Method · From request to delivery

Every request leaves a trace. You can follow it.

A ticket, a documentation page, time logged against it, a dated delivery. Our tools are connected by the ticket key, and you see what concerns you in your project space. AI prepares, a human validates.

One ticket key · one project space · human review before production

Illustration: a design office shown as a scale model; a glowing thread starts at the planning board, runs along the archives, passes under the time clock and along the meeting room, and stops at the desk where the client is received.
FIG. 01 — One thread, from request to delivery · 3D illustration

§ 01 / The thread of a ticket

One identifier connects everything.

The ticket key appears in the documentation page, in the code branch, in the time logged and in the version delivered. It is what lets everyone find everything, us and you alike.

Illustrative example · key PRJ-142

The thread of a ticket, from request to delivery A request arrives by email, chat or video call and becomes a ticket. The ticket has its documentation page. After your approval, the code is written, then reviewed by a human before delivery, which is a dated release with a sign-off report. The same ticket key (for example PRJ-142) connects the ticket, the documentation, the code, the time logged and the delivery. All along the thread, the ticket statuses, a summary of the documentation and the sign-off reports flow into your project space. Same ticket key · PRJ-142 (example) Time logged Clockify Your go-ahead Request Gmail · Chat Meet Ticket Jira Documentation Confluence AI prepares Code GitHub Claude Code AI prepares Human review A principal A human validates Delivery Dated release Sign-off report statuses summary report Your project space · app.sooatek.com Progress, documentation summary, fix status, deliveries The thread of a ticket, from request to delivery A request arrives by email, chat or video call and becomes a ticket. The ticket has its documentation page. After your approval, the code is written, then reviewed by a human before delivery, which is a dated release with a sign-off report. The same ticket key (for example PRJ-142) connects the ticket, the documentation, the code, the time logged and the delivery. All along the thread, the ticket statuses, a summary of the documentation and the sign-off reports flow into your project space. Same key · PRJ-142 (example) Your go-ahead · nothing before Time logged Clockify · same key Request Gmail · Chat · Meet Ticket Jira Documentation Confluence AI prepares Code GitHub · Claude Code AI prepares Human review A principal A human validates Delivery Dated release · Sign-off report To your space statuses summary report Your project space app.sooatek.com Progress, documentation summary, fix status, deliveries
The diagram as text
  1. The request arrives by email, in a shared chat space or on a video call (Gmail, Google Chat, Google Meet).
  2. It becomes a ticket in Jira, with its key.
  3. The ticket has its documentation page in Confluence, titled with the same key: analysis, technical appendix, acceptance test plan.
  4. After your approval, the code is written on a branch named after the ticket key (GitHub), with the help of Claude Code.
  5. Time spent is logged in Clockify with the same key.
  6. A human reviews before any production release; no merge is automatic.
  7. The delivery is a dated release, with its sign-off report.
  8. Progress, a summary of the documentation and the deliveries flow into your project space (app.sooatek.com).

§ 02 / One tool, one role

Each tool has a role. Only one is the source of truth.

We do not add tools for the sake of it: each one has a precise place in the thread.

Role of each tool and what you see of it
Tool Its role with us What you see of it
Jira Tickets, their statuses and the versions delivered. Each request gets its key here.
Confluence One page per ticket: analysis, technical appendix, guide and acceptance test plan. One page per release.
app.sooatek.com Our platform: your project space, our scoping pages, the delivery sign-off reports.
Google Chat For each project, an internal space and a steering space shared with you.
Gmail, Google Calendar, Google Meet, Google Drive Formal exchanges, meetings, video calls, documents and release procedures.
GitHub The code, one branch per ticket, tagged versions before any production release.
Claude and Claude Code Assistance with analysis, development, checks and writing.

Jira, Confluence, Clockify, Google Workspace, GitHub, Claude, OpenAI, Gemini are trademarks of their respective owners. Sooatek uses these tools or models; no partnership or certification is claimed.

§ 03 / The path of a request

Your approval starts the development.

The steps below follow the statuses of our web project workflow. A request only moves to development after your approval.

  1. Step 1: Analysis

    We read the need and the existing code, then we write down what we understood.

  2. Step 2: Proposal

    A scope, an estimate, the exclusions.

  3. Step 3: Your approval

    The ticket waits for your approval. Without it, nothing starts.

  4. Step 4: Preparation and development

    The effects on the application are listed before the first line is written.

  5. Step 5: Tests and review

    The review checks access rights by running the application, not only by reading the code.

  6. Step 6: Delivery

    A dated release in pre-production, then in production, with its sign-off report.

