We've all encountered it; an automated helpdesk, usually pretending to be a human, which persistently misinterprets your requests.

The only time I ever succeeded with one was by typing what I hoped were machine insults like "why don't you dissolve yourself in fuming nitric acid" ; after a few such it asked me if I wanted an email address (I described my problem in an email, a few hours later I got a working solution).

So does anyone know a [life]hack for getting access to a human via an automated helpdesk?

PS xkcd had a take on it, transcript here, but his trick didn't work for me.

EDIT I maybe wasn't clear what I meant by "automated helpdesk". It was a while ago and I don't remember much about it, but the conversation went something like this (I made this up, but it is a realistic indication of the level of the conversation). I had already given my account details, so it knew my name and the like.

PC: Welcome to FooBar's help site, NL_Derek. I am Lisa, the automated 
   assistant. How can I help you?  
Me: How do I convert my mail account from pop3 to imap  
PC: You can access your mail via the webmail button on our site www.foobar.com  
Me: I already have a pop3 client, I want to convert to imap  
PC: Your client account details can be accessed via www.foobar.com  
Me: email account type?  
PC: You can access your mail via the webmail button on our site www.foobar.com  
Me: imap/pop3  
PC: I don't understand. I am Lisa, the automated assistant. How can I help you?  
Me: You are useless, how do I contact a human?  
PC: On www.foobar.com/usability you can read how our site works  
Me: I've visited every damn page on foobar.com, all I want to do is to go from pop3 to imap  
PC: There's no need to become emotional. I am Lisa, the automated assistant. How can  
   I help you?  
   [more of the same]  
Me: You are worse than your namesake  
PC: You can change your display name on www.foobar.com with the Profile button  
Me: Go dissolve yourself in fuming nitric acid  
   [several more insults]
PC: Would you like to mail your question to our support team?  
Me: [Yes yes, please, pretty-please ... oh-oh, better not confuse it] Yes  
PC: Send an email to shibboleet@foobar.com  
PC: Please rate my performance on a scale of 0 (very bad) to 10 (very good)  
Me: -100  
PC: Thank you  
  • I'm iffy about this question... I feel like its off topic... but I'm not sure – Just Do It Jan 7 '16 at 21:15
  • @JustDoIt I wondered about where best to ask, and settled here. Open to suggestions of course, but I believe it's a useful question. – NL_Derek Jan 7 '16 at 21:50
  • I just realised (after posting my answer) that maybe you did insult an AI. I wonder if there are some insults that it recognizes, or if it is something that is build in when it doesn't make sense of what you way. Speech recognition is weird! I have seen a system to learn English that requires you to speak the words and is picked up by AI... my brother had to say "the newspaper" and the AI didn't get it. Then he said some insult in my native language and it got it! o.o – Theraot Jan 8 '16 at 5:47
  • @Theraot it was (IMHO) only an AI in the most generous definition of "intelligence". I've added a (fictional) transcript to indicate the level of "intelligence" involved. And it wasn't speech-recog, just typewritten "chat". – NL_Derek Jan 8 '16 at 23:07
  • Having RTFM I have to agree that it is off-topic (slaps own wrist). But what now? @Theraot provided a number of useful tips about helpdesks and chatbots, which I think should remain accessible to SE users. Should I delete the question, or just leave it as closed? – NL_Derek Jan 10 '16 at 22:11

The Chatbot case

As per OP's clarification we are dealing with a text only chatbot. It is a very basic AI if any, and it does not use Speech recognition.

A way to understand how these run-of-the-mill chatbots as follows: they have a list of categories. They internally have a set of keywords for each category (some keywords may appear in two or more categories). They scan the text you write for keywords... each keyword they find contributes points to particular categories. Afterwards they take the category with more points, and so they assume you are talking about something in that category.

Once they have stablished on what category your text falls, they will have to choose a response from that category. The worst chatbots I have seen have around ten categories and only one response per category.

Slightly more advanced chatbots may have memory of what has been told, they may try to use spell checking or sounds-alike searches to try to force the text into a category. The next level of chatbot technology is able to do rudimentary phrase analysis such as ´{pronoum} {verb} {object}´ - this is not the case here, tho.

It may be worth noting that - while no chatbot in the world has passed the Turinh Test (for real) against a complete set of experts examiners - some chatbots can actually fool unsuspected people. Aside from being used in customer support, they have uses in video games, they have uses as part of scam schemes, etc...

Note: top-of-the-line chatbots are way better than the one involved here. Although even the best won't convince everybody to be human, and hardly will convince anyone to be a particularly smart human.

