Last week at a local testing meet-up I help run we had an open chat about AI and the future of testing. One point that really stood out was the time-scales associated with AI ‘taking my job’. While it feels like developers and testers are safe for now, those working in teams such as customer support may not have such a long buffer.
The business case to replace large teams of humans answering questions using pre-compiled information and workflows with a digital service is quite attractive. You remove the overheads of those staff, the management structure and you have an easier means to scale up the number of requests you can deal with.
The immediate impact to the end user would be minimal, if you mostly handle support requests via an online chat then there would be no change to the interface the customer has with your business and if the AI is good enough they may never know that it’s not a real human on the other end.
The long-term impact to the end user and to the business itself is quite large as such a move is akin to chopping an arm off when it comes to the business’ means to sense how customer’s feel and what to change to improve their product.
Customer support are the ‘front-line’ between a business and their customers when they have issues using a product and those working in such positions are fonts of knowledge for user’s actual needs at the time of using a product.
An AI may be able to answer questions and possibly even learn how to create workarounds as it engages with queries over time but unless it’s able to understand complex human interaction patterns it’s going to fail to provide the same level of feedback to the business.
The AI ‘bubble’
Much like the dot-com bubble in the late 90’s I think once AI hits mass adoption for replacing those front-line jobs which on paper feel like a good way to reduce overheads, we’re going to see a downturn in product quality caused by companies removing these important feedback loops and a crash as a result.
It may take a few years for this to happen, after all PR & marketing do a pretty good job of masking product misalignment by replenishing your pool of customers when existing ones get frustrated and leave but as the cost of user acquisition rises due a bad reputation the business will realise it’s not a sustainable business model.
Those businesses insightful enough to realise they traded an important sensing organ for some quick growth to keep the shareholders happy may be able to pivot back to having human teams and regain some control over things but it will take a long time to get those teams giving the same level of quality feedback an existing team could provide so it may be too late.
The counter-argument of course would be that the reliance on AI that got the business into that situation is the answer to the problem.
AI to the rescue
During the meet-up’s conversation there were a few techno-optimists who suggested that AI will become so good at understanding humans that it will be able to provide such a function so there’d be no need to go back to employing humans to do such a job.
I’m not sold on this. If I think about my own interactions with customer support and the thoughts and feelings I had during those messages & calls I can’t see how a data-scientist would even begin to start categorising that dataset into something to train an AI with and I like to think I’m a rational person!
During a customer support call there are so many variables that can influence the outcome. A strong emotional intelligence is needed by customer support agents to navigate the minefield that is upset humans explaining problems to other humans.
While an AI could potentially extract the words used in the exchange with the customer and run some analysis on it to summarise some actionable feedback for the product team to use, it’s not going to be able to consistently do this mid-conversation and use that understanding to drive the conversation towards learning the root cause of the problem.
Even if such an AI were to exist the cost to train it and the processing power required to run those operations would likely put it at a price-point outside of viability for most businesses who opted to replace their customer support teams with AI.
So is AI going to take your job?
It may, but it won’t be because AI can do your job. It will be because those running the company you work for will not understand the value you bring to the company until it’s too late.