you.
- Everyone is warned at 80%
- New work pauses at the limit
- Owners raise it in one place
A one-time setup fee
We set up each AI employee with you: its role, its services and its pilot runs on your past work. The fee depends on how much setting up your services takes, and we quote it before you start.
- Your mailbox, WhatsApp and drives connected
- Your clients and their history brought in
- You decide if it goes live
The pilot
Bring the Key factor to one service first
We're working closely with a small group of service organisations. We'll set up your first AI employee with you, on one of your own services, and run its pilot on your past work before it does anything for a client.
- Any country, any industry
- Your first service taught with us
- You decide if it goes live
FAQ
Frequently asked questions
How is this different from an AI assistant or an AI agent?
An assistant waits for you to ask. A Keyrole employee has a role, its own services and calendar, clients it looks after, a reviewer and an approver, and a measured record. It notices what has come in, chases what hasn't, prepares the work and puts it in front of the right person, the way a good associate would.
Can it email or message my clients on its own?
Only as far as you allow. Prepared work never goes out before approval. The Coordinator can answer routine questions from addresses it knows are the client's, if you turn that on; anything else it hands to a person. Changes to outside systems also wait for approval.
Does it replace my staff?
It takes the repetitive preparation off them: reading documents, preparing deliverables, chasing what's missing. Your people stay the reviewers and approvers, and the client relationship stays with them. It's how a team takes on more clients without hiring for every one.
What happens when it gets something wrong?
Your reviewer sees it first, with the source of every figure to check against. The note goes back to the AI employee, which revises the work and remembers the correction for next time. It also shows in the Key score, so you always know where it stands.
Which tools does it work with?
Gmail or Outlook, WhatsApp Business, Google Drive, OneDrive or Nextcloud run by us, or files you upload, with secure upload links for your clients. More than 1,000 other apps can be connected, and it suggests the ones each service needs; it only reads them until a job that changes something is approved. It delivers spreadsheets, documents and letters on your letterhead, structured import files, and documents for e-signature.
How much does setup cost?
It depends on the work: how many services the AI employee takes on, how many steps and checks each one has, and how many systems it connects to. We look at your services with you first and quote a fixed setup fee before you commit. After that you pay only for the work it does, from prepaid credits.
Is our clients' data used to train AI models?
No. It's used only to do your organisation's work. Each job runs in its own sealed workspace, and you can erase a client's files and memory at any time.