Chatbot Project Plan for a Business: Work Back From Your Go-Live Date, Task by Task (Free Backward Scheduler)
Chatbot project plans in small businesses usually fail the same way: someone picks a launch date, the widget goes on the site in ten minutes, and the real work (collecting questions, writing answers, checking facts, training the team) gets squeezed into the last two days. This guide treats the bot as a proper small project. It covers who owns what, which tasks take the time, the decisions to make before you start, and how to work back from your go-live date so you know the day you must begin, with a free scheduler that writes the dates for you.
One owner, one goal, the real questions.
Write, then fact-check every number.
Widget, flows, handover, phones.
Safe pages first, then everywhere.
โก In short
- Installing the widget takes minutes. Collecting questions and writing checked answers takes most of the hours.
- Give the project one owner, one written goal and a short list of what is out of scope.
- Pick the go-live date, then work backwards to the start date. If it is already past, change scope, hours or date, not quality.
- Plan a short soft launch and a review two weeks after go-live.

Why plan it
A small project still needs a start date
Most people think of adding a bot as a technical job. The technical part is small. With a hosted tool you paste one snippet, or install a plugin, and the chat bubble appears. Our installation guide covers every platform in a few minutes each.
What takes time is everything the bot needs to be useful: knowing which questions customers really ask, writing short and correct answers, deciding when a person takes over, setting honest hours and making sure the team knows how to reply from the inbox or the phone app. None of that is hard, but together it is often 20 to 40 hours of work, spread across people who already have full days.
When that work isn’t planned, two things happen. Either the launch date slips again and again, or the bot launches on time with thin, untested answers and customers learn to ignore it. Both are avoidable with a simple plan: a list of tasks, an estimate of hours for each, and a start date worked back from the day you want to go live.
You don’t need project software for this. A notebook page, a spreadsheet or the scheduler further down is enough. What matters is that the dates are written down and someone owns them.
Before you start
Three decisions that save weeks later
One owner
Name a single person who owns the chatbot project. They don’t do all the work, but they decide, chase and sign off. In a small shop this is usually the owner or the person who already answers most customer messages. Two owners means no owner.
One goal, written down
Write one sentence: what should be different a month after launch? For example, “Answer delivery, COD and return questions instantly, and send everything else to us within business hours.” A goal like this tells you which answers to write first and which to leave for later.
What is out of scope
Write a short list of what the bot will not do at launch. Taking payments, handling complaints on its own, giving medical or legal advice, or quoting custom prices are common examples. Scope lists stop the project from growing every time someone has a new idea. New ideas go on a “month two” list instead.
Write these three things at the top of your plan and share them with everyone involved. If you are not yet sure a bot is the right move, the chat bot worksheet helps you decide first, and a short business case helps if a partner needs convincing.
Who does what
Roles in a small chatbot project
Even a two-person business has these roles. One person can hold several; just make sure every role has a name next to it.
Owner
Decides scope, sets the dates, signs off answers and the go-live. Chases tasks that stall.
Question collector
Gathers real questions from WhatsApp, email, phone notes and the counter. Usually whoever answers customers today.
Writer
Turns questions into short, plain answers. Can be the owner, a staff member or a helper who writes well.
Fact checker
Checks every price, date, limit and phone number against how the business really works.
Builder
Installs the widget, sets hours, saved replies, flows and handover. An hour or two with a guide.
Answerers
The people who reply when the bot hands over. They need a short training session and the phone app.
Free tool
Work back from your go-live date
Type the date you want the bot live, how many answers you plan to write and the hours a week your team can spare. The scheduler lists each task with its dates and tells you the day you must start.
Work your plan back from the day the bot goes live.
Go-live date (YYYY-MM-DD):
Answers you plan to write:
Hours per week your team can give it:
Runs only in your browser; nothing is sent or saved. Hours per task are typical estimates for a small business; change your inputs if your answers are long or need a lot of checking.
The tasks
Where the hours actually go
The scheduler uses eight tasks plus an optional soft launch. Here is what each one involves, so your estimates are realistic.
- Agree goal, scope and owner (about 2 hours). A short meeting or a phone call, and a written page.
- Collect the real questions (about 3 hours). Scroll back through a month of WhatsApp chats, emails and phone notes. Write each question in the customer’s words. Group the duplicates. Our question grouping tips show what a typical list looks like.
- Write the answers (about 15 minutes each). Short, specific answers with one next step. Thirty answers is roughly seven or eight hours.
- Check every fact (a few minutes each, plus an hour). The owner reads every answer for wrong prices, old dates and promises the business can’t keep.
- Install and configure (about 2 hours). Widget colour and position, team name, welcome text, opening hours and the away form that collects an email or phone number when nobody is online.
- Build replies, flows and handover (about 3 hours). Saved replies for the answers, a proactive message on one or two pages, and clear rules for when a person takes over.
- Test (about 3 hours). Ask a fixed list of questions on a phone and a laptop, in and out of hours, and note every wrong or weak answer.
