Ada Support

Questions to ask before you onboard a conversational AI platform for customer service

Sarah Fox
Content Producer

Finding a conversational AI platform for customer service can be stressful. Unless you know exactly what you’re looking for, it’s tough to find the best conversational AI platform for your business.

In this guide, we discuss seven key questions you should ask the provider when looking for the right conversational AI that will transform your customer service. We’ll show you what a good answer to these questions looks like, so you can identify the top providers from the average ones. Let’s jump right in.

What AI technology does your platform use?

If you want the best solution for your business, start by looking for a conversational AI platform that offers a powerful combination of cutting-edge technologies. Some of these technologies have a long way to go, but adopting them now has benefits. As Ross Gruezemacher, Assistant Professor of Business Analytics at the School of Business at Wichita State University, explains in his HBR article :

"Aggressively adopt new language-based AI technologies; some will work well and others will not, but your employees will be quicker to adjust when you move on to the next. And don’t forget to adopt these technologies yourself — this is the best way for you to start to understand their future roles in your organization."

Here’s the bare minimum you should look for in a modern conversational AI platform.

Machine Learning (ML) and Natural Language Processing (NLP)

ML and NLP are the core technologies powering conversational AI platforms. All the other conversational AI technologies (like sentiment analysis) depend on ML and NLP. A platform that doesn’t use ML or NLP is just an scripted chatbot responding to basic FAQ inquiries already found in your KB, not a generative AI agent.

  • Machine learning: ML is like a muscle that flexes with experience. The more data it ingests, the smarter it gets. For example, an ML algorithm can monitor conversations to find a pattern and “learn” the best answer for a given query.
  • Natural language processing: NLP enables conversational AI to understand human language. Suppose a chatbot can’t offer a solution to a customer’s query and asks the customer if they’d like a callback from a support agent. The customer responds, “Wow, that’s just what I needed! Another callback!” A scripted chatbot would derive the literal meaning and respond, “No problem! Is there anything else I can help you with?” Embarrassing, right?

Speech Recognition

Speech recognition helps a computer understand human speech. Speech recognition systems involve multiple components, including acoustic modeling, language modeling, and speech decoding to accurately convert and interpret speech.

Once the system receives input from a human, it breaks the speech down into basic components called phonemes. The system processes phonemes using a complex series of algorithms to interpret the spoken words.

A speech recognition system must overcome multiple challenges to accurately interpret speech — accents, background noise, and mumbling are common culprits. But thanks to machine learning, systems get better over time.

Which out-of-the-box integrations are available?

Data is a key driver of customer experience. But data gets trapped in silos when your tools aren’t integrated.

An AI agent needs access to multiple data sources to offer a seamless and personalized experience. You can always custom-build integrations, but that’s time- and resource-intensive. Ready-to-use integrations are a lot more cost-effective.

Before you ask this question, prepare a list of integrations you need. Add all the tools you use daily — your CRM, ERP, and invoicing tool — to this list. Ask the provider if they offer an out-of-the-box integration for these platforms or if they have an open API so you can build an integration yourself.

If an API isn’t on the table, ask the provider if they can cook something up for you. If they’re not keen, swipe left and consider another provider.

Does the platform offer analytics and reporting?

The platform’s analytics should be as live and dynamic as the conversations it facilitates. Analytics and reporting offer value in three ways:

Measure performance

The analytics feature collects data and uses it to measure performance. You can measure and monitor customer experience KPIs like customer satisfaction score (CSAT) , containment rate, and Automatic Resolution Rate (ARR%) in real time right from your analytics dashboard.

For example, you could check your ARR when there’s a sharp increase in support tickets. A low ARR indicates that the AI agent is unable to offer solutions customers are looking for.

Make predictions

Predictive analytics can make predictions based on historical data. AI agents use predictive analytics to predict user intent, recommend next-best actions, and drive conversations toward conversion.

“Those with an eye toward the future are boosting their data and analytics capabilities and harnessing predictive insights to connect more closely with their customers, anticipate behaviors, and identify CX issues and opportunities in real-time.”

- McKinsey

Once the ML algorithm makes predictions, generative AI uses them to generate engaging responses that provide an immersive experience. For example, when a customer contacts a travel booking AI agent and asks, “I’m looking for flights from New York to London on March 20,” the AI agent recognizes that customers generally follow up with questions about prices for the economy and business class. The AI agent proactively responds, “We have 11 flights from New York to London on March 20. Would you like to see the prices for economy or business class?”

Generate real-time reports that impress the C-suite

Reporting helps you get a quick performance overview, but preparing reports can be a massive headache. Crunching numbers is one thing, but you also need to invest time into making reports look aesthetically appealing — let’s face it, nobody wants to read through hundreds of rows of numbers.

A built-in reporting feature substitutes manual effort by auto-generating visually stunning and informative reports that helps you and your team make a great impression on C-suite executives. We’re talking sleek dashboards with charts and pie graphs. When you need to look at your customer support KPIs, just head over to the reporting and dashboard, and voilà!

How customizable is the platform?

