How to launch a chatbot successfully:
9 essential steps +1 bonus tip

Natalie Smithson
AI enthusiast | Tea addict | Focused on using AI assistants to win the working week
chatbot ai success

Gone are the days when you needed a bottomless budget, teams of engineers, and many months of workshops and meetings to launch a chatbot. Today, you can get one up and running in less than 15 minutes to meet the needs of every customer promptly, 24 hours a day, without increasing your support teams or spending a fortune, but success hinges on taking the right approach.

Here are the 9 essential steps you should take to make sure your chatbot launch is successful. We’ll also highlight a widely held assumption that often leads people astray, so you can sidestep it with ease.


  • Half of companies are already using AI to better customer experiences and there are many upsides to launching with a recognised provider rather than build a DIY chatbot in-house
  • Large language models (LLMs) like ChatGPT are revolutionising the way we launch and train AI assistants, making it easier and quicker than ever before
  • Defining your unique customer service goals is the most important first step to launching a successful next-gen AI assistant that delivers on customer experience long-term
  • Real world results show Barking & Dagenham Council successfully spared their overwhelmed customer contact team more than 10,000 calls in just six months saving the Council £48,000
  • It’s common to assume the more information you give your AI assistant in the beginning, the more helpful it can be and the faster it can improve, but the opposite is true

Why launch a chatbot for customer service support?

If you want to launch a chatbot to give your customers exemplary service, you’ve likely been thinking about it for a while. Chatbots have been around for decades, but conversational AI technology has come a long way in recent years. The humble chatbot has morphed into a powerful AI assistant that uses natural language processing (NLP) to recognise what your customers want, ready to deliver instant responses. It integrates with any business system you use to manage customer accounts and contacts to personalise experiences.

Quick and effective customer service

McKinsey states “AI-enabled customer service is now the quickest and most effective route for institutions to deliver personalized, proactive experiences that drive customer engagement”. Its popularity is documented by IBM who found almost half of companies are already using AI to better customer experiences, since there are many benefits to customer service automation:

  • Nobody likes to wait for a response to their query. Even if it’s not urgent, getting an instant response from an AI assistant at any time of the day or night can eliminate frustration and quickly drive up customer satisfaction.
  • An AI assistant can respond simultaneously to tens of thousands of customers and that’s a real benefit to you behind the scenes too. You can quickly update your AI assistant with any changes to your business information and it will deliver the most up-to-date detail to your customers across all channels, keeping everyone informed and satisfied with your service.
  • Whatever time of the night or day, in any time zone and using any language, an AI assistant can be there for your customers. It can keep your visitors engaged for longer, helping in any way they need it to, and make all the difference between people browsing and buying.
  • The smartest AI assistants are convenient for your customers to use and give them choices. Not only can people engage with your AI assistant at any time and in any place, they can do this on any channel that suits them, whether that’s your website, a messaging platform like WhatsApp, social media channels, or your mobile app. The customer decides.
  • Communication works both ways and your AI assistant can receive information back from your customers to guide you, letting you know exactly how you can improve your offers or service. Get real time instant feedback from your customers at the end of every interaction.

Hyper-personalised responses, hyper-fast

Adding an AI assistant to your customer service team isn’t just a trendy move. It’s a way to transform your customer journey. With an AI assistant’s instant and tailored support, you can not only solve people’s problems quickly, but also show them you care. An AI assistant can serve the needs of every customer individually with great ease, even when it’s speaking to thousands of them at once. And we all know happy customers become loyal fans.

Real world results show 533% ROI

Launching an AI assistant can have a huge impact on your customer service department. Barking & Dagenham Council, who used an AI assistant to reduce costly phone calls by 6%, increase customer satisfaction by 67%, and free up their waste management team to focus on more complex issues saved £48,000 in their first six months. A few months later they were using their AI assistant to help out another five departments and their Return on Investment (ROI) hit 533%.
⇒ Use our handy calculator to estimate your own ROI

Can (and should) you build your own customer service chatbot?

