Transcript
Welcome. In this video, we will be tackling three major business problems you may be facing. Number one, maybe you’re wasting a lot of time answering the same questions over and over and over again, which leads to loads of hours wasted every single week and reduces the team capacity. Number two, it takes you too long to even reply to the people and then you lose them.
Or number three, your DMs are a complete mess and you can’t keep up with answering them. All of this leads to misleads, lost sales, and a negative customer experience. But not to worry, right now we will set up an FAQ automation together. The slow response fix automation as well as the inbox overload relief.
Let’s go. So, let’s get started to set up an inbound voice AI agent that can handle all of these FAQ questions. Click AI agents on the left. Select voice AI.
Click create agent. This time we are going to choose create custom agent. Hit continue. This one we will name the inbound FAQ assistant.
Right here, you can adjust the business name, the language, the voice, the time zone, the language model you would like to use, and an initial pickup message. Once the voice AI agent picks up the phone, what is he or she going to say? Hit next. Next up, you need to connect a knowledge base.
If you haven’t created one yet, click create knowledge base. Give it a name, for example, FAQ assistance, and add a description. Hit create again. And right here, you can add a source.
You can add a web crawler, meaning the AI agent is trained on all the information that is contained on your website. You can add FAQs. You can add tables, rich text, or even upload files, PDF documents, word documents, and so on. So for FAQs, for example, think business hours, services, general pricing ranges, location, or availability.
If someone were to have to stop and think before answering, it’s not an FAQ. Voice AI is really great at clarifying and judging. Judgment and nuances should still belong to humans. So first, define your boundaries and ask yourself questions.
What can AI answer and what must go to a human? Pricing is a very common example. AI can share a price range but should never give the final pricing. So here we just enter your question and also add your answer.
Then hit save and keep adding questions that are suitable for the FAQ catalog. Then let’s go back to our AI agent and let’s connect the knowledge base we just created. Click the dropown menu, select your knowledge base and now the AI agent will be able to answer questions that comes from this knowledge base. Here you can define rules when your agent should be answering questions.
Always or anything specific. You can get really into detail what you would like the bot to do. You can also define fallback behavior. Decide what will happen if the question isn’t clear.
For example, and fallback responses should acknowledge the question, set expectations, and write the route to a human or call back path. AI should always escalate instead of inventing answers. Down here, you can also collect the data from the person who’s calling in, which I would highly recommend. Just select the data points you would like to collect.
You can also trigger a workflow if you like. Check mark this box and then select the workflow in question and hit next. From here you will be able to select the phone number that you would like to connect to the agent. We want to turn off the backup because that’s the whole point of building an FAQ agent.
So the human doesn’t need to pick up anymore and we don’t want to turn on the working hours because we want it to work 24/7. Then on the right hand side you have the option to test your AI agent. You can start as a web call meaning people don’t even need to pick up their phone. They can just click a button inside the live chat that you can put on your website and it will then ring the AI agent or you can choose to call from a particular phone number.
Once everything is set and you’re satisfied, go ahead and hit save. Next up, let’s set up a slow response fix. This setup is designed that every message that comes in gets an immediate response. Many many leads get lost in this process because the business is not fast enough to reply and this is the fix for it.
So in this example, we are going to create a conversation AI board, not a voice AI bot. So, let’s click conversation AI at the top and create new bot. For the slow response fix, we have two great options right here. The promptbased bot, which will give you more flexibility or the guided form setup for more structure.
At this time, we’re looking to avoid appointment booking unless a booking is a part of the response flow. The key here is speed, not complexity. Let’s choose guided form setup, general Q&A, and hit continue. Here, you’ll be able to connect your brand voice if you haven’t created one.
Click create new brand voice. Add your brand name and again you can train this brand voice through your website content. Okay. So just enter your website here.
Fill out the details and under additional information you can add your target audience, brand values, brand promises. But this will be pre-populated once you add your website. Once you’re happy with the changes, hit next. Next up, you will give your board a name.
We will call it inbound first responder. Here you will select all the channels that you would like the bot to reply to: Instagram, Facebook, SMS, chat widget, live chat or WhatsApp. Make sure all of these are connected in your account under integrations. The wait time here is set to 2 seconds before the bot responds to the lead which is fine.
And then we will hit next. Then here we’ll decide what kind of data we collect. Name, email, phone, street, address or city. Then describe the question or objection and hit save.
Then we will select suggestive mode and save and close. Now we have an inbound first responder created. Next up we are going to create a knowledge base for this particular bot. Click on knowledge base at the top.
Create new. Name it inbound first responder. Hit create. The primary goal here is simple.
Respond immediately and engage the lead. Enable human handover for pricing or booking intent or if somebody asks, can I talk to somebody or if there is repetitive confusion and again same as before, click add source. Here you can add your website for the bot to crawl through the entire website. You can add FAQs, a table, rich text, or even upload a file.
Once it’s created, head back to conversation AI. Click on agents list and select your bot. Click on bot training and from here you can select the knowledge base. Now the agent knows exactly what it needs to do in any situation.
And again on the right hand side you can test your bot. Once you’re happy, hit save. You can see all the test conversations showing up under the conversations tab as well. If you don’t like it, go back and adjust your prompts and your instructions.
So go ahead and test this bot yourself by sending an SMS, a WhatsApp, a live chat, a Facebook message, or even an Instagram message depending on what channels you’ve selected earlier. And that is the setup for the slow response fix. Now let’s build a solution for the inbox overload and relief. This setup is built to help businesses stop losing leads because messages haven’t been answered fast enough.
So very similar to the last part we just built, the problem is slow response rate. So again, the key value here is every inbound message receives an immediate response even outside of business hours. No more slow replies, no more missed DMs. So to build the inbox overload relief bot, we are going to select a conversation bot again.
Click create bot, but this time we are going to select the promptbased bot. Create new bot. We are selecting general Q&A. Hit continue.
We are naming this bot inbox assistant. And here we’re going to select suggestive. We are going to move to autopilot after validation. Then again you will be selecting all the channels that you would like to connect.
Give it a business name. Decide how many seconds the bot will take until it responds. How many messages it can send maximum and if you would like to accept images and or voice notes as input. Then hit save.
Then we head over to bot training and we’re going to connect a knowledge base. This knowledge base should focus on common FAQs, business hours, service explanations, basic next steps and so on. The goal here is to create clarity. Remember the purpose is to reduce repetitive questions and not to qualify leads.
This inbox overload and relief is about absorbing volume. Next up, let’s configure the behavior under B goals. Hit save. The primary goal here is to resolve common questions without human involvement.
We will need to define human handover and only enable escalation for sales ready intent, complex or emotional messages, explicit human requests like hey let me talk to somebody or connect to human. But remember too many escalations will defeat the purpose. And here we have some optional enhancements as well. The auto followup for start conversations or dynamic channel switching but keep it simple at first.
So for example human handover here you would add your I want to talk to manager, I want to talk to human and so on. Here you need to select the user that the conversation is assigned to and hit save. Next up is to test the bot. Here on the right hand side you can test your messages.
You can ask it some general FAQs and see if it responds the way you would like it to be. All these conversations would show up under conversations. Once you’re ready, hit save and get your bot to work. And remember always to test your bots.
Some of the bots may need a little bit twisting and tweaking until they really hit your brand voice and they act according to your instructions. If your inbox still feels overwhelming, AI isn’t set up correctly. Once it’s working, this inbox overload turns into inbox control. And that’s a shift we’re looking for.