Select the right conversational AI platform for your business.

In our 15-minute advisory call, we'll share our Conversational AI platform selection methodology with you. Our method focussed on your time, budget and feature requirements to identify the top platforms suiting your needs.

There are hundreds of conversational AI platforms out there, which can make the selection process overwhelming. Quickly narrow down the list of potential vendors by identifying your business requirements for platform features, innovation capabilities, product roadmap, cost, and implementation time.

Then take the time to thoroughly investigate your preferred 2-3 platforms. Selecting the right bot building platform is the most important strategic AI decision you'll make, because switching platforms later on is challenging and costly (we know from experience). We're platform agnostic and can guide you through this essential decision.

These are the platforms that we work with regularly:

• Sinch
• Cognigy
• Kore.ai
• Amelia
• Ebbot
• Dialogflow
• Copilot

If you have questions about another platform, book a call! We work with more than the ones listed above.

To ensure success, follow these best practices:

1. Focus on the User: Design your AI with the user at the center. Anticipate a wide range of inputs and ensure your bot can handle unexpected interactions.
2. Comprehensive Planning: Approach your project with a robust plan that considers all potential user behaviors. This affects project management, scoping, building, and testing phases.
3. Expect the Unexpected: Users can say the darndest things. Your bot should be equipped to manage a variety of inputs, ensuring it remains functional and helpful no matter what.
4. Iterative Testing: Continuously test and refine your bot. Use real user data to identify and address gaps in your bot's understanding and performance.
5. Diverse Scenarios: During testing, simulate diverse scenarios to train your bot to handle both common and uncommon interactions effectively.
6. Out of Scope Matters: Out-of-scope interactions are just as important as in-scope ones. You don't control what users will type, so while building your happy flows, ensure your non-happy flows are well-covered too.
7. Start with the Basics: Begin with a scope that includes frequently asked questions which are easy to answer and easy to understand.

By focusing on these best practices, you can create a chatbot or voicebot that delivers a seamless and effective user experience.

Unrecognized user expressions are unavoidable, but we do aim to minimize the dreaded "I'm sorry, I don't understand, can you please rephrase?" response.

Here’s our approach:

1. Remove Rephrasing Requests: Rephrasing a question rarely helps if the NLP model is set up correctly. We suggest removing this request altogether.
2. User-Friendly Responses: In fact, rewrite the entire "not understood" answer to be more user-friendly, always providing helpful options to guide the user.
3. Entity Fallback: Sometimes the user expression isn't understood, but we may recognize a keyword. We can then answer with: "It sounds like you have a question about your card. Can one of these options help?" This redirects the conversation effectively.
4. Gibberish Detection: Detect and discard interactions where the user typed gibberish, keeping the conversation relevant and focused.
5. Generative AI Rephrase: Use Generative AI to extract the essence of complex sentences or multiple intents in a single expression, then send this refined result back into the NLP model for better processing.
6. Generative AI Knowledge Base: Utilize Generative AI trained on your knowledge base to answer complex or unexpected questions, ensuring the user gets a helpful response.
7. Provide another option: If the above options fail, provide the user with a next step. This might include directing them to your website, offloading to a live agent, or offering an alternative means of communication.

By employing these strategies, we can improve user interactions and keep conversations on track, even when faced with unrecognized expressions.

AI is a double-edged sword. While it empowers users with more control, it also relinquishes some control from the interface side. this brings with it certain risks. No one wants to be the next Chevrolet or Air Canada incident.

That said, we’ve built secure bots for multiple banks, insurance companies, and telecom providers, including those handling user login and transactional flows. We ensure the AI remains within safe boundaries regarding what it can discuss and how it communicates. Given that a successful chatbot can have thousands of interactions a day, it's crucial to build it right.

Here’s how we ensure safety:

1. Platform Security: We only work with platforms that adhere to robust security measures, including data encryption, secure access controls, and compliance with industry standards such as GDPR, CCPA and ISO certifications.
2. Traditional IT Safety Measures: Beyond AI-specific controls, we ensure traditional IT security practices are in place, including secure data handling, regular security audits, and stringent access controls.
3. Multi-Iteration Testing: Before go-live, we conduct thorough multi-iteration testing, including testing from our side, client-side, and intense monitoring during the first days of launch. This ensures AI quality is kept at a high standard.
4. Controlled Prompt Design: When using Generative AI, we carefully design prompts to limit AI responses to safe and relevant topics, maintaining a professional and controlled interaction.
5. Controlled Input: For Generative AI, we use NLP to limit which questions we send to Gen AI for processing, making sure we control the input.
6. Internal Safety Checks and Controls: We implement multiple failsafes and internal checks to monitor and manage AI behavior, ensuring it stays within predetermined guidelines.

By employing these best practices, we help our clients use conversational AI safely, ensuring robust security and control while delivering powerful user interactions.

We sure do! Our extensive auditing methodology is explained in our Best Bots of Belgium 2023 Report. If you have a bot that you want us to take a look at, make sure to include the link, phone number or shortcode in the Notes when you book a 15 min. call with us.

We’ve worked on virtually every channel available to ensure we meet users wherever they are. Our experience includes:

1. Webchat (website widget)
2. Integrated in-app chat
3. Facebook Messenger
4. WhatsApp
5. Instagram
6. SMS / MMS
7. Slack
8. Teams
9. Telephone (Voice)

By leveraging these diverse channels, we ensure your chatbot or voicebot can provide seamless interactions across multiple platforms, enhancing user engagement and accessibility.

‍

We sure do! We typically recommend starting with an intent-based bot and then adding Generative AI at a later stage. The two are compatible and serve different purposes. Intent-based bots are ideal for frequently asked questions where we want to control the flow completely. Generative AI, on the other hand, is perfect for handling the long tail of user questions—those that aren't asked often individually but collectively represent a high volume. This combination ensures comprehensive coverage and enhances the overall user experience.