In the digital world, which is overloaded by an overwhelming quantity of promotional content, the advert is becoming less effective. People are bombarded with a constant stream of colourful but meaningless content, which develops ad blindness. The problems in marketing mean that it fails to satisfy both consumers and advertisers.
To those who have planned advertising campaigns this year, this state of affairs will seem disheartening, but this is also a sign that the marketing field is ripe for modernisation. It is time to make use of the capabilities of artificial intelligence (AI) in order to take a step forward in advertising.
Artificial intelligence systems have long worked in the popular products of such companies as Amazon, Google, and Netflix. In recent years, technology has been incorporated even further into marketing to help brands reduce the number of steps consumers take to purchase. Furthermore, the use of artificial intelligence in marketing is now accessible to small and medium-sized businesses, not just giant companies.
Formerly, when choosing goods people always gave preference to those products which friends recommended to them. By contrast, nowadays many rely on the opinions of independent reviewers on the Internet. In this way, potential customers form their views on future purchases or particular brands in advance.
Naturally, brand owners want to influence societal opinion and facilitate the creation of a positive image for their products. They can train AI to write articles and flattering reviews on a multitude of websites and social networks. It is enough to upload the characteristics of a product to the system, and descriptions of its specific advantages will disperse across the Internet in the form of unique reviews and use cases.
A recurrent neural network (RNN), created by five researchers from The University of Chicago in August 2017, became a pioneer in the writing of human-like reviews. They trained it using thousands of genuine reviews that customers had posted on Amazon, TripAdvisor, and Yelp. As a result, the RNN was able to write reviews that were almost indistinguishable from those written by people. Moreover, since the RNN independently generated the reviews, rather than assembling them from phrases on websites, the texts successfully passed plagiarism checks.
In February 2019, the organisation Open AI presented GPT-2 — an AI with an open-source code. They even demonstrated a version that was configured for the generation of an infinite number of positive or negative product reviews. Since then, companies have been using similar models for marketing for over two years. It has reached the point that other companies have been forced to train AI in order to identify fake reviews, both machine and human, because human moderators are no longer able to differentiate them.
Internet users are often overwhelmed by advertising which does not correspond to their interests. Of course, targeting methods that allow you to show the target audience more relevant content exist, but it remains difficult to group people together for effective marketing. In order to achieve the ideal results, a personal approach is essential.
For this, AI-based solutions are being developed that can show or send personalized promotional messages to prospective customers. Almost every individual or filtered narrow group of people with specific interests will see only those adverts which pertain to their desires. Advertising will finally provoke the desired response and improve user experience, as well as bring more convertible traffic to advertisers.
A striking example of the use of artificial intelligence for personalized marketing is Amazon. It became popular because it registers and remembers customers’ interests: each viewed product, each purchase, and the place where the parcel is received. On the basis of this information, the retailer recommends relevant goods and suggests further items which were chosen by shoppers who have previously purchased the same things. Amazon’s recommendation system works with an AI engine called DSSTNE. Having spent years creating its own algorithms, the company posted an open-source library on GitHub in May 2016. You can download it to study and then order an enhanced version from us in accordance with your needs.
We would like to illustrate the capabilities in formulating advertisements with an example of a simple use case of artificial intelligence in the marketing of Chase Bank. Promotional texts for it are generated by a model from the startup Persado. From 2016 it was only utilised to generate slogans promoting bank cards and mortgages. In July 2019, having tested the model on new products, they assigned it the responsibility of writing texts relevant to other banking activities. According to the management, the machine devises more appealing adverts than marketing specialists. For example, users clicked on the headline “It’s true — You can unlock cash from the equity in your home” (machine’s version) twice as often as on “Access cash from the equity in your home” (people’s version).
AI can increase sales if you allow it to modify commercial offers. Let it set prices for products, taking into account users’ behaviour and their needs, in accordance with the availability of the product and the level of demand for it. This tactic will attract more new customers and incline them towards a purchase or a target action.
In turn, machine learning for business and marketing allows you to classify existing customers and anticipate their value to the company on the basis of past behaviour and current intentions. After training, such data will be relayed through identified signals from the customer base and converted into a unique message to those who can make more frequent and larger purchases.
For those who have reservations about the financial capabilities of machines, there is the case from the Peak platform. Its AI, CODI, helped a leading British retailer (the name cannot be disclosed) to optimise markdowns and maintain its profit margin when clearing warehouse stock. CODI collected and standardised data across the entire chain of price generation, presented projections for the level of demand to merchandisers, and recommended an ideal price range for particular goods. The team of merchandisers used its recommendations for 15% of the stock, and they helped optimise markdowns, increase productivity, and save time. Consequently, the retailer saved £2.4 million.
As a rule, PPC (pay-per-click) campaigns are either carried out by company employees or by advertising agencies, that is, by people. If you have launched campaigns on Google Ads or Facebook Ads, then you have already used the neural networks of these IT giants in order to achieve a low price in their automated auctions of contextual advertising with AI.
AI will help you to discover new ad channels, which are unknown to your competitors, and optimise older ones. With its help, you can test more ad platforms and optimise targeting. Facebook is currently doing just this with its ad delivery optimisation. This approach can also be applied to data from omnichannel PPC campaigns, if you use third-party or custom AI-based tools.
