CX Codeaxis OÜ AI apps & automation
The delivery rail

From first note to steady daily use

This page explains how CODEAXIS OÜ delivers a project, step by step, so you know what happens after you write to [email protected]. The process is built for small teams in Estonia who need a working tool, not a long study.

Six stages, one workflow owner on your side, an agreed pilot window, and a written scope before anything is built.

01 Enquiry
02 Discovery
03 Scope
04 Build
05 Launch
06 Handover
Stage one of six

Step one: short enquiry

Send two paragraphs. We answer within two working days.

Send two paragraphs: what repeats, how often, and which tools you use. Example: 80 RFQs per month in Outlook plus an Excel price list. We reply within two working days. If the task clearly doesn't fit AI, we say so and suggest a manual fix. No discovery fee for this triage.

The reply is short on purpose. We tell you whether this is a workflow we would take on, which of the six stages we would start with, and what we would need from you to run the first check. If it is a manual fix — a better Excel rule, a routing rule in your inbox — we say that instead and tell you what to change. Two paragraphs is enough to make that call; you do not need to prepare a deck.

Bring to the first note
  • 01 What repeats, in the words you would use with a colleague.
  • 02 How often it happens in a normal week or month.
  • 03 Which tools hold the data today — inbox, helpdesk, spreadsheet.
A vertical work bench seen from above with tools laid in a single ordered row.

Tools in order, work in order.

Live stage · 01 Enquiry — the note is being read.

The six stages

What happens after your first note

Each stage has an output on your side and an output on ours. Nothing starts before the scope is signed; nothing moves to a second workflow until the first one is running.

02

Discovery call and file check

A 45-minute call, online or in Tallinn, with the person who owns the workflow. We look at sample emails, price rows, or tickets. We check language mix, from Estonian to English, and note terms that must stay locked. You learn what the system would quote and where a human must approve.

Output: a written summary of the workflow and where a human stays in it.

03

Pilot scope in writing

One workflow, one team, four to six weeks. We list inputs, outputs, approval steps, retention, and access. We name the workflow owner on your side. Both sides sign this short scope before any build starts. Changes after start go to a second phase rather than stretching the pilot.

Output: a signed one-page scope and the weekly review slot.

04

Build with weekly reviews

We connect to your email or helpdesk, set up knowledge tables, and tune drafts. Each week you see real outputs on anonymised or live samples, depending on what you approved. Your staff correct drafts directly, which trains both the templates and the team. Model tuning stays visible in a change log.

Output: corrected drafts, knowledge tables, and a change log you can read.

05

Launch with limits

Go-live starts in assist mode: drafts for staff, direct answers only for low-risk topics you name. Confidence thresholds route the rest to humans. We provide a one-page guide covering what the assistant may quote, when to escalate, and how to read logs. Most teams keep this mode for the first month.

Output: assist mode live, escalation rules, and a one-page guide.

06

Handover and upkeep

We hand over source tables, prompt rules, and log access. We agree review dates for price lists and guides. You can run daily use alone or keep us for monthly tuning. Data handling — retention and access — is documented in plain Estonian or English.

The working side

What we need from you

One owner, two hours per week during the pilot, and access to sample files. Delays usually come from missing price rules or conflicting guides, not from modelling. If your files disagree, we pause and ask you to decide. That decision becomes the attributable source.

The owner is the person who already knows when a price is right. Two hours a week is enough because we do the connecting and the tuning; your time goes into correcting drafts and deciding the disputed terms.

Distribution · live stage moved

Weekly reviews run while the work is real

Nobody signs off on a demo they have not seen on their own files.

A night procurement desk from the side with a lit laptop edge, keyboard and a cold mug.

Stage 04 · Build with weekly reviews — real samples, on your desk or ours.

Costs and timing

Plainly priced, plainly scoped

Pilots run four to six weeks because scope is fixed. Extensions add one workflow at a time. We don't bill by vague story points. You get the scope, the weekly review slot, and the handover list before we start.

Write to [email protected] for a scope estimate based on your volume. The estimate names the workflow, the weekly review slot, and the handover list — so you see the whole shape of the pilot before anything is signed.

01

Four to six weeks per pilot because scope is fixed before it starts.

02

One workflow at a time. Extensions add a workflow, never stretch the pilot.

03

Scope, weekly slot and handover list are all agreed before we start.

Before you write

Three questions teams ask us

Straight answers up front. Everything else is worked out on the discovery call or in the scope document.

Do we need technical staff?
No. The workflow owner plus access to email files is enough. We handle connection and tuning.
Where does our data stay?
Agreed per project and documented. We use minimum access and defined retention.
Can we stop after the pilot?
Yes. You keep the documentation and decided sources. No further obligation.
Stage 06 · Handover

Write one note and we will tell you which stage you are on

Send the enquiry to [email protected] or start with the contact form. Two paragraphs is enough to know whether this is our work or a manual fix. We reply within two working days.

Hours
Mon–Fri · 9:00–18:00
Bench
Sepapaja tn 6, Lasnamäe, Tallinn 15551, Estonia