Chatbots can make interactions easier for both customers and businesses—if they’re carefully developed. If the bot can’t answer common questions or guide a customer through all the steps needed to resolve their issue, it will lead to frustration and added stress for both your service team and the customer.
It takes a blend of human understanding and leading technology to create a chatbot that will successfully serve customers and become a valuable, time-saving tool for your team. Below, 20 members of Forbes Technology Council share their tips to help companies launching or refining a chatbot ensure it does the job as intended.
1. Assess The Client Benefit
Will your clients or customers use the chatbot as you intend? When investing time and resources in creating this solution, a company needs to assess how the project will benefit the client enough for them to want to use it rather than reaching out via traditional methods. Even with all the data and analytics, it will remain dormant if it’s not adopted. – Robert Giannini, GiaSpace Inc.
2. Define The Scope Of The Dataset
A common mistake in chatbot development is not defining the scope of the dataset. If it’s too small, responses may become repetitive and lack depth. Conversely, an overly extensive dataset can yield inaccurate and inconsistent responses. Continuous testing and refinement based on feedback is key to optimizing the dataset’s scope. Managing the dataset closely helps ensure high-quality responses. – Megan Kelley, Fidelity Investments
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
3. Invest In NLU Technologies
A common mistake in chatbot development is neglecting natural language understanding, which leads to poor user interactions. Organizations should invest in advanced NLU technologies and train their chatbots with diverse, industry-specific datasets to improve understanding and contextually relevant responses. Regularly updating the bot based on user feedback can also enhance its effectiveness. – Saif Sultan, Volos Portfolio Solutions, Inc.
4. Define The Problem Scope
Leveraging large language models to create robust chatbots has significantly improved the customer experience. However, a potential for misinformation can arise from a chatbot’s ability to understand complex inquiries combined with an inability to solve them, which may require human action. It is crucial to define the problem scope that chatbots can effectively handle to ensure a seamless customer experience. – Narjès Boufaden, Keatext
5. Leverage Customer Conversation Data
When creating chatbots, leveraging customer conversation data is key to avoiding robotic experiences and high frustration rates. Advanced conversation intelligence technology can help bots learn and personalize, allowing for better, more tailored user interactions. The goal is to provide users with a tailored experience with a chatbot that reflects personal preferences and past conversations. – Robert Beasley, Tethr
6. Enable Added Self-Service Abilities
Organizations tend to enable the same functionality in chatbots that is already available via other self-service channels (such as websites, apps and so on). People only want to talk to someone when they are unable to use self-service options. Thinking through what is already available through other self-service options and enabling additional services through the chatbot is critical. – Ahmed Zaidi, Accelirate
7. Address The Knowledge Powering The Chatbot
While creating a chatbot is fairly quick and easy, making the chatbot useful and scalable requires a lot of input, especially when it comes to the knowledge the chatbot has access to, how it is updated and more. If organizations don’t take the time to address the knowledge powering a chatbot, they risk having a chatbot that lacks accuracy, performance, security and more. – Danny Fields, Avalara
8. Ensure It’s Channel-Agnostic
Don’t develop it all yourself. In today’s tech era, no company should be an island. Whatever your platform, whether it’s in-house or third-party, it’s key to ensure your bot is reusable. That means being able to tie different communication flows into more channels than Web chat alone. A great chatbot should be channel-agnostic so you can leverage it across platforms such as WhatsApp or WeChat. – Matteo Gatta, BICS
9. Give The Chatbot Agency To Meet Customers’ Needs
Customers are either looking for generic information, personalized or specific information, or to resolve actionable or transactional requests. Many organizations launch a chatbot with no agency to really solve the majority of customers’ needs. If most customers contact support to reverse a charge or cancel a subscription, and the chatbot won’t have the authority to perform such actions, rethink your strategy. – Lucas Persona, CI&T
10. Leverage Full AI Capabilities