The short answer is yes, you can build your own chatbot, but it’s not as easy as it might sound, especially since today’s chatbots are now highly sophisticated AI assistants that are as much about customer experience and empathy as they are about coding and technical know-how.

The challenge of building DIY chatbots in-house

Creating an advanced AI assistant requires a lot of time, money, and expertise. You need a whole team of developers to take NLP, integration, cloud hosting, redaction, training, and UI tools and merge them together to create an AI assistant. On top of this, you need data scientists, linguists, writers and designers to help create, train, and maintain an AI assistant that can keep up with the ever-changing needs and expectations of your customers.

It’s a truly multidisciplinary effort and if you’re a very large organisation all set up for this, DIY is an option. If you don’t have this kind of capacity or would rather spend your time and resources developing other areas of the business, choose an experienced provider with all of this baked into a platform that lets you quickly and easily create your own AI assistant, or ask them to do it all for you.

The upside of using an experienced provider

These days, you’ll find fantastic options for getting to a fully customisable AI assistant without the need to hire anyone or spend months in project discussions. In light of this, you have to question why you’d go through all the hassle and expense of launching an AI assistant many months or even years from now when you can do it in minutes online for free and only pay for what you use going forwards.

Use a trusted provider and you’ll benefit from their cutting-edge technology and experience in creating AI assistants for different industries and use cases. They’ll also take care of things you might not have thought of or specialise in, like security and compliance, conversation review, and making sure your AI assistant is scalable for continued success, so you can focus on doing the best for your support teams and customers without any worries or distractions.

9 steps to launching a chatbot successfully so you can aim high and scale fast

If you’re launching your first AI assistant or reinventing an old chatbot, you couldn’t have chosen a better time. It’s never been so easy to dive into advanced AI with no technical know-how, nor has the technology ever been so affordable. Large language models (LLMs) too, like ChatGPT, have revolutionised the way we launch and train AI assistants, making those processes even faster and reducing what is already minimal work you need to do.

On the flip side, there’s also never been so much hype around AI. New tools and platforms are popping up faster than a whack-a-mole on double espresso and customer service leaders everywhere are starting to embrace the benefits of AI automation. It’s important to recognise it might be easy to launch an AI assistant now, but there are appropriate steps to take to make sure it’s highly successful.

Here are the nine steps you’ll need to focus on to get to what really matters to your organisation when you launch an AI assistant for customer service.

1. Define your goals for customer service success

It’s exciting launching an AI assistant. You might start thinking about what’ll it’ll look like or how you’ll choose a name, but before you even select the platform you’ll create it on, decide on the criteria for success to give you the best chance of achieving it. This is the very first and most crucial step because you can use an AI assistant in so many different ways to benefit your business, it’s important to select goals that will make the biggest commercial difference.

Using a framework like Key Performance Indicators (KPIs) or Objectives and Key Results (OKRs), your list will likely include one or more of the following:

  • Reduce contact volumes
  • Increase conversion rates
  • Increase customer lifetime value (LTV)
  • Improve customer engagement
  • Increase customer loyalty
  • Reduce contact team resource costs
  • Reduce customer wait times
  • Get more efficient automation between sales and marketing
  • Get better data to help make decisions
  • Provide a 24-hour service
  • Improve customer satisfaction
  • Learn more about customers

Once you’ve chosen your goals, record how your performance sits today, so you can accurately track your progress weekly after you’ve launched your AI assistant.

2. Get your teams on board

An AI assistant is there to serve your customers, but also support your customer support teams, so make sure everyone’s aware of the launch and knows what you’re working towards. By choosing an experienced provider, you won’t need any engineering or technical knowledge to get started or manage your AI assistant, so any team member can become responsible for it knowing the technology part is all taken care of.

If you work in a highly regulated industry, read our extra guidance to help you prepare for conversations and approval with compliance officers where onboarding is especially strict.

3. Create a strong brand connection

Launching an AI assistant isn’t a once-and-done task. It helps to think of it as a new team member more than anything else. It’ll need training and assessment on its path to becoming a true expert in everything you do, so your AI assistant can support your customers in every way, just like the rest of your teams do. Companies like Stena Line Ferries have been using an AI assistant successfully for almost a decade ― your launch is just the beginning.

