Forcing customer journeys | Community
Skip to main content

Forcing customer journeys

  • February 13, 2024
  • 4 replies
  • 0 views
Federico11

Hi all!

Posting this for sourcing tips and ideas from the community – either technical or "tactical".

Mine it's a very small team and we decided to implement ZD in order to facilitate customer service, of course.

Turns out that customers usually simply ignore the help centre, ignore the articles the bot suggests and simply skip to "chat with a human".

Furthermore, some requesters end up putting their phone number in place of the "name" field in the widget, asking to get phone support.

I must specify that, being a small team, we cannot offer a wide phone support (especially because the usual questions are to retrive info that are clear on the website – e.g. how much is it?). Secretly we do offer phone sessions, but only to high potential value customers that are further down the funnel and have more complicated questions.

I wonder whether any of you has gotten through the same issues and if you would have any good advice or tip to work around all of this.

Might there also be a way to block name strings that contain numerical figures? That would be a wonderful fix to begin.

Thank you in advance!

      4 replies

      Ola11
      • Ola11
      • February 14, 2024

      Make sure your bot is helping and not just suggesting articles. If it gives the answer in the conversation flow and links to an article with more info then the customer is far more likely to take the answer on board (clicking an article is effort!).

      Maybe ask for the customer's name earlier in the flow, rather than as part of the handover, so they're more likely to give their actual name. Also look at the wording you're using in the handover options.

        Harper11
        • Harper11
        • February 14, 2024

        We had a similar problem when we were using AnswerBot. Balancing good CX with efficiency is never easy.

        Eventually I removed all intent training for "Talk to a human." I re-named that as "Transfer me" and did not add any other keywords or trigger phrases for it. Only the exact phrase "Transfer me" would work. This effectively turned the direct transfer option into a hidden menu item that you could only access if you knew the secret passphrase.

        We then removed "Transfer me" as a visible option except at the end of topic flows where the customer indicated that they still need help. This forced customers to at least try to ask a question of the bot before they jumped straight to asking an agent.

        You could also train a second branch where if the customer asks for an "agent," or "representative," instead suggest some popular topics / button options and invite them to type in their question if they need help with something else. If the customer types in a question the bot doesn't understand, you can configure it to transfer to an agent at that point.

        We had okay success with this, but ultimately it's not a very customer-centric approach. We are instead moving to an LLM AI that's capable of engaging customers and answering their questions directly (not just suggesting articles or spitting out walls of text based on a keyword).

        Customers can usually tell very quickly whether they're dealing with a sophisticated AI or a "dumb" keyword-based bot like AB; if the latter, they'll unfortunately almost always ask for an agent without even trying to use the self-service option.

          Brandon12
          • Brandon12
          • February 15, 2024

          Hey Federico - 

          Ahh yes, the chatbot equivalent of smash 0 until someone picks up.  You've got a few options here.  Keep in mind, all of this is predicated on the assumption that you are using Zendesk Messaging.

          First, I would remove "Talk to an agent" as an immediate answer, encouraging them to chat with the bot.  Second, I would update your "Talk to an agent" answer to offer one last chance at deflection before transferring (ie, before we transfer you - many answers can be found faster just by asking me, the friendly virtual assistant). 

          Finally, should they persist you can adjust the prompts to explain why you need specific pieces of information. "ie please provide your email so that we can provide you with a copy of this conversation."

          Should you need further assistance, I'm happy to connect with you for a free Zendesk Health Check.  Best of luck with keeping your humans out of the loop.

          Brandon

          Grace17
          • Grace17
          • February 23, 2024
          Hello Federico, 
           
          Thank you for your question, and thank you to our Community members for their answers so far! 
           
          The suggestions made here are very helpful! It's okay to make your customers do a little bit of work before connecting them to an agent. When it comes to self-service, the goal is to have more help center views than ticket submissions (this is how you can calculate your self-service ratio). 
           
          I agree, by removing the "Talk to an agent" as an immediate answer, you're encouraging your customers to fill out more information before getting in touch with an agent; whether you have them go through the Bot Builder process or a form submission, by having the customer fill out information at the front end of an engagement, your agent will know in more depth what the customer needs assistance with and also reduce the amount of time spent on a ticket. 
           
          Another thing I'll point out is that you can optimize your bot for better article recommendations and flows if need be - you can learn more about it here and here.
           
          Finally, we offer our article called Fine Tuning: Best Practices for Ticket Deflection, please feel free to check this out! 
           
          Best,
          Grace
           
          Terms & ConditionsCookie settingsAccessibility statement