Automaton and on and on

I've been doing tech support in one form or another for about 15 years. The work has gotten better over that time, largely because the landscape of tools available has grown and evolved dramatically. For the most part, support is more interesting. We have deeper, more complex problems to solve and more useful ways to solve them. We sometimes complain that we're doing the work of engineering or sales or QA, but everyone has more to do, and the expanded scope has made many of our jobs more satisfying. Part of this is a shift in culture and part is the advancement of technology over the past decade.

The past five or ten years have seen a huge spike in automation capabilities. There are so many things we can do now that I could only dream of 15 years ago. A couple of years ago, I worked for a company whose support team would frequently get crash logs from customers. Whenever the app failed in some way, it would generate a crash log and the customer would send that text file to us in a support email. Support staff would open each file and scan it for specific data related to the crash and use that to find the source of the problem (whether it was user error or a bug). That happened a lot, several times a day for each support agent. I built a small app, hosted on Heroku, to take crash files from our help desk, scan them for the data we needed, and post it as an internal comment in the support ticket. Any ticket with a crash log attached would automatically send the file to the app, saving the team a bunch of time and frustration every day. It was grunt work that we no longer had to do. Little things like that add up quickly; they reduce resolution times and ultimately increase customer happiness, which then reduces churn. Everybody wins.

This kind of automation is possible all over the place now. It requires some internal knowledge of the team's product and processes, and it requires an internal developer to build an application or widget or service. But it's possible. And that's amazing.

But we can go so much further.

Where we are now

Now that AI has saturated the support tools market, everyone is obsessed with automating everything, where "everything" means " the things that are easy to automate." Look, I'm glad we can let computers handle password resets and refund requests and product exchanges. That's great. If you're an e-commerce shop, that probably offloads a lot of work that humans don't need to do, at least as a percentage of inbound ticket volume. But is that a large percentage of the time spent by the customer support team? Is automating the easy, known set of requests making a real dent in the amount of grunt work a support professional does each week?

AI (and by "AI" I mean LLMs and generative AI, specifically) has given us some useful tools. But those tools are more often than not only applicable to a predetermined subset of support problems and workflows. Support professionals are still stuck doing a ton of work every day (like opening and scanning crash files) that should be automated, that could be automated. Maybe by AI, maybe by other technology. It's almost like nobody understands the work of support.

Where we should go instead

Truthfully, it's not an "either/or," it's a "yes, and..." We should definitely let computers handle password resets and refund requests. They’re boring, known problems that computers can handle faster and cheaper.

We should also give support professionals the ability to automate the things that suck up their time and emergy every day. But not all teams are bogged down by the same things every day, though, so it's not like the same automations (AI or otherwise) will work for everyone, or be as useful to everyone. Support teams need tools that can be configured to work with their processes, their customers, their products

I'm not suggesting that every team is a unique snowflake. (Note: I'm sure that you, personally, are a unique and special snowflake who contains multitudes, but your support team is probably mad at the same things other support teams are mad at.) I am instead suggesting that all support teams benefit from having flexible, customizable automation capabilities. And not just "hooked up to Salesforce" or "hooked up to HubSpot," either.

Teams need to inject automation in different ways at different points of the customer support cycle. Some have crash logs that need scanning, others have internal CRMs to update. The set of problems we all have are similar, but they show up in different ways and in different places and at different frequencies. We all know our own processes and we know what we don't want to do manually and what we need to do manually. Our tools should help us get organized around those processes.

The fact is, support teams' problems aren't going to be solved by slapping some APIs and generative AI on top of tools that don't address our core needs. Our tech stack needs to be built with modern support organizations and processes at the foundation. We can't continue to use software that piles shiny feature on top of shiny feature while the underlying structure crumbles under the weight of a shifting industry. Teams work differently now than they did 15 years ago, and throwing some AI chatbot on a pile of bad assumptions isn't going to make anyone's job - or life - better.

(If only there were some type of software that understood all of this and gave support teams the power to model and automate their processes and workflows...)