Consider this case:

Me: You are worse than your namesake
PC: You can change your display name on www.foobar.com with the Profile button

The chatbot has - probably - only recognized the word "name", has put the text into the category of "thing related to the name of the user" or something like that. Ans has come up with a pre-cooked answer for those type of questions.

Another example is this one:

Me: I've visited every damn page on foobar.com, all I want to do is to go from pop3 to imap
PC: There's no need to become emotional. I am Lisa, the automated assistant. How can I help you?

I'm guessing the recognized word here is "damn", and this is probably a standard fallback for all insults in his database. Being it a fallback, it ignores all else.

Knowing this you may try thinks like: Contact human, Contact email or something like that, that may trigger a response in the right category. Avoid unecesary words, brevity is your friend.

See this case:

Me: You are useless, how do I contact a human?
PC: On www.foobar.com/usability you can read how our site works

It could be the case that they expect - maybe they noticed, they could have word statistics - that people may complain that the service of www.foobar.com is useless. Or simply it matched it to "how to use" category. Whatever the case they have a predefined answer to direct people to www.foobar.com/usability. The same query without "You are useless" may have worked - may be not. But this a particulary dumb chatbot, it doens't understand that a sentence may have more than one phrase - in fact, it probably ignores all punctuation (and short words).

Addendum: It has ocurred to me that perhaps the reason why why don't you dissolve yourself in fuming nitric acid may work is because the words didn't fall in any category, not even the insults fallback. So you may try very specific words that has nothing to do with the web site, along the lines of phosphate, eletromagnetism, manifold, tetragrammaton, malapropism, valetudinarian, ragamuffin, etc... you get the idea.

Here is the deeper problem: there may not be a human to answer the chat. It is possible that the developers of the webpage has included this catbot mechanism as a "user friendly search", they may have worried about usability of the page and used it as a bandaid. Or it could be a gimmick to appeal to sales. Or somebody has used steak and sex to sell the chatbot to their boss and now they have to add it because reasons, etc... Whatever leaded to this chatbot, it doens't imply that there may be humans. And even if sometimes there are humans, they may only be part time - hiring lots of people for covering different shifts and hight demand hours is expensive.

Of course, checking every damn page on foobar.com^1 is the way to go. I usually don't bother with chatbots. Yet, since you say you did and didn't find what you were looking for...

Take the high route on the chatbot, there is no need bother with it.

Try using a search engine. You can use site:foobar.com on most search engines^2 to restrict a search to a particular internet domain. Using keywords such as contact and email, you may also try common email providers such as gmail.com (or with at @gmail.com it may or may not match depending on the search engine), etc...

Find them on social media. You can try using Facebook search, Twitter search, Google+ search, etc... it is common for there to be somebody hired to check those places on a regular basis. If there is no email, this could be the next best thing. Although, due to the bulk of input they may be getting this way, they could just ignore it.

Phonebook. Some telephone companies - and similars - have online search for their phonebooks. Try finding one from city of the company. If you find une, you could use it to get a phone number or email address of the company you want.

^1: Didn't you miss any? There are solutions to crawl and download a web site. HTTrack comes to mind (it may go against Terms of Use). Once downloaded you can search there using - for example Agent Ransack or similar. Although I would consider that a last reasort option.

^2: Tested on Duckduckgo, Google, Bing, and Yahoo! Search.

If none of the above works, it means that this company is particularly not approachable. You may try to send a letter, yet if you are still insisting you probably want to try some not offical ways to contact them.

The options below can be considered competitive intelligence - barely

Use whois service. Some webpages such as internic.net among others allow you to get registration information about internet domains. This information may include an email. If there is no email, look for the entry "Whois Server" visit their page and try searching for the domain there - they usually have more information. It could be the case that the email you find is fake, that is the case if it was obscured by "privacy reasons", which is becoming more common, but it is still worth a shot.

Specialized services. I have come across the EmailHunter which will search for email addresses for a particular web domain. Think of it as a web search but it can only find email accounts for the domain you enter - it may have fake results, as it goes by what is on the web, you will need to try to use the email to know if it is real. There is a paid plan, yet it can be used for free with some restrictions. Another alternative comes from CEOEmail which is a database of well known contact information of high staff members of big companies - query by the name of the company.

Find employees. Again, in social media. If you can find some approachable employee [I opt not to explain means to try to do this other than regular search], get a way to ask them. Who you find is hardly who may help you, but may know who can. Since you don't want to pass as a stalker, you will restrict yourself to publicly available information and will not try to contact a person insistently (nor from various means at the same time).

Finally, according to CleverBot:

ME: How do I contact a Human?

CleverBot: Try signalling out of the window on the first Friday of the month when the moon is full.

Sorry, I have no more to offer.