Imagine a world where all AI agents look and feel the same and generate the same answers. Boring, right? That’s why customization is non-negotiable — it enables you to add your own flavor to the AI agent. Here are some customization options you should look for:

  • Visual customization: Visual customization options allow you to customize the bot’s branding to make it blend into your website. Branding customization options can include the bot’s name (like Bank of America’s Erica ), colors, layout, fonts, language, and the fallback message that appears when there’s a technical issue with chat.
  • Personality: Is your brand’s personality friendly, plainspoken, playful, or sophisticated? Top conversational AI platforms let you choose a personality, emojis that can be used during conversations, and a speaking voice.
  • Language: Your AI needs to speak your customer’s language. In fact, 71% of respondents to a survey said it’s extremely important for a brand to promote and support their products and services in their native language. Check if the platform offers multilingual support. Look for languages native to regions where you have a large group of customers.
  • Analytics and reporting customization: One size doesn’t fit all when it comes to analytics. Look for a platform that offers customizable analytics and reporting dashboards that allow you to monitor metrics that matter to your business. From CSAT scores to ROI, you should be able to slice and dice the data to extract valuable insights.
  • Accessibility: Can you customize the conversational AI platform to make it more accessible? See if it offers features like screen reader compatibility, keyboard navigation, and voice commands to ensure accessibility.

Some providers offer even more customization options. For instance, Ada lets you customize your AI agent’s generative content with guidance . Guidance helps Ada answer questions based on your preferences. For example, you can guide the AI agent on how to respond when the customer’s text includes or begins with specific text variables.

What security measures do you use to protect customer data?

Conversational AI platforms collect and generate data that can supercharge your business . However, this makes them an attractive target for cybercriminals. Security breaches can lead to legal troubles and, more importantly, a damaged reputation. This is why security is mission-critical.

Here’s what you should look for to ensure your data remains secure:

  • Data security and privacy: Your data should be locked up in a secure digital fortress. The conversational AI platform needs a secure encryption protocol and cipher to protect your data — TLS 1.2 and AES-256 are the current industry standards. Also, ask if the platform retains any data after you terminate your contract with them.
  • Compliance and certification: Ask the sales rep for information on third-party audits and compliance with privacy regulations like GDPR.
  • Infrastructure and network security: Is the platform’s infrastructure hosted on secure servers and does the team have full control over the infrastructure? Has the platform undergone penetration testing? Ask the sales rep about how they monitor suspicious activity on their infrastructure — do they use intrusion detection systems (IDS)?
  • Business continuity: Ask if the platform has made provisions to continue services even in case of deployment errors. In the event of data corruption or loss, the platform must have a backup to prevent data loss.

How does the platform generate safe and accurate responses?

An AI agent must interact respectfully with the customer and shouldn’t engage in dangerous or harmful topics. Responses should be safe and accurate. Conversational AI platforms that use publicly available LLMs may generate incorrect or offensive answers because they contain information from all over the internet.

Not only is this bad news for your reputation, it can have legal and financial repercussions for the company. Using first-party support documentation and internal data minimizes the risk of misinformation or hallucination.

Ada was built with a focus on safety and accuracy to optimize for Automated Resolution . At Ada, we use multiple methods to ensure that the responses AI Agents generate are retrieved from the documents found in your knowledge base.

When speaking to a sales rep, ask how their conversational AI platform tackles hallucinations and knowledge base management.

How have previous clients benefited from the platform?

Fancy features are nice, but they don’t mean much without results. Ask the provider how their solution has benefited previous clients before you put money on the table. Top providers publish case studies to showcase how exactly other clients used the platform to achieve the results you expect to achieve from the platform and have multiple testimonials from happy clients. Ask the provider to share these with you.

Here’s a general overview of things to look for when the provider offers case studies or other examples of successful implementation:

  • Testimonials: Testimonials tell you exactly what previous clients loved about the platform. For example, if they’re raving about a specific feature or the support team’s proactiveness, those are your green flags.
  • Tangible results: Look for case studies that talk about tangible results and go beyond vague success stories. Did the platform boost CSAT by 50%, reduce support costs by 30%, or cut response times in half? Numbers don’t lie.
  • Innovative solutions: Has the provider helped clients solve challenges through innovative solutions? Ask for examples. While your challenges may be different, these examples tell you if the provider is willing to go above and beyond to help you succeed through their platform.
  • Before and after scenarios: Compare the before and after scenarios when looking at a case study. If you’re in the same industry, you’ll know if the platform truly made a difference or if it was just a shiny distraction.
  • Long-term success: Successful implementation isn’t a quick, one-and-done task. It’s a long-lasting transformation. Look for signs of sustained success when going through case studies, not quick fixes.

Time for a demo

Once you find a conversational AI platform that checks all the boxes, request a demo . Be sure to make notes of things you like and questions you have during the demo. The more questions you ask, the more confident you’ll feel about investing in the platform.

How to interview an AI agent

Looking to better understand the difference between chatbots and AI Agents? Download this guide to become an expert on the topic, and discover the success criteria you should be testing to get the most ROI.

Get the guide