3 Mistakes You're Making With Customer Support Chatbots (Every Company Is)

October, 31 2018

Subscribe and stay up to date

No spam, we promise! You will only 
receive essential emails.
Written By
Rob May

Every time technology shifts, people are caught off guard and there is this phase in the market where the new technologies are being thought of using the old frameworks. That’s where we are with AI today. Buyers are looking at enterprise AI software like it’s one of the SaaS solutions they bought in 2014. It’s not.

At Talla, we have calls on a daily basis with people deploying bots for customer support. The bots aren’t working and they wonder why. I’ve seen the same problems crop up again and again and again, so I want to share the mistakes that companies are making and how you can avoid them.


Mistake #1: Not Understanding That Bots Need To Be Trained

In a world of SaaS software, you deployed your software and started using it. On day 1 it was just as good as it was on day 1001. AI software, including true AI chatbots that learn (most don’t), needs to be trained. The software works more like a person than a program.

The good thing about it is that it gets better over time. On day 1001 it will be 100x more powerful than on day 1. The bad thing is that on day 1, it will need a lot of hand holding.

When you deploy a bot for customer support, make sure to bake in training time. Our best customers at Talla typically start with a 30 day roll out to a subset of support reps, before opening up the tool to the broader group.


Mistake #2: Not Understanding That Content Needs To Be Trained For Bots

In more organizations today, someone in product, marketing, or a technical writer of sorts writes product content for sales and support materials. This person is not usually connected to the day to day questions being asked by customers. They tend to write content like it’s a user guide or research paper or something similar, and as a result, there is a very large gap between the way the content is written and the way customers would like to digest it.

AI is great at translating this gap. What we do at Talla is to add special structure to your content as you create it so that we can provide extra information to the NLP models we build. This means we don’t have to use a generic language tool like Watson or Lex, but instead can train our bots on the particular vernacular of your business.

Think about it this way — you wouldn’t create a tradeshow handout in plain text would you? No. You would add some formatting so that it looks good to the prospects who walk by your booth. Talla allows you to add some formatting so that your content looks good to the machine learning models that process language. This is why we outperform all the other bots on the market.

If you are going to use a chatbot solution, take the time to format your content in a way that is best consumable to the bot, and thus to the customers or support reps who use the bot. If you want to “train your content” based on what customers are really asking about, only Talla has that capability. If you aren’t going to use Talla, at least assign a person to look through it and put it in a format more consumable to bots and to customers.


New call-to-action


Mistake #3: Favoring Point Solutions Over Integrated Products

Success in AI is all about the data. The more data you have, the better the AI can perform. The corollary then is that the more integrated AI is into various steps of your process, the more data you will have and thus the better the outcome.

Most buyers are stuck in a SaaS world. SaaS was all about highly targeted point solutions that could be integrated with APIs if necessary. AI tools work differently. If you pick a knowledge base from one vendor, a NLP solution from another vendor, a bot from a third vendor, and a human-in-the-loop training tool from another vendor, you get a complicated setup with subpar AI results. The most you can look at integrated solutions, the more refined the final output will be. This is why Talla built a new kind of knowledge base that merges content management with automation, machine learning, and chatbots. Our integrated solution delivers a better result.

Every business is under press to deliver more with less, and if you run a Support team, your boss probably gave you conflicting goals of providing a better experience while simultaneously lowering costs and increasing productivity. Until AI came along, that was impossible. With the right AI tools, you can achieve all of those goals together.

Be careful as you evaluate solutions not to make the same mistakes that so many others are making. Choose the right tool for your organization, and understand how AI is different than what came before it. Build a deployment plan that includes a training period, and make sure your content is AI-ready. That will be the difference between success and failure with AI deployments.

And of course, if you want to try Talla, please schedule a demo with us. We know your problems well, and we can help.


New call-to-action

View all posts

Subscribe and stay up to date

No spam, we promise! You will only 
receive essential emails.

Subscribe and stay up to date