Orignal Answer

I'm keeping this answer because it may be useful for thirs parties that may find this question - yet, sice the AI is a text only chatbot, this doesn't address the OP's question.

We've all encountered it; an automated helpdesk, usually pretending to be a human, which persistently misinterprets your requests.

If we are talking about a machine, it is common that 0 will take you to a human. Otherwise listen if it give you a menu, and see what is the right option to talk to a human. If this is an AI, ask for a human operator - insist! (speech recognition is not perfect).

On the other hand, If you started by offending the person who attends you...

First understand that helpdesk job is low entry and with high personal rotaton. People may end there because it has been particulary hard to get a job doing whatever they studied for - and probably don't want to stay there longer than needed. Also, a good portion of the requests they get are dumb people who doesn't know how to operate a machine, a another good chunk are simple operations. By policy they make you follow some simple instructions that usually are expected to cover around 80% of the cases. The call is recorded - and sometimes a supervisor may be listening in real time - and if they don't do as the policy says, they may get in trouble.

I have friends who did work in call centers, I have formed my opinion from there.

While I would agree they are not particulary tech savvy, they are humans.

Work with them

The first option is always to try to work with them. To make them skip the normal steps provide as much information as possible, also explicitly say what checks you did. For instance - in the example of failing Internet connection - say that you have a modem XYZ (where XYZ is the manufaturer and model if you know it), that it has XYZ leds on, that it is properly connected, that you have already disconnected it from the power and conneced it back again, and that you have checked with another network cable and another computer... and none did work.

If you are tech savvy as I presume, you should be able to do - and proably already do that kind of stuff to rule out all else before calling. If you are a bit like me, you want to avoid calling if reasonable.

Tell them all you did, and they will usually move on to check the connection remotely and try to fix it on their end, they may schedule a tech visit, or give you another mean for contact. All depending what is the case, and what are you calling for.

Note: politeness doesn't buy you anything, but insult usually won't work either. The reason it did seem to work is because they may ask you if you want to use other means of communication to get rid of you. But that is not a solution, just a diversion. If you want a solution, just go to the point, and be explicit.

It did not work. Ok...

Ask for a supervisor. If you insist in talking to a supervisor and they don't allow you to do it, they are sure to get in trouble. Also, it is a call attended for them - and they may be happy to get rid of you. It is win-win.

Once you get a supervisor in the line - sadly you may have to tell again what is the case. The supervisors are usually more experienced, and they they may have the power to put on the line somebody who can assist you better. Note: this changes from company to company, some do outsource the call center service and so there is no way to put on the line the people who actually can fix the problem.

It did not work. Ok...

It usually means they don't have a solution for you. You cannot force the company to do whatever you want. So back off! This is probably a company you don't want to be stuck with, they have bad customer service, look for alternatives.

But there may be a way to get information from them. If you wanna put more resources in the matter.

Depending on your country and region/state there may be some law that you can invoke to make a formal petition for infomation. Learn about laws related to right to access information, it may be under freedom of the press, customer support, or something like that.

I don't know if that is the case in USA, but it is the case in my country.

You may be able to get free judicial consulting from law schools, where students near graduation do this kind of service. Find those near your area, and ask if do. You will have to pay a visit for that, tho.

If there is a law that can apply. Invoke the law! they will now have the obligation to respond your inquiry. Which maybe something like "how to get tech visit" or "where do I take a defective product for revision or replacement", etc... just ask for information on how to get your problem fixed.

It is worth noting that it usually better to invoke the law in a letter, that you deliver personally, and for which you have a receipt copy signed by whoever receives it. That you have evidence that you did - in fact - invoke the law.

Note: At the start of a call the person attending you will usually introduce themselves. Write down the name of the person who attends you, and the time and date of the call. If they give you a number for the report write that too. If in you need to call again, you can say who attended you the last time, and if in the future you need to write a letter about the problem you can include that information.

In the extreme, If legal action is taken - because they are denying a right, or failing to comply, or similar - a judge may give the order to get a copy of the recording of the calls. Which should be futher evidence in your favor.

For example, if you have been trying to cancel (as in, stop) a service that is renewed automatically each year, and it is required to be requested to cancel it with anticipation of the renewal... and you did call in time, and they didn't help you... you may take legal actions, and demostrating that you did request the cancellation in time will win you the case. And you probably want to take legal actions in that case, as not doing so may mean to be stuck paying for another year.

| improve this answer | |
  • Good stuff Theraot, most I knew but a few useful tips; however I don't think it addresses my problem. I'll add a clarification to the question. – NL_Derek Jan 8 '16 at 22:40
  • @NL_Derek updated. – Theraot Jan 9 '16 at 18:40

Not the answer you're looking for? Browse other questions tagged or ask your own question.