I’m ready for AI be my customer support representative

Can we pause for a moment on developing more AI learning models that can write blog post, SEO and social post copy and focus on LMs that can help a user with a technical support question?

Alan Jones
3 min readApr 27, 2023
Image by Dall•e

In the past few days I’ve used a few SAAS platforms that have thrown up a cryptic error message and left me blocked from moving forward with my task.

There is never a layperson’s explanation of what the error message means, why it has happened, or what to do next.

Have you checked the FAQs?

Yes, in the faint hope that one day, a FAQ might include an answer to a question that I actually have. Every day, that hope gets fainter.

Inevitably, a SAAS’s FAQs are useless once you become an actual user of the product, containing only answers to those obvious user pre-onboarding and onboarding questions about pricing, features and free trials that the product management team were able to think of, back when the first public beta was going live.

Please call us between 9am-5pm Monday to Friday

I don’t want to wait 20mins on hold to speak to an exhausted, inexperienced customer support representative, only to have them read from a script, while they walk me through all the obvious things I’ve already tried that have failed to fix my problem.

Our friendly chat bot is here to help you

Nor do I want to speak to a dumb chatbot that only knows what’s in the FAQs I’ve already read, can’t actually understand my question, and seems mainly designed to delay and discourage me from dialling the support line.

What I want, what I really, really want

I want a Learning Model (LM) that learns how to help me and other customers from studying the actual questions and problems that users are facing right now, today, in their use of the product, as well as the way tech support representatives communicate solutions to those problems.

This seems possible right now, if the LM were pointed at tech support representative chat and call transcripts about how to help customers and told to learn. Learn not just the platform jargon that software engineers and tech support reps use, but from whether/how users comprehend the help they’re given.

There are challenges, sure. We can’t allow the LM to hallucinate imaginary causes of (or solutions to) customer’s problems. We have to train the LM to adhere to the brand voice while still being efficient and more helpful than a human support rep. And obviously we can’t have the LM telling you that it loves you and wants you to divorce your wife. But these are fixable!

Does this already exist?

In a fast-changing sector with a massive amount of hype and not a lot of facts, there may well already be products which do this. Please leave a comment with anything you’re using that you can vouch for, that you’d like me to try out and see if it’s something I can recommend to the early-stage startups I work with.

You see, I often get asked to help an early-stage startup with improving their churn (the number of customers who stop using the startup’s product) and my first area of investigation is to dig into the customer support experience being offered. Early-stage SAAS startups often under-invest in great customer support solutions because they’re focused on polishing a product and acquiring more users, figuring that customer support is a second-order problem that can be dealt with later. But great customer support is the best customer retention tool in your tool box, and if you’re experiencing high churn, it’s often a leading cause.

--

--

Alan Jones
Alan Jones

Written by Alan Jones

I’m a coach for founders, partner at M8 Ventures, angel investor. Earlier: founder, early Yahoo product manager, tech reporter. Latest: disrupt.radio

Responses (2)