Effortlessly become a more visible, dependable brand

Using an AI assistant means you can scale your customer service support in a way that hasn’t been possible before. You can be there for your customers 24/7 in any time zone using any language. Your AI assistant won’t ever need a holiday or sick leave and can do all the tasks your agents wish they could do consistently without distraction. Being able to rely on an AI assistant to carry out routine tasks without any input from them means those workers can focus on developing new ways to increase customer loyalty as the favoured brand within your market.

Get your teams and your AI assistant working strong together and you’ll be able to show a unified support service that customers come to depend on. With that in mind, think about the language and style your teams use to speak with your customers and replicate that with your AI assistant:

  • Is the language formal or informal?
  • Should your AI assistant be soft and caring or clinical and serious in tone?
  • Are touches of humour appropriate in its responses, or should they be simple using plain communication?
  • What kind of mood might people be in when they get in touch? It’s crucial to recognise how people interact with your AI assistant so you can show empathy in all areas of your customer support.

If you have a marketing team or access to a copywriter, it’s good to get their input, but other great sources of reference are any existing tone of voice guidelines, support call scripts, traditional FAQs, and existing content from your website, emails, and social channels.

4. Think about where your AI assistant will live

An advanced AI assistant can live on any channel of communication, whether it’s your website, a mobile app, smart speakers, a social media channel or messaging platform. Thinking about where your unique customer base likes to hang out and the channels they’re most familiar with, decide where to launch your AI assistant so it’s most beneficial to them.

Meet your customers where they are

Customers now expect seamless support across multiple channels and you can rapidly increase the omnichannel customer experience over time, but you don’t have to go all in from the off if it’s not right for you. For most companies, the contact page on your website is a good place to start since customers that want to make contact are already there. Your website home page or main social selling channel is also a good starting point, so customers don’t have to navigate to a second page to find your AI assistant. Add more channels as you need them — it’s a doddle to add more channels and features, but less easy to take them away.

Take one step at a time to build confidence

Add one channel at a time or focus on one page and, if you’re multilingual, start with only one language, and you’ll get a clear picture of the conversations taking place and the customer data coming in, so you can build quickly and with confidence from there.

Wherever you put your AI assistant, be sure to promote it. Failing to promote your AI assistant is one of the reasons why chatbots fail when they shouldn’t do.

5. Decide who’s responsible for your AI assistant

Launching an AI assistant means you’ll need the right people to take care of it – the guardians of your AI assistant. Anyone working closely with your AI assistant should be a subject expert, someone who understands the heart and soul of your business and is committed to giving your customers the best support. They don’t need to be an AI expert or a writer or a data analyst. With an advanced conversational AI platform so much of the work is done for you:

  • You can launch a part-built AI assistant that’s already a specialist in your industry, so you won’t need to craft any initial everyday responses to routine queries, you simply add responses that relate specifically to your operation.
  • Or, using a provider with LLM (large language model) capability, you can automatically pull in content from your website and use that to launch an AI assistant that knows everything there is to know about your business, so you can develop it quickly from there.
  • The more your AI assistant learns about your customers and what they need, the more suggestions you will automatically receive via the platform to make improvements by adding new responses, channels, or integrations that better serve your customers. Simply accept or reject them.
  • Training your AI assistant is also easy with the help of LLMs. On our platform, you won’t need to think of training examples that teach your AI assistant to recognise what your customers need, you can simply generate ideas automatically using AI and click on the ones you want.
  • More than anything, whoever is championing your AI assistant must have a customer-first mindset. Are your customers getting everything they need? Does your AI assistant show empathy for your customers? You can check progress and view reports any time on your AI assistant dashboard and report back to the rest of the team.

6. Try out your AI assistant and fail well

Before you launch your AI assistant to the world, make sure you talk with it to try it out and ask others you trust to do the same. Remember, your AI assistant can only answer questions it’s trained on and this knowledge grows over time, but you can make sure you ‘fail well’.