The marketing platform with self-learning AI from Albert Technologies has demonstrated itself well here. This startup became a pioneer, having shown the possibilities of its AI in July 2017, prior to when Oracle and Salesforce launched their equivalent services. Albert processes and analyses data about its audience and tactics, autonomously allocates its budget, and optimises advertising campaigns on paid search and social channels. Here are the results of recent cases where AI has helped to manage digital advertising and increased its effectiveness:
Artificial intelligence possesses a huge potential that businessmen, website owners, marketing professionals, content managers, SMM specialists, and webmasters can exploit. It helps to create, fill, and optimise websites, as well as facilitates promotion on social networks.
Today neural networks write readable news, reviews, and articles, drawing from a simple set of rules and formats. The texts generated by them are assumed to be written by humans. The analysis of data and authorial style depend on the settings that a particular company chooses so as to appeal to their audience more directly. For example, Associated Press uses neural networks for the automation of financial reports of surveyed companies, and USA Today uses them for the creation of short video clips.
Above, we mentioned OpenAI’s GPT-2. In May 2020, the organisation released a new, third version of the natural language processing algorithm. GPT-3 is now considered the most advanced generative language model. GPT-2 had: number of parameters used—1.5 billion, context—1024 tokens, dataset size—40 GB. GPT-3 has: number of parameters used—175 billion, context—2048 tokens, dataset size—570 GB.
Here are the popular services for a content generation which work on the basis of GPT-3 and promise to alleviate the work of marketing specialists, content managers, and copywriters:
They all require payment, as OpenAI provides commercial companies with access to its API at two basic rates:
For development enthusiasts, OpenAI also offers either 100 thousand tokens for free or a 3-month trial period. As a point of reference, 2 million tokens are roughly equal to 3000 pages of text.
Besides content for a website, neural networks can create unique posts for social networks. Subscribers to your network will engage with such posts without suspecting that they were written by a machine. For example, The Washington Post has long used its own AI, Heliograf, for the creation of news and posts on social networks. This machine reporter produces 900 articles a year.
Although AI is a long way from being able to create powerful websites from scratch, it is capable of improving user experience by intelligently personalizing two components of your website or web application:
Use AI in order to automate a large part of the personalization of your website. As a result, visitors will see the most relevant content, notifications, and offers, depending on their preferences.
AI will also be able to help you detect when the flow of data stops or when unexpected traffic arrives at the website. It is unlikely that your colleagues can check analysis every second, but AI can do this with ease. This information will help you maintain the smooth operation of the website and deal with emerging anomalies.
If you deploy artificial intelligence in advertising and sales, it will consult customers 24 hours a day, transferring them to employees only in special cases. Upon receiving a text or voice request through any channel, it will analyse it, comprehend it, and automatically respond to the customer. Usually, the support service implements a simpler system, while the sales department requires a smarter machine.
Intelligent chatbots (those which use AI rather than scripts) provide services to clients in many areas: from fashion and entertainment to health and insurance. Moreover, they interact with customers more amicably than human managers. This is so as chatbots are free from prejudice, stress, or fatigue.
Chatbots also have a deeper understanding of what clients want, since they have access to thousands of data points related to clients. Bots aggregate location-specific queries, identify patterns, and add recurring questions to the database. This raises their awareness of the interests of people who contact them.
Today, chatbots have become an essential tool in personalized marketing. They are already ubiquitously used by companies for handling each online visitor, and not only for the existing client base. It is easiest to implement a ready-made chatbot from a service, for example:
If their capabilities do not fit your requirements, you can create your own chatbot on such platforms as Botkit.ai, Botsify.com, Morph.ai, Wit.ai. If there are no programmers in your company, then we will develop a chatbot with artificial intelligence for you.
The cost of attracting new customers is higher than that of retaining old ones (sometimes up to 10 times greater). When the speed of attracting customers declines, it is necessary to increase the loyalty of existing ones. To do this, you need to run churn prediction.
AI is able to identify those who have become less active or who are preparing to leave you in favour of your competitors. With the help of machine learning, it is possible to collect data, build a predictive model, and then run it on customers. This information will let you know the extent of each customer’s dissatisfaction.
Predicting customer churn with the assistance of AI allows you to analyse omnichannel events and detect a fall in customer engagement. If the system identifies behaviour that indicates a loss of interest (for example, a reduction in use time), it will begin to send users related offers, push notifications, and emails in order to sustain their level of interest.
Here are some examples of functioning prediction services:
You may have noticed how well Google Photo has begun to recognize images and the people in them. In recent years, developers have advanced the capabilities of recognition software, providing a rate of accuracy that exceeds 99%. Such major companies as Amazon, Facebook, and Pinterest use artificial intelligence for image recognition in order to identify people and objects.
In advertising, machine learning and image recognition are useful for improving the synchronisation of online content and store visits. Retailers use video facial recognition software to identify regular customers and link these videos to their profiles. Thus, 59% of British fashion retailers install facial recognition software in their stores. And utilising AI for push notifications, they send visitors personalized greetings and discount offers.
Marketing has changed a great deal. Previously, it was product-oriented, but now it is customer-oriented. The development and implementation of artificial intelligence technology have played a key role in this. McKinsey predicts that the potential of AI in the marketing field amounts to $1.4–2.6 trillion.
Deloitte estimates that one in two companies already uses machine learning in marketing. Businessmen and marketing specialists rely on AI, and they plan to benefit further from its application. Therefore, it is not simply a matter of following trends, but it has become essential for survival in a competitive environment.
If you need to use AI in marketing, then Polygant is ready to develop programs and applications which are tailored to your requirements. To accurately estimate the cost and time of development, as well as to get more detailed information about our services, fill in a short form, and our specialists will contact you immediately.