Hacker News new | ask | show | jobs
by laborcontract 732 days ago
I was on a small team of 10 that experienced very strong growth in our product. One of the consequences of success is that we had to take hundreds of calls per day. We eventually had to hire dedicated customer support people but after hours were an issue.

There are large outsourced customer service companies but those cost $1 a minute and those people have to deal with CS calls from many different companies, so it took them forever to find scripts relevant to the customer’s request. Most calls failed - that outsourced CS was a glorified voicemail box.

Our success rate with outsourced CS was very low, partly because we didn’t have the resources to fly over to Austin to train a workforce that didn’t show to be particularly promising.

AI voice bots would have been able to helpfully answer and deflect 90% of our calls, and do so in a much faster and humane way.

I understand the sentiment behind your comment, but I cannot agree with your assessment of the outcome.

1 comments

>AI voice bots would have reduced been able to helpfully answer and deflect 90% of our calls, and do so in a much faster and humane way.

Would they have? I've worked at companies that had AI voice bots, they required a ton of care and feeding like a script. Only reason we used them is they were better at following the script and they didn't cause massive customer backlash compared to customer backlash we got when it was offshored.

So many of our calls were informational with a predictable/scriptable answer set. 90% may be high, but I know 80% of the calls would have been squashed by an AI bot.

There's definitely a ton of care that you have to put into creating it aside from just retrieval, like control flow, but the state of our outsourced CS was really, really bad. 5 minute wait times just to get connected to someone that mispronounced our company name and mostly told people to call us in the morning. The costs were absurd for the quality of service we got and they ended up making customers angry.

> 90% may be high, but I know 80% of the calls would have been squashed by an AI bot.

Which one is it 90% or 80%? Is it possible it could've been 70%? 60%? 20%?