Google chat support shows how intelligence and stupidity can coexist

I used Google’s chat support. The result was infuriating, because the support person didn’t listen. Humans doing support need to listen and use context — and you’d think algorithms could help with that, especially at Google.

Here’s what’s happening. I just set up a new web site for my nonprofit, The next step is set up emails for this domain on Google’s G Suite.

Unfortunately, it appears a previous contractor already did that and didn’t share the password with us. So now I need to gain control of the Google account.

This means talking to a human at Google, because all the automatic options require knowledge of the passwords or backup email that the previous contractor set up. I contacted Google support through my other G Suite domain,, to try to get help. I waited with 30 other people in a queue for 20 minutes until a live support person came on the chat. (They offer phone support, too, but that just plays music for 30 minutes after telling you “a support person will be with you shortly.”)

Notice in the transcript below the number of times that the chat support person, Kai, repeats information that they should already have, or that is not helpful. I’m showing those in bold. I’ll also highlight in italic the comments that are intended to reassure but instead just make the customer angry.

Kai: Thanks for contacting Google Cloud Support. My name is  Kai, and I’ll be working with you today. I’m reading over your message right now. Was there anything else you’d like to add?
Kai: Hi Josh, thank you for contacting G Suite support
Kai: I apologize for the long wait on queue
Josh Bernoff: Yes. I need help getting into this domain for email:
Josh Bernoff: My domain is working fine. But I am trying to set up that other one, and the employee who set it up has left the company so we are locked out.
Kai: Just to make sure I got that correctly, you need assistance accessing

Repeating back what you just read is supposed to be reassuring, but in this case it’s like talking to a 2-year old. Why not say “OK, let’s get you the help you need to access”

Josh Bernoff: yes.
Josh Bernoff: Can you help with that?
Kai: I see, no worries, I will do my best to help you with your concern today
Josh Bernoff: The admin account email is [email redacted from this transcript] and we do not have the password or past passwords. I do have access to the domain CNAMEs on GoDaddy if you want to verify that way.
Kai: Got that noted
Josh Bernoff: So how do I get in?
Kai: Have you tried doing a password reset?

I would think the answer to this is obviously yes, but I guess Kai has to be sure.

Josh Bernoff: It asks for an old password and I don’t have. I also don’t have access to the backup email because the employee has left the company.
Josh Bernoff: Can we verify a different way?
Kai: I completely understand Josh
Kai: And there’s no way to get in touch or reach out to the previous admin or the one who had the account registered?
Josh Bernoff: That is correct. I don’t like to repeat myself, I said that already.
Josh Bernoff: Can we verify by way of the CNAME record?

Yes, I’m beginning to lose my cool. I know exactly what I want to do, which I am sharing with Kai — that is, to verify ownership of the domain through the records at my domain registrar, using what’s called the CNAME record. I know that’s possible, and Google will eventually open that option up to me. But Kai insists on repeating what’s in their script, including paying no attention to the information I shared previously — that I don’t have access to the previous passwords. (If I did, why would even be talking with them?)

Kai: Alright, my apologies, just wanted to make sure I have everything noted
Kai: Reviewing your concern now
Kai: Thank you for patiently waiting Josh
Kai: So I have here a form that you can fill out, your case needs attention by a specific team that will further help you gaining access to your other domain
Kai: Please click on this link – [link to support form]
Kai: Once you have submitted the form, it will be sent directly to our account recovery team

I originally contacted the chat because the form — which I’d already seen online — said it would take 72 hours to respond. But now I see that talking to a human just redirects me to the form after the human fails to listen and wastes my time. Is Google sending me the message that chat support is supposed to be annoying so I should just communicate with its Web site instead?

Josh Bernoff: How long to get support after I fill out the form?
Kai: They will reach out to you as soon as possible
Josh Bernoff: 2 hours? 2 years?
Kai: It will be within the next 24 hours

Never say “as soon as possible” to a customer. It has no meaning. If you know it’s 24 hours, say so.

Josh Bernoff: thank you.
Kai: Don’t worry, I’ll have your case forwarded to them as well with the highest priority. That way, you have the form and this case on their end.
Josh Bernoff: Thank you.
Kai: No problem at all
Kai: Before we end the chat, are there any other concerns I can assist you with?
Josh Bernoff: no.
Kai: It was a pleasure to have assisted you today, Josh. Thank you for contacting G Suite support! Have an awesome day ahead and always keep safe!

Apparently Kai doesn’t know how much I hate exclamation points and cliches. I don’t enjoy “Have an awesome day” after I’ve just wasted half an hour.