Avoid customers getting stuck in a loop

When someone asks a question your AI assistant isn’t trained to respond to, or perhaps it’s irrelevant to your business ― asking an insurance company what the football scores are for example, make sure your AI assistant still has a response for them:

  • Your AI assistant has a default fallback message it sends out when it can’t answer and you can edit this response any time to make sure customers get the most appropriate and helpful response.
  • Set up live chat, so your AI assistant can pass customers over to your team if it can’t help or it’s a sensitive subject that needs the human touch.
  • Include your contact details as a last resort, or let customers know when they’ll hear back from you, so they always know exactly what to expect next to avoid frustration.

Focus on overall customer experience

You can preview any responses you add to your AI assistant before they go live to check you’re happy with the content, but when you’re testing out those responses, consider what experience your customers are having on the other side. What vibe do they get? Are they left with any unanswered questions? If you’re nervous about customer reactions, you can always soft launch your AI assistant on a more discreet page of your website and go from there, but the sooner you set your AI assistant live, the sooner it starts to learn and then quickly improves.

7. Keep training your AI assistant

Once your AI assistant’s live and having conversations with real customers, it can evolve at a rapid pace and its capacity for learning overtakes anything we can do as humans. It’s vital to keep a human in the loop to check your AI assistant is always operating at peak performance and we do that by offering a conversation review service. That means you get to rely on our expert team to help train your AI assistant and guide your decisions about how you’ll grow it next without barely having to lift a finger; just accept (or reject) the suggestions you’re given.

Top tip: Your teams can also start to use your AI assistance as a reference library since it will learn and remember everything there is to know about everything you do.

8. Prioritise customer feedback

Having an AI assistant on the support team can transform your customer insights:

  • How are customers feeling? Ask them to rate every conversation using emojis 😠 😒 😐 🙂 😃
  • Where can you improve your service? Customers can tell you about their experience in real time while they’re clearest on what worked well and what could work better.

Real time engagement with your customers is one of the easiest ways to get to know what your customers really think and it’s super easy to do with an AI assistant on the team.

9. Keep measuring the impact of using AI

Once you’re up and running with your AI assistant, it’s time to consistently revisit those goals you set back at the beginning. Are you meeting them? How can you make sure you do?

Act on all the richness of the customer data you get from your AI assistant and make changes to your processes or procedures, to the information you give to customers, or to your products, service and offers any time you need to. Your AI assistant might not be a living, breathing entity, but it’s learning and improving all the time to give you everything you need to improve service. Use it to keep on track with achieving your every goal.

Real world results show progress is fast

In its first six months, Barking & Dagenham Council’s AI assistant spared their overwhelmed customer contact team more than 10,000 calls to ease the pressure and saved the Council £48,000. This gave them more time and resources to work on more rewarding or profitable tasks and in nine months their AI assistant was helping another five of their departments.

Bonus tip: Why less is more when you launch an advanced AI assistant

It’s common for customer service leaders to assume the more information you give your AI assistant in the beginning, the more helpful it can be and the faster it can improve, but we’ve learnt over years of experience in creating AI assistants that’s not the case.

The one idea you should turn on its head

Whether you’re thinking about B2C or B2B best practices for launching an AI assistant, the recommendation stays the same: launch with only 5-20 frequently asked questions and just let your AI assistant learn. You think you know what your customers will ask for, but most of our clients find it surprising and insightful to see what actually happens when their new AI assistant sets to work. The sooner you can get your AI assistant live and speaking to real customers, the sooner you can start to learn what they want, not what you think they want.

Get ahead quickly in the world of advanced AI

Where old-style chatbots needed packing with as much information as you could muster, and even then, you had to hope the customer asked a question in just the right way, next-generation AI assistants are far smarter and learn for themselves as they go along. Being so easy, fast, and affordable to launch, success is in your sight if you launch following the appropriate steps.

Follow our advice to launch your AI assistant and you can enjoy the benefits of having happier customer service teams and happier customers in no time.