In this guide, we will show you how to integrate SnapCall Flow into your Messaging Widget, allowing you to automate and simplify the process based on videos uploaded by customers.
This approach offers significant advantages over traditional check-in/check-out methods:
The application is entirely web-based, enabling customers to start and complete the process from any device without the need to install apps or compress videos.
Powered by AI, the system ensures that customer-submitted videos contain all the necessary information for agents to verify the process manually. Customers will only be able to complete the flow once all steps and criteria are fully met.
For the configuration we will use mainly the following tools:
SnapCall with a Flow Subscription: This will enable the creation of three distinct flows: two for the check-in and check-out processes, and one for customers to report any damage.
Zendesk Messaging Widget: This is where customers can record a video or take a snapshot. Once the media is collected, it will be organized either into a ticket or a conversation, allowing agents to review it offline at their convenience.
Ultimate Bot: While the standard Zendesk Flow Builder is sufficient for basic automation, the Ultimate Bot offers enhanced capabilities. It allows you to create advanced workflows based on the outcomes of the videos recorded by customers. For instance, if SnapCall detects that a video lacks specific required information, the system can prompt the customer to record a new one.
Configuring your SnapCall Flow - step by step guide
From your SnapCall Dashboard, navigate to the Flow section and click the "New" button. Give your Flow a name and select Video as the media type. On the next screen, define the instructions customers need to follow when recording their media. Then, move to the following screen and activate the Transcript, Summary, and Content Recognition AI skills. On the AI Output page, you can group all AI outcomes into a single private note within your Zendesk ticket.
Repeat the same process for the check-out flow. While editing the prompt for Content Recognition, write instructions that will prompt the AI to produce different outputs if the customer forgets to show certain data you'd like to capture from the video. This is a good example of suitable prompt:
You will be provided with media related to a customer returning a car to a rental company. Your task is to ensure that the customer has shown all required details, including: all four angles of the car (front, back, left, and right sides), the fuel level, and the car plate number.
As an output, provide a single string based on the content of the media:
If the customer has shown all the requested information, return 'success'.
If the customer did not show all four angles of the car, return '360'.
If the customer did not show the fuel level, return 'fuel'.
If the customer did not show the car plate number, return 'plate'.
Save changes and create a Flow to flag damages as well. While editing the prompt of Content Recognition, let the AI know it will have to flag if the video show any damage of a car. A prompt like the one below should do the job:
If you detect any scratch, damage or similar in the video, please write as an output: ''Damage detected and logged on our system.''
Save all links of your Flow as you will be using them into your Ultimate bot.
Adding your Flow to your Ultimate bot
Now that the SnapCall configuration is complete, we can proceed with adding these flows to a brand-new Ultimate flow configuration. You can define the process in a way that best suits your organization, but here is the approach we used to streamline and simplify the check-in/check-out process:
First, we ask the customer (using a carousel) whether they want to initiate the car check-in or check-out process. Then, we prompt them to enter their car's license plate number. For a more advanced setup, you could ask the customer for the car plate number first and then query your database via API using Ultimate to determine which process the customer should initiate (check-in or check-out), depending on whether the previous step was completed.
Next, we send two messages to the customer. The first message includes a sample video demonstrating a completed check-in/check-out process, making it easier for the customer to understand how the media should be recorded. The video is uploaded via the SnapCall Library and opened directly in the Ultimate bot via Webview. Click here to learn how to open a webview via Ultimate.
The second message contains the link to the Flow we just created, which the customer can also open via Webview. Be sure to add the following parameter at the end of the flow link so that SnapCall can associate the media with the conversation that Ultimate creates in the background.
&ticket=smooch_{{conversation_id}}&customerEmail=CLIENT_EMAILThe customerEmail parameter can be skipped or you can add a dynamic value within the URL in case you'd like to either ask the client to provide this value earlier in the flow, or you'd like to take it from your database. If unspecified, client will be asked to type an email address before having the possibility to record.
On this second step, make sure to save the conversation id as a parameter, so that SnapCall can use it to add the media directly in the conversation. Learn more here
Once the two bot messages are configured, add a 'visitor message' step. When the customer uploads a video using the Flow, SnapCall will post a new message within the chat using the related conversation ID. This message will appear as a visitor message, ensuring that the flow continues only once the media has been received. In the same step, make sure to save all conversation metadata from SnapCall, as it includes the AI output. You should configure it similarly to the example in the screenshot below, but feel free to refer to this FAQ article for additional guidance if needed.
At this stage, you can either prompt the customer to record evidence of any car damage using one of the previously created Flows or skip this step and proceed to the next one. If you choose the first option, follow the same instructions as outlined earlier, but update the Flow URL to the second one you plan to use.
Building conditions on your Ultimate bot
We now want to build conditions according to the AI output provided by SnapCall. To do so, let's use a Condition (green) step from the Ultimate Flow Builder.
If you're using the exact prompt I shared earlier in the article, you can set up the first condition as 'description' (which is the variable where we stored the AI output from SnapCall) IS 'success', so that the first thing Ultimate will check is whether the video satisfies all the parameters we defined in our prompt. In the ELSE part, we then build a cascade where Ultimate checks each possible value of the 'description' variable, as shown in the screenshot below.
In the example above, we use SnapCall AI to determine if the video uploaded by the customer includes all the necessary details and information. If it doesn’t, we ask the customer to record a new video, eliminating the need for an agent to request the same at a later stage. This significantly reduces resolution time, as it ensures the customer cannot complete the check-in or check-out process until all criteria are met.
Submit the ticket to an agent for manual review
The final step is submitting all the information to an agent so they can finalize the process offline on behalf of the customer. For this, we recommend using the 'Escalation' > 'Forward to an agent' step. This approach is beneficial for the following reasons:
The agent will have access to the full conversation history between the customer and the bot.
All media recorded by SnapCall will be available on the right side of the Zendesk interface, along with a summary, transcript, and additional details.
Agents can handle the case either offline or live, depending on your company’s processes.
Have any questions? Feel free to reach out via our chat widget or email us at support@snapcall.io!