Josh Bernoff: bye
Kai: Bye bye
Kai: Thank you for contacting Google Cloud support today. Now that the conversation ended, you will receive the chat transcript in your email shortly. Have a nice day.
Kai: Kai ended the conversation

In case you are wondering, they did get back to me and were similarly unhelpful on the phone and by email. In one case, the support person typed the wrong domain name in his email to me about the problem, which disturbed me. One solution they are proposing is to delete the account altogether, which would be fine — there are no files and no emails stored in the Google account for Wellness Campaign. But now I’m terrified they will mistakenly delete because that’s in my email address. If that happens, you will find me curled up in a tight little ball on the floor, crying.

Imagine if you applied intelligence to a customer’s support interaction, not just advertising

Google (and Facebook) are very smart. They know so much about you. Google reads your Gmail and shows ads related to it. It customizes your searches based on your history and the searches of millions of other people. Facebook decides what you’ll love in the news feed and what ads you would like to see, and is uncanny in its ability to show you opinions that you and friends share.

Why do they throw that intelligence out the window when it comes to support?

Why can’t Google identify that I’ve been repeatedly trying to access an account? Why does the chat rep have to even ask? Why ask me questions that I’ve answered already, and why not connect me to the resource I’ve asked for if that’s what I want?

Similarly, in the recent congressional hearings, Al Franken asked why Facebook didn’t reject American campaign ads paid in Russian rubles. It certainly seems as if applying this basic level of intelligence could have prevented some of the Russian meddling in the election. The Facebook representative wouldn’t even commit to making that the policy in the future.

I’ve begun working with some technology vendors who apply artificial intelligence to customer support and marketing. The result is communication — including chat — in which machines anticipate customers’ needs based on intelligent algorithms and all known context (who is the customer, what is their history, what are they doing right now, what did they just try to do). A smart machine chatting is way better than a stupid human. Even when the machines fail and turn you over to a human support rep, the rep knows all the context, which makes the resolution much faster and more painless.

This is going to be the standard for customer experience in the near future. Google and Facebook ought to be leaders at this. They’re not. It’s time they applied their intelligence to the problem of creating the best possible member experience, not just deriving as much money from members as possible.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.


  1. Mostly all technical support is abroad or automated with Alexia-like systems. Google is one of them with Smart Home and probably business or Artificial Intelligence (AI).
    If you getting frustrated with AI,… welcome to the unreal world where stupidity and mediocrity become main human interaction.
    Did notice that mediocre interaction is everywhere?

    A little off reasoning with Russian Rubles and Intervention in USA Election, but this is your post. Will be great if somebody makes the same questioning using Artificial Intelligence to find about How a private Russian company controls approximately 20% of USA Uranium Reserves

  2. I also feel your pain. This happens with just about every chat session I am involved with. I am convinced that each of these companies has at best one person that knows what you need to know and all the back and forth repeating is just a delay tactic till that person becomes available. Just when through this with my contract renewal with my internet/cable provider. He spent 40 minutes going back and forth with me, finally offered me a rate/services combo that I wanted. Then went to check something and came back and said “sorry, I cant offer you that any more” Which meant we had to start back at the beginning. Two topic ideas for you…1) Internet “Self Service Customer Support” is euphemism for no support (on some websites it is impossible to find a phone number or a street address) 2) In the age of bundled phone/TV/internet services, why is it impossible to lower your monthly bill? Unbundling seems to always cost more.

  3. Kinda makes me wonder what if Elon Musk saw this thing happen at one of his companies serving customers.

  4. The other end of the chat was probably an individual in an ESL country where they haven’t the context/flavor of the American dialect and are using an FAQ script to respond. A bio-bot.
    To test this, use a foreign word for the Anglo one ie: Si for Yes and wait for a response.
    It should prove interesting.

  5. How many of these chats does one human handle at any given time? I always get the feeling that they person I’m chatting with is repeating everything I say as a delaying tactic, so that he has time to remember where we were in our conversation, because he’s handling three other conversations simultaneously.

  6. Hey Josh,. Just a thought, but to me this transcript reads like the output of a chabot – not a human. I don’t know how GSuite support does it in particular, but I have built chatbots using Google’s chat technology that respond like this. The only part that seems to me to be more likely a human is where Kai replies that the response will be within 24hrs. It may be that the conversation started between you and a bot, then at some point it’s handed over to a person who can then read the”notes” that the chatbot has collected.

    Of course that doesn’t solve your problem and isn’t any less frustrating, but it’s interesting to understand where the handoff between computer and person occurs – and then even more important is to try to improve that.

  7. So frustrating. I will say, though, that I try not to assume the person on the other end of the chat is stupid. They might not be Albert Einstein, but chances are good they’ve been conditioned to follow the script and not deviate from it even if they know the script sounds stupid. That’s the culture part of this issue. Companies need to stop turning human employees into robot-like beings by treating them like 2 year-olds.