Do you remember the last time you called your bank to complain about a payment delay or check your balance? How long did it take you from hearing monotonous and boring ringtones to talking to a human operator? And even more important — what emotions did you feel while waiting for your turn in an endless queue of calls? Did you get irritated or furious? Did you wish you’d rather spend this time for something useful and meaningful? I bet you did. And then, when your call was finally accepted, did you throw your fuming and irritation straight to the operator who just did their job and provided service to a different client while you were waiting? I bet you did.
If we look at this situation from a business perspective, we’ll see the following picture: your clients are unhappy that their calls aren’t responded immediately. On the other hand, you have unhappy employees who’re stressed out by unhappy clients — such a vicious circle. So, as a business, you’re at risk of losing both a client and an operator. To eliminate this risk, you now have an excellent option to build and deploy a chatbot and supercharge it with artificial intelligence.
Although building robust conversational AI solutions requires time and money, the outcome will pay off and provide an appropriate ROI. If developed and deployed the right way, chatbots offer great UX and prove to be much more efficient than running a 24/7 support team or building a native mobile app.
To explore what it means to build chatbots the right way so that they drive business value, I’ve talked to an expert from the company that builds custom chatbot solutions. I also checked out what the crowd expert has to say about it.
A chatbot can perform all repetitive actions that can be arranged in the form of dialogue. From answering simple questions and showing directions on a map to adding products to the cart, segmenting catalog items, providing automated support service, collecting user feedback, booking hotels and flights, buying tickets, processing payments — nearly every single interaction between brand and user can be pre-programmed and streamlined with the right chatbot solution. If you’re a small business selling handmade toys on Facebook, you don’t need a website anymore. Instead, you can spend money on building a high-quality AI-based chatbot for FB messenger to create an automated sales funnel and strain leads across all stages leading them to buy your toy.
Everything starts with an idea and its validation.
“The main thing is to understand what to expect from your chatbot and what KPIs and success factors you want to measure,” says Kate Haskova, Head of Software Delivery at 8allocate.
“ You can’t launch a chatbot just for the sake of hype or because it’s trendy. You’ll end up wasting your time and precious dollars on building a completely useless product if you don’t know its mission and can’t communicate it properly to all stakeholders and developers.”
Afterward, according to Kate, it is necessary to describe the logic of chatbot performance based on its goals and mission. Make sure to answer the following questions before developing your chatbot:
- What rich media and content do we need to ensure the effective work of our chatbot?
- What rules will apply to each particular piece of content to be distributed through our bot channel, and how will it match each specific user request?
- What backend integrations are required for the chatbot to do its job? (e.g., CRM, database, price lists, 3rd party platforms, payment gateway, etc.)
- Will it suffice to set “what if” conditions or should we train our chatbot with machine learning and NLP algorithms?
According to Kate Haskova, a typical technical requirements specification for custom chatbot solution development includes more than 20 items, sometimes more than 50, and it’s up to you to decide whether to go with custom chatbot development and hire a professional team or company with extensive expertise and know-how, or build a DIY solution powered by one of many automatic chatbot builders available in the market nowadays.
Because chatbots have three main use cases which are user support, sales, marketing and HR automation, and mobile app alternative, it’s crucial to decide from the very beginning what company department will own and manage the chatbot and who’ll be responsible for the outcome. The chatbot cannot be a one-size-fits-all solution to be effective and bring value to the business.
All professional chatbot developers agree that only a well-crafted and custom-tailored chatbot solution can benefit business regardless of its size and objective.
Chatbot development should be treated just like any professional software product development, and it should undergo all typical stages of product development, from ideation and discovery through to development, testing, final release and support.
Kate Haskova of 8allocate agrees that your decision to build a custom or DIY chatbot should be based on your business objectives and other factors such as availability of budget, access to AI dev talent, resources, timeframe, etc.
“It makes sense to use automatic chatbot builders if you need to check and validate hypotheses about your future chatbot performance and how it’ll help achieve your business goals. Or if your chatbot features a very simple functionality such as arrange email send-outs, conduct basic surveys with no complex logic, or automate your sales funnel.”
There’re many smart chatbot builders flooding the internet today including Chatfuel (which can process up to 5,000 subscribers for free and more for an extra fee), Botsify (which integrates easily with Shopify, WordPress and Alexa), ChatterOn (which AI and machine learning algorithms are pretty solid to support a wide range of rich content, including carousels, buttons, photos, gifs and videos), etc.
One more DYI solution that proves effective for simple tasks is Google’s DialogFlow which enables chatbots to understand natural language and be optimized for a range of platforms and devices. What makes this particular chatbot builder platform different from similar tools is its default integration with Google Analytics where you can build custom reports, track stats and visualize data pertaining to chatbot dialogues with the user.
