Automation Project Timeline: What to Expect Week by Week

You’ve decided to automate something — a Telegram ordering bot, a CRM integration, an AI agent for customer replies — and now you need to know one thing before you sign anything: how long will this actually take. Vague answers like “a few months” make it hard to plan staffing, marketing, or cash flow.

The truth is a well-scoped automation project timeline is predictable, because most projects follow the same rhythm: understand the problem, build the smallest useful version, test it against real customers, then launch and refine. Below is a realistic week-by-week breakdown you can hold any vendor to.

---

Week 0: Before the Contract — Scoping and Discovery

Nothing should start until scope is written down. This phase usually takes 3–7 days and covers:

• Which process is being automated (order intake, lead qualification, support replies)
• Which channels matter — for most Uzbek businesses that’s Telegram, given its ~76% share of internet users
• Which payment rails need integrating: Payme, Click, Uzum Bank, or card systems like UzCard/Humo
• Which CRM already exists — amoCRM and Bitrix24 are the most common in Tashkent
• A rough price band, based on complexity (simple Telegram bots often run 500 000–1 500 000 so’m; ordering-and-payment bots 1 500 000–5 000 000 so’m; CRM-linked AI agents from 5 000 000 so’m up)

If a vendor skips this step and jumps straight to a quote, that’s a red flag worth reading more about in our piece on red flags in an AI automation proposal.

Weeks 1–2: Architecture and Design

Once scope is signed off, the team designs the actual flow: bot conversation trees, CRM field mapping, API contracts with payment providers, and data storage. For an AI agent, this is also when you decide what the model should and shouldn’t answer, and how it escalates to a human.

Deliverable by the end of week 2: a clickable prototype or flow diagram you can react to before a single line of production code depends on it. This is the cheapest point to change your mind — use it.

What Slows This Phase Down

• Missing access to existing systems (1C, MoySklad, an old CRM export)
• Undecided ownership of content — who writes the bot’s replies in Uzbek and Russian
• Scope creep disguised as “just one more feature”

Weeks 3–5: Core Build

This is where most of the calendar time goes. A developer (or small team) builds:

• The bot or web interface itself
• Integrations: payment gateway (often via a unified library so Payme, Click, and Uzum Bank are handled consistently), CRM sync, and any AI/NLP component
• Admin dashboard or reporting, if scoped

For a focused MVP, three weeks is realistic; a genuinely custom AI agent with multiple integrations can run four to six. If a vendor promises a full CRM-linked AI system “in one week,” ask how — that pace usually means corners are cut on testing, and it’s exactly the kind of claim worth scrutinizing using the checklist in 10 questions to ask before hiring an automation agency.

Week 6: Internal Testing and Pilot

Before real customers see anything, the team runs it against test data and a handful of real staff. Typical checks:

• Does the bot handle a full order-to-payment cycle without a human touching it?
• Does a failed Payme/Click transaction retry or notify support correctly?
• Does the CRM record every lead with the right status?
• Does the AI agent hand off gracefully when it doesn’t know an answer?
• Is there a fallback if Telegram’s API is briefly unreachable?

Many Tashkent teams run a soft pilot with a small customer segment during this week — a subset of Telegram subscribers, or one branch instead of the whole chain — before a full rollout.

Week 7: Launch

Launch itself is usually low-drama if weeks 1–6 were done properly: DNS/webhook cutover, staff training (10–30 minutes for most Telegram-bot setups), and a monitoring window for the first 48–72 hours. In market patterns reported by retailers running similar setups, a well-tuned ordering bot can end up driving 40%+ of monthly turnover within a few months — not an overnight jump, but a real trendline once customers form the habit of ordering through Telegram instead of calling.

Weeks 8+: Stabilization and Ongoing Support

The first month post-launch typically surfaces edge cases nobody scoped: a customer paying twice, an unusual product combination, a Telegram update that changes button behavior. Budget for a stabilization period and an ongoing support arrangement — commonly from around $50/month for monitoring and small fixes, more if the system is complex or growing fast.

Typical Timeline at a Glance

Phase	Duration	What you get
Discovery & scoping	3–7 days	Written scope, price band
Architecture & design	1–2 weeks	Clickable prototype/flow
Core build	3–6 weeks	Working bot/system + integrations
Testing & pilot	~1 week	Verified checklist, soft launch
Launch	2–5 days	Live system, staff trained
Stabilization	ongoing	Bug fixes, support plan

A simple Telegram bot can realistically go from kickoff to live in 3–4 weeks total. A CRM-integrated AI agent is more often 6–10 weeks. Anyone promising a complex system in under two weeks is either underscoping the work or overselling — worth comparing against how vendors are evaluated in how to choose an AI automation company in Tashkent.

Frequently Asked Questions

Can a Telegram bot really be built in one week? A very simple bot — FAQ answers, static menu, no payment integration — yes. Anything touching payments or a CRM needs more time for integration testing.

Why does the CRM integration phase take longer than expected? Because most delays come from data quality on your side: incomplete product catalogs, missing customer fields, or an old CRM export that needs cleaning before it can sync.

Should I pay everything upfront? No reputable vendor asks for 100% upfront. A staged structure tied to milestones (scoping, prototype, build, launch) protects both sides and is standard local practice.

What determines whether a project takes 4 weeks or 10? Number of integrations (payment + CRM + AI agent stacks faster than a single feature), how much content/logic needs to be written in Uzbek and Russian, and how much of your existing data needs cleanup before it’s usable.

If you want a realistic timeline and price band for your own project before committing to anything, see what Fera Tech builds at /#services or browse recent work at /#work, then get in touch and we’ll scope it honestly.