- Train the team (about 2 hours). Show answerers the inbox, saved replies, tickets and the Android app.
The writing task is the one that grows. If your answers need input from suppliers or a manager, double its estimate.
Building it
What to set up, in what order
When the answers are ready, the build is quick if you do it in this order.
First, the basics. Create your account at livechatbot.in, add your website and paste the snippet or install the WordPress plugin. Set the widget colour, position, team name and welcome text. Set your opening hours so visitors outside them see the away form instead of waiting for a reply that won’t come. See opening hours.
Second, the answers. Load your checked answers as saved replies. The free keyword bot can answer from saved replies, and the AI Agent answers from the knowledge you give it, so the same checked text serves both. The AI Agent page explains how they work together.
Third, the flows. Add one or two proactive messages at most: perhaps a delivery note on product pages and a help offer on the checkout page. Flows can use conditions such as page, device, country or number of visits. Start small; how to create a chat flow walks through it.
Fourth, the handover. Decide what happens when the bot can’t help: a person replies from the inbox or phone, or, after hours, the visitor leaves details and the message becomes a ticket for the morning.
Only then do the tests. Testing before the answers and handover are in place just tells you what you already know.
Testing and soft launch
Prove it works before everyone sees it
Write a test list of 15 to 20 questions before you test: the most common questions, a few awkward ones, a complaint and one question the bot shouldn’t answer. Ask each one on a phone and on a laptop. Mark pass or fail. Fix the failures and ask again. Keep the list; you’ll reuse it every time something changes. Our regression testing guide has a ready-made set.
Then do a soft launch. Use the “where it shows” settings to hide the chat on pages where a mistake would hurt, such as checkout or payment pages, so it starts on safer pages like contact and product pages until you trust it. Watch every conversation for a week. Read what people actually typed; it is usually different from what you expected. The staged launch plan explains how to choose the first pages.
After the week, widen it to the whole site. If the soft launch turned up many problems, it is fine to extend it by a few days. The go-live date matters less than the first impression your customers get.
When the date is too close
Change scope, hours or date, never quality
If the scheduler says your start date is already in the past, you have three honest options and one bad one.
Cut scope. Launch with the top 15 or 20 answers instead of 40. Most customer questions repeat, so a short list often covers the majority of chats. Put the rest on the month-two list.
Add hours. Can someone give the project two extra hours a week for a month? Writing answers is the easiest task to share, as long as the fact checker reads everything.
Move the date. If the go-live was tied to a sale or festival, consider launching a smaller version before it and the full version after.
The bad option is to keep the date and the scope and skip the checking and testing. That is how bots end up quoting old prices or promising delivery dates nobody can meet, and customers remember that longer than they remember a launch that came a week later.
If your timeline is tied to a busy season, the peak-day playbook has extra steps for surges.
After go-live
Close the project properly
A chatbot project ends with a review, not with the launch. Put a 30-minute meeting in the calendar for two weeks after go-live.
Look at four things. Which questions did the bot answer well? Keep them. Which did it fail or hand over? These become your next answers. How fast did people reply when the bot handed over? If chats waited, adjust hours or the rota. Did anything go wrong in front of a customer? Fix the source and add a test question for it.
Then hand the bot over to normal upkeep. One person checks the unanswered questions weekly and updates answers when prices or policies change. The 30-day launch plan covers the first month in more detail, and our maintenance calendar covers the routine after that.
Finally, revisit the month-two list. Some ideas will no longer seem important; others will have become obvious. Pick the top three and treat them as a small follow-up project with its own owner and date.
Getting started
Make the plan today
Write your owner, goal and out-of-scope list on one page. Put your go-live date into the scheduler, along with a realistic number of answers and the hours your team can spare. Copy the dates into your calendar and share them.
Then start with the task that matters most: collecting real questions. You can create a free LiveChatBot account at livechatbot.in on day one and keep the widget hidden until you are ready. The free plan covers two websites with 100 visitors per website each month and has no per-seat charges, so the whole team can log in and practise. See pricing for packs and limits. A well-planned chatbot project usually takes a few weeks of part-time work, and every hour of it shows in the answers.
Frequently asked questions
How long does a small business chatbot project take?
Usually two to five weeks of part-time work. The widget takes minutes; collecting questions, writing and checking answers and testing take the time.
Who should own the project?
One person who knows the business and can sign off answers, often the owner or the person who already answers most customer messages.
How many answers should I launch with?
Twenty to thirty well-checked answers to the most common questions is a good start. Add more from real chats in the first month.
Do I need a soft launch?
It is strongly recommended. A week with risky pages hidden shows how real visitors phrase questions before the whole site sees the bot.
What if I can’t hit my launch date?
Cut scope, add hours or move the date. Don’t skip fact-checking or testing to save time.
Keep reading
Related guides
Chatbot Price Quotes, Side by Side: Turn Three Different Offers Into One 24-Month Number
Plan the bot, then build it
Work back from your go-live date, write and check your answers, and soft-launch first. Start free on livechatbot.in.