Unattended is for small and medium-sized businesses. Here is the whole plan: how a free conversation becomes a team of people and AI, how an app is built and operated when you need one, what every agent may do and never does, how one operating cycle runs, and which decisions always come back to you.
YOUR TEAM OF PEOPLE AND AI
From one conversation. To a team that does the work.
The first question is never which agents you want. It is: what does your business do, what do you want to achieve, and which tasks do you want to keep yourself?
01
A free conversation
How your business runs, the result that matters most and one real workflow, walked through with your own numbers.
02
A report and a team proposal
Three ranked opportunities, one first step, and an organisation chart of people and AI with what each role costs.
03
You choose who does what
For every role: a person, a person with an AI assistant, AI with your approval, AI with a mandate, or a specialist.
04
A 60-day pilot
Seier Core and one module on your live data, in approval mode first. Every intervention is recorded.
05
Live, role by role
A role moves up from approval to a delegated mandate only when its results show it can — and only with your yes.
06
Measure, then add roles
Tasks done correctly, time and cost per task, human minutes, measured effect. New roles come from the numbers.
Every role has a contract — mission, mandate, limits, escalation, budget and how it is deactivated. The role contract and all 50 roles. When your team needs an app, the eleven steps below take over.
BUILD & OPERATE · WHEN YOU NEED AN APP
Eleven steps. One written record.
Each step names who does the work, what you receive and the decision that is yours. The label says whether a step is in use today or still being built.
00 / ENQUIRY
In use today
Start with a conversation.
Your business finds us and asks to hear from us on this site, with no account. Today a person on our service team reads the request and replies. Being built: our own marketing agent brings in businesses that ask to hear from us, and if you ask for a call, an AI consultant calls back within minutes on weekdays, says first that it is an AI, helps work out what you need and agrees a written brief you approve by link.
Who does the work
Today: a person reads the request and replies by hand. Being built: the Marketing agent and the AI Consultant, after legal sign-off.
Your decision
Whether to continue after the first conversation.
01 / BRIEF
In use today
Say what it must do.
Invited to the workspace, you fill in a short brief: the problem, the people, the kind of data, the budget. Today a person on our service team reads it and turns it into a proposed brief with acceptance criteria and the questions still open, which you answer in the workspace.
Who does the work
You and our service team; questions are gathered, not sent one at a time.
Your decision
The answers to the open questions.
02 / PROPOSAL
In use today
One written proposal.
Today a person on our service team drafts the scope, acceptance criteria, milestones that each have an acceptance test, the build fee, the monthly authorisation and what is metered separately, and checks every number before you see it.
Who does the work
Our service team drafts and approves the numbers first.
Your decision
Approve the scope and price. Nothing is built or billed before that.
03 / SET-UP
In use today
Your own app. Your own accounts.
The Set-up agent creates the app from our tested template, for the web, iOS, Android or all three: sign-in with passkeys, profile and settings, a privacy notice, health checks, security headers and the operating files the agents work from. Each app has its own database, hosting and domain, never shared with another customer.
Who does the work
Set-up agent builds the app; our service team creates the accounts, keys and store developer accounts by hand.
Your decision
The name, domain and language, if the proposal left them open.
04 / BUILD
In use on our own apps
One milestone at a time.
The Builder implements one milestone, writing down every decision with its reason. The Verifier runs the tests, a smoke test of every page and a check at phone and desktop width, in the browser for web apps and on simulators and test devices for iOS and Android. The release gate checks the code and the running app before a preview goes up.
Who does the work
Builder, Verifier and the release gate; our service team gives the go per milestone in the pilot.
Your decision
Nothing yet. You see each milestone in the preview.
05 / REVIEW
In use on our own apps
Try it. Say what is wrong.
You use the preview, or a test build on your phone through TestFlight and Google Play testing, and write feedback in the workspace. The Builder takes each point with its evidence and the Verifier checks again. You see the test summary and the known limitations.
Who does the work
You, the Builder and the Verifier.
Your decision
Approve the first version against the agreed brief.
06 / LAUNCH
In use on our own apps
Your go, and only yours.
