Blits Logo

How chatbots outperform human insights

5 min

Key takeaways:

  • Chatbots provide more efficient customer feedback
  • Human filters skew customer feedback
  • People and chatbots working together is the best recipe for success

What a Chatbot does

Chatbots are in essence a gateway between a human and a machine. The visible part is a chat app like WhatsApp or WeChat, the questions asked by a human, and the answers given by the machine. Here’s a Spanish example where someone is asking for their bank account balance after they have been authenticated securely.

(An example of a WhatsApp chatbot built on the Blits Platform)

But what happens behind the scenes is more interesting to watch. Let’s start with the technology, followed by the business insights you can derive from it.

A bit on the technology

Let’s keep this brief. There is a lot going on. The Blits platform connects to various AI (Artificial Intelligence) Engines from Google, Microsoft, IBM, Rasa, Baidu, and more. Every engine is pre-trained with huge datasets and differs in performance. Choosing the right one that fits the context of your use-case and language is crucial.

Typically, Enterprises vary the Engine by use case. For example Microsoft for their English Human Resources bot, Google Dialogflow for their Procurement bot, Baidu for their Chinese customer service bot, etc.

And what’s great is that the Blits Platform provides insights whether you chose the right Engine in the first place.

Did you pick the wrong one? Just switch to another Engine with the flick of a button, and increase your chatbots performance. That’s a feature unique to Blits, by the way. Just imagine, compare that to building a chatbot completely on the IBM Platform and then switching to Google, it will cost an arm and a leg in training, development, process readjustment, time and resources. With Blits, it’s literally one click..

(Switching NLP Engines is as simple as a click)

The feedback loop

Most Enterprises are very curious about their customers, and put lots of effort in gaining insights in a structured way. Customers are asked to fill feedback forms, or are invited in panels to talk about their experience. These events always happen retrospective. As such, observations, even those directly obtained from customers, are often skewed and softened due to the process of obtaining them. Additionally, the process doesn’t catch all customer opinions but just a small subset. It would be unpractical to invite all customers, in most cases 😉

Another method is gathering customer service- and sales reps together in a meeting, and asking what the customers are saying. Again this happens retrospectively. What’s more, the feedback is filtered by the people in between you and the customer you want to learn about, skewed by their memory, their outlook on life, their mood, the relationship with the company they work for, and much more, sometimes leading to socially desirable answers.

Chatbots work differently. Initially you train them. This means you provide answers to questions you foresee. This can be a simple process, we call this a FAQ bot, where the bot provides answers to frequently asked questions. Or a sophisticated one; the bot queries databases to get personalized answers depending on the person, such as an HR bot that answers what the maximum of a salary group is, or a procurement bot that explains delivery options tailored to the product and the customer type.

In all cases, you try to predict the type of questions you get. So far so good.

Then you go live with your chatbot, and an in-tray starts to fill with ‘untrained questions’.

(Untrained questions give insights in customer questions you didn’t anticipate)

Customers have different questions, including topics you did not anticipate. Rejoice, this is when you start to learn, fast.

No filters, just direct questions to your organization that need an answer. And that will get an answer repeatedly, for two reasons. Firstly, after providing the answer once, the system has learned and will answer every next person asking the question. Secondly, the AI Engines will make sure you do not have to ask in 10 different ways, they get it that ‘What can I do with your Platform’ means more or less the same as ‘What are your Platform’s benefits’, no need to type that twice..

Compare that to a human; they will have to either type it repeatedly, click on a canned response in your CRM system time and again, or provide spontaneous answers which vary from human to human.

So, do bots outperform humans at every turn?

No! This is why we have built in a life chat integration with tools like Salesforce. Not just can the bot handover to a human, vice versa is also possible; Imagine a scenario in which a customer wants to book a hotel. The bot handles all routine things well, until the customer starts to get angry as the bot asks for their date of birth, something the customer doesn’t want to give due to privacy concerns. Our ‘sentiment analysis’ function kicks in, detects the mood of the customer getting bad and offers the option to speak to a human representative (rep). If the customer accepts, a rep takes over the conversation. A pleasant conversation follows, the customer calms down and the rep asks if the customer wants to continue booking. The customer confirms, and is asked by the rep if the bot can finish the process, as it will be faster. The customer agrees, and the rep hands back to the bot to complete the booking.

Humans remain crucial in this process, and excel at relating to other humans. But when it comes to handling routine repeated questions, and learning from them to continually increase customer satisfaction, chatbots are the better choice.

Most success is obtained where humans and chatbots work together to serve customers. When the going gets tough, the human takes over. When routine kicks in, the bot takes over.


I’d be interested to learn from you perspective. If you have thoughts you’d like to share, please leave a reaction below, or contact me on this page.


* When we speak of AI (Artificial Intelligence) Engines, we specifically mean the field of Natural Language Processing (NLP). Examples are Microsoft LUIS, Google Dialogflow, IBM Watson, Baidu, and Rasa. There are more Engines, and Blits’ commitment to the market is to continue including additional high quality NLP Engines, to ensure our Platform supports every possible chat- & voicebot business case.

Our Company

© 2023 Blits BV

Warmoesstraat 151
1012 JC Amsterdam
The Netherlands

Blits Conversational AI Ecosystem

Blits is the low-code Conversational AI Ecosystem which combines the AI power of Google, Microsoft, OpenAI, IBM, Rasa, Wit, Amazon, Stanford, Nuance and more in one platform.

Use Blits to build, train, deploy and benchmark chatbots & voicebots at scale, for any type of use-case.

Focus on building a bot with the perfect tone of voice for your audience, and optimize the underlaying AI Technology later.

Always stay ahead of the competition with 'Blits Automate', giving your bots the latest combination of AI tech that fits your use-case automatically.

Reuse templates between bots, creating multi language/country/brand interactive communication on your existing channels (WhatsApp, Slack, Alexa, Twilio, Web, etc.)

Connect backends to build smart bots (Automation Anywhere, Salesforce, SAP, ServiceNow, UIPath, etc).