Chatbot Challenges

Is your chat or voice bot upsetting customers?

By ian tomlin | Jun 09, 2021

Is your chat or voice bot upsetting customers? Cover Image

In this article, we discuss the business consequences of using untrained chatbots.

The hidden scourge of untrained bots

Many companies treat their deployed bots as a ‘fire and forget’ project, but that creates thousands of potentially frustrating customer experiences that bosses really don’t have a clue about. Bots have awesome potential to transform customer experience for the better and to generate lots of lovely new sales leads, but leave them on autopilot and not regulate their performance is bad news.

Sometimes, it comes down to resource. Bots can become a time-consuming ‘hobby’ for tech teams that don’t have the time to invest in their training. Equally, when IT teams lack robust corporate sponsorship,

Like humans, bots need coaching and ongoing support

In most companies today, we’ve yet to see command and control systems evolve to support all of the software, chat and voice bots in use across the enterprise.  Were such as system to exist, it’s likely you’d see vast differences in the levels of performance and usefulness of bots, in the same way as humans today.  Like your employees, the performance of bots needs to be continuously monitored and improved.

Learning and development systems for bots

Mirroring the management mechanisms to improve the performance of human workers, bots require a management system to govern their training; to ensure each bot is performing at its optimal level.  Even when using artificial intelligence, it’s a home truth that robots improve their effectiveness over time, and the more practice they have in their roles, the better they become when an effective learning management system is adopted.

This isn’t ‘futures’ technology though: Companies are already adopting bot learning platforms like INTNT.ENGINE from INTNT.AI.

What a bot training system looks like

At this point, you might be imagining rows of bots in a classroom staring at a chalkboard, but thankfully training bots is simpler to organise than that, and somewhat more efficient.  To train a chat or voice bot requires a Software-as-a-Service platform that uploads chat scripts (that have been stripped of personal data to comply with data protection and privacy policies), to then be processed through an AI-powered Natural Language Processing (NLP).  This engine considers the lines of chat questions and answers, quickly determining where failure points exist.  The common types of failure are ‘false positives’–when the bot has given an answer to the enquirer but it’s the wrong one to satisfy the request made–and ‘missed intents‘ when the bot has been unable to offer up an answer because it doesn’t understand the request.

Having processed and interpreted the data, the bot training platform makes recommendations on how best to optimise the answers.  A ‘bot manager’ decides if they are happy with suggested responses and can accept them with a single mouse click.   After decisions have been made on the recommendations served up by the training engine, ‘the system’ will then upload the new chat scripts to the knowledge databank.

Does bot training work?

The simple answer is yes.  Live case stories of projects delivered in 2021 by INTNT.AI demonstrate a 70 reduction in call transfers, a similar 85% cut in missed intents (that boots customer experience) and a further 80% improvement in new business opportunities (in the form of new sales leads).

To learn more about INTNT.AI and to experience the technology and process for yourself, request our short demonstration video or book a meeting with of our consultants.