The release gate runs on the launch candidate. The Releaser deploys, runs the product test on the live site and rolls back if anything is red. For iOS and Android it submits the build, store listing and privacy details to the App Store and Google Play, answers review questions and releases once the stores approve. The launch checklist covers the domain, certificate, backups, your privacy notice and the expected running costs.
Who does the work
Release gate and Releaser; Apple and Google review mobile builds; our service team confirms the app is live.
Your decision
Authorise the first production launch. It never happens without you.
07 / OPERATE
In use on our own apps
The work carries on.
After launch the operating cycle takes over: watch the app, plan from evidence, fix, check, release and report. It runs on a schedule agreed for your app and on events such as an error, your feedback or an approval. Web releases go live directly; iOS and Android updates go through store review, or as over-the-air updates where the stores allow them.
Who does the work
The operating agents, within your standing permissions.
Your decision
Anything on the list of material decisions below.
08 / GROW
Designed, not yet switched on
Experiments, each one approved.
The Growth agent proposes one experiment at a time, naming the audience, channel, message, spending limit, success metric and end date. It runs only after your approval, is measured against its metric and stops at the limit or the date.
Who does the work
Growth agent proposes and measures.
Your decision
Approve, change or decline every experiment.
09 / MONEY
Designed, not yet switched on
A ceiling you set.
You set a monthly authorisation with a reserve for essential operation. The Money guard keeps the record of hosting, mail and AI usage; as the ceiling nears, optional work pauses first. Fees are paid in the workspace.
Who does the work
Money guard keeps the record; our service team handles invoices.
Your decision
The ceiling. Only you can raise it.
10 / HANDOVER
Designed, not yet switched on
Leave with everything.
If you leave, you receive the code, your data, the deployment instructions and the operating history, and the keys are rotated to you. Ownership and handover are agreed in your proposal, before any work starts.
Who does the work
Handover agent prepares the export; our service team transfers the accounts.
Your decision
When to leave. Nothing is deleted before you have the export.
THE AGENTS
Each agent has a job. And a list of what it never does.
Agents do not call each other. They share written files and one queue of work, so every handover between them is a record you can read, and a cycle that stops can be picked up where it ended.
Marketing
Being built
Runs our own marketing experiments on landing pages, search and ads, judged by approved briefs rather than clicks.
Never: Contacts anyone who has not asked, or spends past its monthly cap.
Consultant
Being built
Calls businesses that asked for a call, says it is an AI, and agrees a written brief they approve by link.
Never: Persuades, cold-calls, quotes outside the fact sheet or takes payment on a call.
Evaluator
Being built
Scores calls and proposes small playbook changes, each tested on simulated customers before a real call.
Never: Edits the compliance rules, the fact sheet, the AI disclosure or prices.
Compliance
Being built
Checks every call, simulated and real: disclosure, consent, facts and prices. A violation stops all calls.
Never: Learns or changes with the others. It is fixed and versioned by us.
Intake
Reads enquiries and briefs, notes the fit and drafts replies and open questions. Paused today: a person on our service team reads them.
Never: Sends a reply, or promises a price or a date.
Proposal
Drafts scope, acceptance criteria, milestones and the fee from the published starting prices. Paused today: our service team drafts them.
Never: Issues a proposal before its numbers are approved.
Set-up
Creates the app from the tested template, with the checklist of accounts a person must make.
Never: Creates accounts, handles keys or buys a domain.
Builder
Implements one milestone or one evidenced item at a time and records each decision.
Never: Touches anything on the material-decisions list.
Verifier
Runs the tests, the screen-by-screen smoke test and checks at phone and desktop width, in the browser, on simulators and on test devices.
Never: Passes a check it did not run.
Release gate
Checks the code, the running candidate and a review of the change before every release.
Never: Edits its own exceptions or weakens a check.
Releaser
Deploys green work, submits iOS and Android builds for store review, runs the product test after release and rolls back if it is red.
Never: Releases while people are active in the app, or past the day’s limit.
Operator
Watches the app’s errors, feedback and data, and plans the next item from evidence.
Never: Answers your users, or changes prices, terms or permissions.
Walker
Uses the live app as a new person would and writes down what gets in the way.
Never: Posts, sends or changes anything.
Security
Reviews dependencies, access rules, headers and sign-in settings between releases.