A fix In progress Tests In review Resolved Done

§ 04 / Your project space

What you see, without having to ask.

Your space on app.sooatek.com is synchronized with our tickets. You follow progress there, you report acceptance testing issues there, and you find the deliveries there.

Progress

Tickets and their status, updated from our tracking tool several times an hour during the day.

Where the fix stands

Development, pre-production or production: you know when to test again.

Acceptance testing

You report an issue with screenshots; it goes straight into our triage queue.

One thread per task

Questions, answers and attachments stay with the task, not scattered across emails.

Depending on the project

Deliveries

Each release has its sign-off report, downloadable as a PDF.

Depending on the project

Progress review

A written review every week: what is moving, what is resolved, what is waiting, what comes next.

Illustrative example · fictional data

Illustrative example: the list of a project’s tickets and their environment. Keys and statuses are fictional.

You only see the projects of your organization. Internal technical discussions and estimates in hours stay internal.

For agency partners: the space currently carries the Sooatek brand. As it stands, it is therefore not meant for your end client.

§ 05 / Where AI comes in

AI prepares. A human validates.

AI saves us time on reading, writing and checks. Decisions stay human: scope, merge, production release.

What AI does, and what stays human, at each step
Step What AI does What stays human
Intake Spots requests in emails, chat and tickets; prepares a reply and a follow-up.
Analysis Reads the code without changing it, lists the effects of a request.
Development Writes and checks the code block by block.
Parallel tasks One session per ticket, which pushes its branch and reports in the ticket.
Your space Structures an issue report, summarizes the documentation, writes the progress review.

Which models, for what

  • Claude (Anthropic)

    Our main tool. Claude Code for analysis, development and reviews; the Claude API in the project space, for the three aids described above.

  • OpenAI

    Whisper, an open model from OpenAI, was the basis of our first Malagasy speech recognition version; the model published today is built on w2v-BERT 2.0. ChatGPT and Codex have been tried internally. See the case study

  • Gemini (Google)

    Google Meet note-taking produces some of the meeting minutes. The illustrations on this site are generated with a Gemini image model, then selected and cropped by us.

§ 06 / Your data

What does not leave.

The rules that apply to our tools, to our assistant and to your project space.

  • Read-only by default

    Our in-house assistant reads without changing anything; write access is opened one action at a time.

  • Secrets masked

    Passwords and keys are masked in drafts and attachments.

  • Sensitive projects: our assistant sorts locally

    For projects with sensitive data, emails and messages are sorted on our own hardware by our in-house assistant: they are not sent to an external AI service.

  • Separation

    You only see the projects of your organization.

  • Pages shared by unguessable link

    Scoping pages and sign-off reports are not public.

Where the data lives, and with whom

This website
OVH, in France
Project space (app.sooatek.com)
Vercel; database and files at Supabase
Email and documents
Google Workspace
Contact form emails
Brevo
AI models
Anthropic (Claude); website illustrations: Google (Gemini)

Main providers. Exact locations and the full list of subprocessors are set out in the data processing agreement (Article 28).

§ 07 / Routines

A rhythm in writing.

Fixed meetings, so that nothing depends on someone’s memory.

Every morning
A team check-in.
Every evening
An internal recap of the day.
Every Monday
Planned time is compared with actual time; significant gaps are addressed.
End of sprint or work package
A review with the project manager.
Each improvement
A ticket and a page: the method itself is tracked as a project.

§ 08 / Questions

Frequently asked questions.

Does AI write your code?

It assists with analysis, writing and checks. A human reviews before any production release, and no merge is automatic.

How am I notified of a release?

Each release is a dated version; we notify you before the production release. Your project space shows whether a fix is in development, in pre-production or in production.

Do I have to use your tools?

No. You follow your project in your project space, by email or in a shared chat space. You do not need a subscription with any of our vendors.

Do I get access to Jira and Confluence?

You see the progress of tickets and a summary of their documentation in your project space, without having to navigate our tools.

Is my data used to train a model?

We never train a model on your data without your written consent. On your projects, our AI tools go through business plans: Anthropic’s (API and the Claude Code coding assistant) and Google Workspace for meeting notes; their commercial terms exclude using your data to train their models. For sensitive data, sorting happens on our own hardware, with a local model. The list of tools used on your project is given to you in writing.

I am an agency: is the project space under my brand?

Not today: the project space carries the Sooatek brand. Your deliverables and your client remain yours, and the channel rule applies.

See the method on your own project.

A 30-minute scoping call: we show you how your request would be tracked, from the first question to delivery.