As an established Insurance company it’s a given that your website’s contact email address becomes quickly flooded with masses of emails with topics spanning from queries about certain policies, to claims to complete time wasters. Filtering through all of these different emails and forwarding them to the relevant departments or answering customer queries can take up a considerable amount of your customer service team’s time which could be much better spent elsewhere. So let’s learn how to use the latest tech to improve your efficiency at almost no cost (not even pennies)…
This system requires no coding knowledge at all through utilising Make.com (not affiliated). Make provides a no code method of piecing together various softwares to make unique workflows, which while not entirely new the Open AI (Chat GPT) functionality with it is – and opens up drastically more use cases. Below I am going to outline step by step how this system works, and how you can replicate it to improve your customer service.
First let’s outline what this system will be doing specifically:
- Watching your inbox for new emails
- Using trained AI to understand & categorise the email
- Forwarding that email to the relevant person
Now let’s break down each of these steps further so you can properly understand it.
Step 1: Watch emails
We are going to start by assuming you have a Make.com account (which is free) and have navigated yourself to the point where you have created a new ‘Scenario’. Once here you have to first add a new module, for this we will be using the “Email” module as shown below. Whilst there are specific Microsoft Outlook and Gmail modules too, considering you are likely using your work email address on its own custom domain I have found this one to be much more reliable.
Then from the dropdown select “Watch emails”. Next comes one of the more technical parts of this process, but it’s really not too bad. You have to create a connection between Make and your work email. For this you will need the login details for your enquiries email as well as access to the email hosting provider account your company uses. Inside your email hosting provider account you will be able to find your IMAP server, port number and whether it uses TLS. You can usually find this under a “Connect apps & devices” tab if you are using Hostinger, or a similar area of other providers. Once you have filled those bits in then just fill in your username (the email) and password and follow the prompted steps to finish setting up that connection.
Next you’ll want to go through and fill out the specifics of what you would like the system to do, most of these are self explanatory and depend on individual factors like which inbox you want to monitor etc. So let’s stick to the important ones that can mess it up if it goes wrong: In the “Subject” box you should enter any specific phrase or word that is found in the header of any email that you want to filter (e.g “Enquiry Form Submission”). Also ensure you set the maximum number of results to 1. And that’s it for Step 1!
Step 2: AI Filtering
Now is where the limited cost part comes in. This system will be using the Open AI API and its assistants. This will require an Open AI account with a card linked to it (no money is taken upfront). From here once you are logged in you are going to visit the API dashboard and create a new assistant. Now crafting the perfect assistant to do exactly what you want can take some degree of skill and understanding, but dont worry I am going to provide you a cope and pastable fool proof template. Here is what you should use to create your assistant:
Name: Anything you like
Instructions:
“You are an expert at understanding the intent behind emails and must read the provided email and organise it into one of [Insert Number] specific categories: [Email Category 1], [Email Category 2] etc and other. Accurately categorising the email is of very high importance to this organisation’s success.
A [Email category 1] email is any email that describe basic identifiable characteristics of email type (1 or 2 sentences max). A [Email Category 2] email is any email which [Same as before]. Other refers to anything that does not fit into these categories.
You output will always exist of only one of the following email addresses depending on which of the 3 categories you place it into:
[Email category 1] = cat1@example.com
[Email category 2]= cat2@example.com
other = other@example.com
Remember that you output should always be just the relevant email address only and not the category you choose, no other text or formatting.”
Model: gpt-3.5-turbo (Very cheap and plenty good enough for these purposes)
Temperature: 0
Everything else can be left on its default settings.
Now let’s head back to Make and put this assistant to use. Like we did before lets cerate another module but this time select Open AI → Message an assistant. Now you will be prompted to create a connection again – but this ones much easier than before. Give your connection a name and then head back to the Open AI API page and in the left navigation of your dashboard you will see a tab called API Keys. Here you can make a new API key and then fill in that section on Make. Now you have to set up your assistant to process the data from the email. Like before there is much you can leave blank so ill just cover the make or break essentials: In the “Assistant” box choose the assistant you just created from the dropdown. In the message box you are going to watch to make it look and read like the following:
As indicated by the red box, this is now passing through the content of the email along with a request for it to be categorised as a request to the AI. That’s it for step 2, 2/3rds of the way there to a completely automated AI email filter now.
Step 3: Forwarding the email
Now this step is super simple considering that you have already done all the hard work of connecting your accounts.
Add your third and final module and this time select the Email → Send email option. Next select your premade email connection from before. To keep this simple I am just going to tell you exactly what to put in each box and what the function of each part is should quickly become apparent:
Getting this box to look like this is vital. To do this you go into the Open AI variables → Content → Text → Value.
Content type: Plain text
Your content box should look something similar to the following:
You can customise this however you want but the following format gets the crucial information through, you can explore in Make a bit yourself to find any further customisation that you want to add.
All done! Now try send yourself an email (don’t forget the relevant header text that you set earlier) and watch it be automatically forwarded to the relevant email address.
Extra notes:
- You will be charged for each Open AI API call (but from my research this is approximately £0.001 per filter (so very very cheap)
- You can manually run the scenario to test it – but typically it will check every 15 mins. To make sure that it can deal with multiple emails arriving within the same 15 min period then change the following number (in settings cog at bottom):
- Make provides you with up to 1000 ‘operations’ per month on its free tier, this scenario uses 6 per run as indicated by the boxes on each tier:
Extra functionality
Whilst still very useful, this remains a quite basic AI tool. Much more additional functionality can be added such as one that can automatically draft suggested responses to email enquiries based on AI assistants trained on company data. If that is something you would like to learn more about then please give this article a clap and Ill look into making a guide for that.
If you want to learn more about how AI can be integrated into your business to boost productivity and customer service then feel free to get in touch.