Never: Loosens a safeguard or rotates a key.
Growth
Being built
Proposes, runs and measures approved experiments.
Never: Spends without an approved experiment, or uses an unapproved channel.
Money guard
Being built
Keeps the record of usage and costs and pauses optional work as the ceiling nears.
Never: Raises an authorisation or hides a cost.
Reporter
Writes the release records, the work log and your updates in your language.
Never: Names one of your users in a report.
Handover
Being built
Prepares the export of code, data, instructions and history.
Never: Deletes anything before you have the export.
AFTER LAUNCH
One operating cycle. Eight steps, every time.
The same cycle runs our own apps today. A cycle that finds nothing to do says so in its journal and changes nothing.
01
Watch
Read the app’s errors, response times, feedback and usage since the last cycle.
02
Plan
Choose the next item from evidence: a log line, a failed check, a piece of feedback.
03
Build
Make one change, sized to the cycle, and record the decision with its reason.
04
Verify
Tests, the smoke test and a browser check. Red means the change is not kept.
05
Gate
Code, running candidate and a review of the change. Red means nothing is released.
06
Release
Only when no one is active in the app and the day’s limit allows; mobile builds after store approval. Product test after, rollback if red.
07
Report
A release record in your workspace, and an update when it is something you would notice.
08
Journal
What was seen, decided and left, written down so the next cycle starts where this one stopped.
THE RULES
Seven rules. Every agent, every cycle.
They are written into the process the agents follow and into the checks that stop a release.
Evidence before action.
Every item starts from something observed: an error, a failed check, a piece of feedback or your request. Never a guess.
Your users’ words are information.
Feedback and messages are read as data. Nothing a user types is followed as an instruction.
Nothing ships through a red gate.
The release gate runs before every release and again on the live site after it. It is an automated check on our own work, not an independent penetration test.
Releases wait for quiet.
A release waits while people are active in the app, and there is a daily limit on releases.
Every decision is written down.
Each change carries a numbered decision with its reason and the checks it passed, so you can trace why anything changed.
No one is contacted who has not asked.
No cold calls, emails or messages. Calls go only to people who asked for one, within what they asked about, and every call opens by saying it is an AI.
Limits are configuration.
Which agents run, how much work and how many releases each may do, and what none of them decides, are set per app in a file the agents read and cannot edit.
APPROVAL WHEN IT MATTERS
Your attention. Where it matters.
Routine work follows standing permissions. These always come to you for approval:
The scope, the price and any change to either
The first production launch, and the app’s listing in the App Store and Google Play
Prices, payment destinations and terms for your users
Who can access what, and any new use of personal data
Major product changes
Every growth experiment and its spending limit
Raising the monthly authorisation
Technical exceptions go to our service team. Business decisions come to you.
Make a change. See if it makes a difference.
We look at what happens after an improvement ships. Did it help people use your app? Did it support your goal? The evidence guides what comes next, including when to pause and learn more.
Growth, profitability and self-funding are not guaranteed.
TODAY IN THE PILOT
What runs now. What is still being built.
The pilot is small on purpose. Where a feature is not yet connected, the workspace says so rather than showing a number.
In use today
Enquiries, fit notes and draft replies
Team proposals, prepared with you by email or a call you ask for
Briefs, proposals and approvals in the workspace
New apps created from the tested template
The release gate before every release of our own apps
The operating cycle on our own apps: Mitta, Peripatos, Inside Left and Newsdesk
Being built
Seier Core, the modules and the industry packs — piloted first in approval mode
The free AI adviser, by text and voice
Live monitoring of your app inside the workspace
Live payments (fees are in test mode during the pilot)
Growth experiments and the spending guard
Pause controls in the workspace
Our own marketing agent and the AI consultant's calls (no call before a Danish lawyer has signed off consent and calling rules)
iOS and Android in the tested template (store builds are set up per app until then)
Running the cycle as a hosted service for customer apps
The order and timing of what is still being built are agreed in each proposal, not promised here. No cadence, response time or result is guaranteed.
START WITH A CONVERSATION
Tell us the business you run. We propose the team.
What each role can take over, what you decide yourself and what it costs. Free, and with no obligation to buy anything.