What is the first thing you do each morning? For me, it is reaching for my phone to check my emails. It is a terrible habit, I know, but I think most people are the same. This alone shows us that email is still relevant, and so is email marketing.
It is one of the oldest forms of Internet marketing and is still going strong. As it turns out, prediction algorithms and machine learning are useful here too. With enough data, algorithms can measure what worked best in a campaign, and adapt and improve your email outreach over time.
One of the most common uses of AI in email, though separate from marketing, is a popular feature in Gmail. You might even have used it yourself! Gmail recommends suitable short replies to incoming email automatically with its smart reply technology.
Email is still relevant, and so is email marketing.
The Gmail inbox analyzes incoming messages with natural language processing (NLP) and natural language understanding (NLU) algorithms, and creates a suitable short response automatically using natural language generation (NLG).
You can simply reply with the proposed text using a touch or click, or extend the reply by writing more in the auto-generated email draft.
A company called Conversica takes this concept further. They offer an AI-driven virtual assistant that initiates fully automated email correspondence with your leads in plain English.
In effect, you get a virtual sales assistant that can send email messages back and forth, thereby nurturing your new leads into customers automatically. It is like a chatbot, but over email. With solutions like this, your sales reps can focus on the leads who are interested in buying, instead of wasting time of cold leads who will never covert.
Your customer service team can also be relieved of answering common and simple questions.
I had a conversation with Gary Gerber, senior director and head of product marketing at Conversica. He says the platform combines natural language processing to understand incoming messages, an inference engine for basic decision-making, and natural language generation to create human-like messages.
He says, “Think of all the money and resources being spent on marketing activities to drive that interest, and then not reaching people. So, an AI-based assistant, who will follow up with 100% of the leads as many times as it takes to engage them, makes perfect sense.”
But does it work? His answer is to give an example from one of their customers: “Leveraging their AI assistant, the Los Angeles Film School generated a 33% increase in sales pipeline, which wouldn’t have been humanly possible otherwise.” Conversica works on mobile text (SMS) messages too.
A company called x.ai also does magic with email. They offer an AI-based scheduling assistant called Amy or Andrew, depending on your preference, that can be used for automated meeting bookings and scheduling of other things in the calendar.
It does so by parsing your email correspondence automatically and integrating with your calendar. The main benefit here is to get rid of the endless email chains negotiating a phone or meeting time that works for both parties, or when rescheduling is needed.
The scheduling robot can reach out to your contacts and propose a meeting, and it does the time selection negotiation for you using automated email correspondence and calendar integration.
We will probably see a lot of virtual sales and customer support assistants from an array of vendors in the near future
We will probably see a lot of virtual sales and customer support assistants from an array of vendors in the near future, and it is interesting to see some of them are getting email and text/SMS capabilities as well, in addition to the more common chat windows.
A crucial point is whether the natural language correspondence feels sufficiently human-like, of course. Your mileage may vary here, dependent on what solution you choose, but we can expect the bar will be raised significantly in this area in the coming years.
It is only a matter of time before conversational AI-bots appear fully human, at least most of the time.
It is only a matter of time before conversational AI-bots appear fully human, at least most of the time.
Let’s move on to a completely different use of machine learning with emails. Marketing automation suppliers are beginning to use machine learning to optimize email send times individually for each recipient of an email broadcast.
This helps improve email open rates and can help reduce the load on customer support centers as well.
Seventh Sense goes even further to predict how frequently you should send to each contact in your database, optimizing email cadence based on their unique behavior and engagement patterns. In effect, this enables throttling of email delivery frequency to reduce email fatigue on a per-recipient basis.
Writing well-crafted email subject lines that give a high open rate has been an important task for improving email campaign results for years. Copywriters use their experience and gut feeling to come up with compelling subject lines that boost the likelihood recipients will open the email.
Unopened emails will not help your marketing efforts, as we all know. In fact, writing compelling subject lines can be automated using AI, which optimize this step better than humans can.
Writing compelling subject lines can be automated using AI
Examples of this are Phrasee and Persado, who offer language optimization solutions that uses machine learning to analyze historical response metrics and recommend subject lines that they think will have a higher open rate.
Self-optimizing subject lines are great, but it is possible to take this one step further. A potentially disruptive use of AI in email marketing is the automatic generation of email copy, hyper-personalized and optimized for engagement from each recipient.
In fact, this personalization can go well beyond just the text. The imagery in the email can also be personalized based on the digital footprint of each recipient, or be dependent on external factors, like the current weather, football match results in the recipient’s area, or other near real-time data that may be available.
A few years down the road, AI tools might even generate email graphics automatically using conversion optimization algorithms that know how to generate compelling images—hyper-personalized for each email recipient, of course.
Boomtrain and OneSpot are examples of tool vendors that do email personalization. Their solutions include user behavior and content analysis to surface content that is likely to resonate well with a lead, and delivers it automatically using channels like email, the web, or push notifications.
Personalization is also available for product recommendations included in the email copy. An email can promote the product you are most likely to buy based on your previous behavior or other data.
For example, I might be recommended red shorts in an email campaign from a clothing shop, whereas my neighbor gets an email that recommends blue t-shirts instead. Machine learning algorithms mine the digital footprint and other data to conclude which product each email recipient is most likely to buy next.
You can also use machine learning to understand what email to send, to what person, at what time.
Use machine learning to understand what email to send, to what person, at what time.
This type of email sequence optimization is offered by companies like Optimail. This helps optimize the overall efficiency of automated drip email sequences.
Automatic generation of customer segments can also be done by machine learning algorithms to figure out what groups of contacts should receive a certain email campaign.
Not surprisingly, machine learning can be used for housekeeping duties too. A company called Siftrock uses machine learning to analyze email correspondence and update the lead and customer database automatically.
Machine learning can be used for housekeeping duties too
For example, an AI bot can remove contacts that bounce (are no longer with a company), update contact information when someone has changed job title, mine phone numbers out of email footers, or scan out-of-office auto replies to find new leads within that company. This is done by integration with the existing CRM, emailing, or marketing automation system.
It is clear that machine learning has many uses for optimizing email content and improving email campaign performance overall, as well as leveraging email correspondence for completely different usages as well.