One mistake is confusing decision-tree based chatbots with “AI.” Decision-tree based bots allow some control over what the bot says, what branches to take and which scenarios to handle. However, this isn’t much different than interactive voice responses on phones: “Press 1 for billing; press 2 for account changes.” When deploying a chatbot, leveraging full AI capabilities—ideally via a full-service AI suite—will yield the best results. – Deon Nicholas, Forethought
11. Think Through The Questions Customers Will Ask
When building a chatbot, organizations often make the mistake of not researching and brainstorming the possible queries of customers. Instead, they only add their own ideas to the list. All possible queries should be researched and addressed, and then the chatbot must be released only after final testing. – Praveen Andapalli, Vitel Global Communications, LLC
12. Try To Anticipate Customers’ Wants
Companies often build a chatbot to save costs, and in doing so, they’re gambling with their most important asset: their customers. Think “customer first” and implement a chatbot that truly anticipates what a customer wants. Leveraging modern technology that uses natural language processing and empathic AI to react to customer sentiment helps maintain trust, and that trust is the foundation of your customer relationship. – Derek White, Galileo Financial Technologies
13. Extend The Chatbot Across All Customer Touch Points
To prevent inconsistent experiences, companies should adopt an omnichannel approach, utilizing chatbots across all platforms where interactions happen. This requires understanding user preferences, using integration tools for easy deployment, ensuring consistent responses, optimizing with feedback, and maintaining security and compliance. – Jon Jessup, 1440
14. Go Beyond Merely Answering Questions
One common mistake brands make is developing chatbots that are all talk and no action. To make them cost-efficient, chatbots must go beyond merely answering questions—typically information already found elsewhere—and be able to handle real, complex actions like a human agent. – Jeff Fettes, Laivly Inc.
15. Offer An Easy Opt-Out
One key mistake is not offering an easy opt-out from chatbot interactions. Companies should ensure their chatbots are user-friendly but also unobtrusive, allowing users who prefer navigating the site on their own to easily bypass the bot. Not all users are the same, and not all users see the value of chatbots. – John Kuhn, Integral
16. Carefully Establish A Search Foundation, Knowledge Access And Guardrails
AI-powered tools such as chatbots have the power to dramatically increase productivity by automating everyday tasks, but having the right search foundation, access to your company’s collective knowledge and specific guardrails is crucial to getting accurate, useful output. When these ingredients are at the forefront of chatbot creation, you protect sensitive information and increase efficiency. – Arvind Jain, Glean
17. Protect Your LLM Against Attack
The explosion of generative AI apps has created a new threat to sensitive enterprise data that didn’t exist a year ago. This created a new data loss vector almost overnight, as AI apps (including chatbots) can inadvertently process confidential data, and AI models can leak sensitive data. Enterprises worldwide must protect their large language models against attack and ensure employees practice the safe use of these powerful tools without putting sensitive data at risk. – Anand Oswal, Palo Alto Networks
18. Avoid Misinterpreting Extra ‘Clicks’
Customers shop online and will study your product (and your competitors’ products) carefully before buying. Support that journey. Don’t let a myopic focus on response rates to calls to action derail a positive customer experience. If your bot interrupts the customer journey and generates “clicks” in your data, be careful not to misinterpret that frustrating experience as positive. – Garrett Silver, Critical Insight
19. Maintain A User-Centric Focus
Instead of blindly following trends or competitors, ask yourself, “Does this chatbot really solve my customers’ problems?” Approach chatbot creation like product development—with a user-centric focus—and launch the new solution only when there’s a clear need. As a result, you’ll create a chatbot that users love, and your business will reap the rewards of increased satisfaction, loyalty and ROI. – Leon Ginsburg, Sphere
20. Make The Experience Engaging And Fun
Don’t be too transactional. Chatbots are a brilliant way of galvanizing your brand’s fans into a community that can provide you with rich insights, marketing feedback and data. With AI, chatbots can supercharge your marketing engagement across platforms. Consider using an avatar, adding AI video content and gamifying the experience to ensure people return for fun, content and connection. – Catherine D Henry, EPAM