However, you need to remember that most of the chatbot builders are tailored specifically to Facebook or Telegram and let you leverage their proprietary subscriber databases. If you decide to build a custom chatbot in the future, you’ll lose the lion’s share of your potential outreach and need to think about data migration mechanism as early as possible. While there’re some, none gives a 100% guarantee your data will be migrated properly.
If your chatbot is expected to leverage and extract big data from external sources such as CMS or data brokers and modify and match them with NLP, there’s no way for you to avoid custom chatbot development.
If you decide to use custom development, your steps will depend on the type of chatbot you’re building: conversational or transactional.
According to Jorge Peñalva, CEO at lang.ai, “in conversational chatbots, you have data, you are trying to improve customer service, and thus you have customer questions and agent’s replies.” As such, steps in making a client-tailored chatbot will be as follows:
- Train an ML model to extract the intents (first tag the data to extract intents in case of supervised approach and review the intents in case of unsupervised approach).
- Train an ML model to extract relevant front-end data points such as company information, entities, etc., and define how to connect all your dynamic data.
- Train an ML model to match answers to questions and replies.
- Build a user feedback system and integrate it with the chatbot testing stage.
When it comes to transactional chatbot development, you don’t have a lot of data to work with, as you typically build a new process for customers. If you want to power it with AI, you’ll have to find ways to find and collect at least some useful data and insights and follow the steps described above.
“I live in San Francisco, and you should remember that when offering me tickets. If you are really an AI chatbot, you should even remember where I fly most often and have some kind of recommender system based on similar users!” continues Jorge from lang.ai.
In any case, custom chatbot development covers the following stages:
- Data collection and preparation for ML training (supervised or unsupervised);
- Data analysis for intents;
- Data analysis for building answer system (if needed);
- UX/UI design;
- Analysis of relevant entities in questions and answers;
- Chatbot architecture;
- Chatbot coding;
- Chatbot QA and testing (including development and integration of user feedback system);
- Deployment to a target messenger or website.
Most of the chatbot builders offer subscription-based packages which rates depend on the number of subscribers and messages and typically start at $15 per month. In most cases, when you reach and exceed a certain number of free messages, you’re incurred a fee per message. For instance, ChatterOn charges $0.0010 per message once you pass the 15,000 mark, while Pandorabots charges $0.0025 per message for up to 100,000 messages.
When it comes to custom chatbot dev, your cost of development will depend on several factors with the location being the most decisive one. You have four options to build your chatbot as a custom product:
- Build an in-house team of chatbot developers;
- Extend your software team with highly-competent and experienced chatbot developers and testers offshore (in countries like Ukraine or India) or onshore (in your home country) and manage the project on your own;
- Build a chatbot dev team from scratch in your outsourcing provider’s R&D Center and manage it like your remote and standalone development department;
- Outsource your entire chatbot development process to a 3rd party that will build and deliver it as a turnkey managed solution (T&M model).
While the pricing will be different under each model, offshoring options will always be cheaper than in-house or onshore project development because offshore locations offer faster time to hire developers, lower rates, and faster access to AI and ML expertise.
In the USA, for instance, the cost of a simple bot platform that works with a variety of messaging apps can go as high as $41,000 (plus a low-end fee of $500 per month for maintenance depending on backend integrations, scripts, reporting, etc).
An enterprise-level AI chatbot solution built in-house usually starts at $150,000 depending on complexity, including 15–20% on top for annual support and debugging.
Christian Rennella, CTO and Co-founder of elMejorTrato, revealed that “after 9 years of working hard with an internal marketing and sales team to answer our client’s questions through live chat, it was not until 8 months ago that we started developing our own AI chatbot. It cost us $340,000 where the greater part was invested in the salaries of our employees.”
When you leverage offshore resources, the cost will be 2–3x lower. Once again, it all depends on the size of your chatbot dev team (the larger the team — the higher the price), the location of the development, access to global talent and expertise, etc.
To top it all, before making an investment and figuring out your chatbot development approach, you need to weigh all the pros and cons and make sure your chatbot solution will be your “rescue ranger” and not another headache that will result in wasted hours, dollars, and tech debt you’ll have to handle afterward.
So, analyze all of your routine and repetitive tasks which can and should be automated, prioritize tasks and decide on their complexity, choose a DIY platform or hire a custom chatbot development partner to help scope and build your conversational AI solution, allocate the appropriate budget, release, analyze, and improve based on user feedback and behavior. Now when we have so many data analytics tools and software development models and cost-saving options, the sky’s the limit when it comes to building and monetizing highly efficient chatbots.