Magnus Unemyr

As AI solutions not only analyze and understand text, but generate human-sounding text too, there are many use-cases in content creation. Let’s start with AI-assisted manual content creation.

A great use of AI in content creation is advanced grammar checking. Since I am Swedish, English is my second language.

While Microsoft Word or Apple Pages have built-in spell checkers with grammar control, they aren’t best-in-class and I could do with some more help.

I have thus used a stand-alone grammar checker for years. It is called Grammarly and uses machine learning to improve its grammar checking over time. In fact, this area is so hot that they raised over $100 million USD in venture capital in 2017 to improve their product even further.

Grammarly is an AI-based natural language processing tool (NLP) that processes and analyses the copy you have written and recommends improvements.

Another interesting area is natural language generation (NLG). NLG tools doesn’t read and analyze your content, they write it.

NLG tools doesn’t read and analyze your content, they write it. While it may sound like science fiction, already several tools can do this.

One of them is Wordsmith from Automated Insights. Their tool converts structured data (say, information in an Excel spreadsheet or a database) into natural language text. They claim to generate 1.5 billion pieces of narrative content each year for customers like Microsoft, Cisco, the Associated Press, Yahoo, PwC, and more.

Wordsmith can be used to write product descriptions in e-commerce, property listings for real estate agencies, stock market portfolio descriptions in the financial industry, and similar documents for practically any other industry.

It can also be used to generate data-driven and highly personalized email copy. In effect, Wordsmith can generate unique copy from the inputted data. Think robot journalism here.  

If you are an e-commerce company and want to upload 1250 product descriptions to your web shop, you can feed the raw product data from an Excel spreadsheet into Wordsmith. The software will then auto-generate 1250 engaging product descriptions in a matter of seconds.

The Wordsmith team has extended the usability of their writing robot with an API (an application programming interface), thus enabling external software solutions to integrate with it to produce fully automated text generation solutions for your company’s needs.

Narrative Science is a similar company, with an NLG product called Quill. It too transforms raw structured data into natural language narratives using artificial intelligence and machine learning.

Their NLG engine is designed to produce human-like, intuitive language. It automatically utilizes variable word choice, understands when it’s best to combine sentences, and identifies the appropriate pronoun usage, verb tense, and grammar to produce fluid, conversational language that is easily consumable.

This is how Aimee Rowland, Director of Product Marketing at Narrative Science, explains their product:

Our technology not only utilizes NLG to generate natural language explanations, but it also utilizes a range of analytical, reasoning, and learning capabilities to identify the most interesting and important insights, all with domain-specific knowledge so that it understands the context behind the data.

Their technology highlights the most interesting and important facts from data, providing relevant insights for each intended audience.

It does this by performing context-aware analysis focused on the core of what needs to be communicated. After all, it is not only about generating text, it is also important that it covers the right information.

Both Automated Insights and Narrative Science offer report generators that turn Google Analytics web usage data into insightful narrative reports. This can be a great aid to marketers who need to write management reports.

The news media is also turning to AI and NLG for automatic generation of articles.

The Washington Post, for example, uses their own in-house AI robot for article creation. It is called Heliograph and is said to have produced 850 articles in the last year, including many relating to sports.

Many local newspapers, even relatively small ones, already use robot journalism to turn numeric data into articles. Short articles on local sports results are prime examples here, as are weather reports, traffic information, and police blotters.

Now, content marketing and content production can be automated at scale.

Now, content marketing and content production can be automated at scale.

Don’t see it as a replacement of your copywriter, but an opportunity to automate parts of the content production efforts, in particular areas where massive amounts of data just couldn’t be converted into natural language text due to the time and cost required.

Add to this the possibilities for hyper-personalization, where each customer gets different content based on the data you have on them. 

Other interesting tools include ArticleForge, Articoolo and AI-Writer. With these, you simply type a phrase in an edit field – for example, “car insurance”. The tools then harvest the internet for relevant information, and return an auto-generated article on that topic, of your desired word length.

At least one of these tools can even auto-generate hundreds of different variations in one go, enabling you to publish many similar articles that use completely different text.

While the results may vary with different tools, they are getting better each day and can act as a quick-start that you refine further. Beware of copyright infringements, though, if you use phrases from other people’s websites.

Did you know AI tools can auto-generate videos? They can, at least semi-automatically. I’ve used Lumen5 for a while and I find it simple to use. This innovative tool aims to automate the process of creating short videos, in particular for promoting blog posts on social media using video clips of a minute or two.

Feed it with the URL of the blog post, and it finds what it thinks is the most relevant captions to use as headings.

It then lay out the headings on scenes (similar to PowerPoint pages) and uses AI to find free images and video clips on the Internet that relate to the keywords in the selected headings. At the end, it adds animated scene transitions, allows you to select music, and exports a video into standard file formats.

You can transform text into video in minutes, even if you do some manual editing of the preselected options. This is pretty amazing and a clear sign of where we are heading.

I can’t wait to see what tools like this can do in five or ten years.

Are you interested in marketing automation, artificial intelligence, or chatbot development? Contact us for more information!

Many companies are now extending their marketing automation stack by adding AI technologies to predict churn, in order to prevent loosing customers by using customer retention initiatives.

Most businesses suffer from churn – that is, customers who defect and stop purchasing your products or services.

Most businesses suffer from churn – that is, customers who defect and stop purchasing your products or services.

This is a major problem, as conventional wisdom commonly claims it is about five or six times as expensive to recruit a new customer compared to keeping an existing one. If not more.

Existing customers are also more likely to buy additional products from you. Understanding why customers stop purchasing and assessing the risks are key aspects of a data-driven retention strategy.

Churn prediction may be particularly important to companies offering subscription plans, such as mobile phone operators or cloud software vendors. To them, it is of utmost importance to keep the subscription plans active as long as possible.

Detecting unhappy customers early gives you the opportunity to offer incentives to stay.

Detecting unhappy customers early gives you the opportunity to offer incentives to stay.

In effect, if you know in advance which customers are likely going to abandon you, you can reach out to them ahead of time and try to prevent this from happening.

With AI-driven predictive marketing, it is often possible to detect what factors and patterns indicate a customer is likely to leave.

Dependent on the customer value, AI-driven churn prediction algorithms could trigger the marketing automation system to send a sequence of nurturing emails, possibly with a discount offer to re-engage the customer.

This might be a suitable approach for low-value customers and in the B2C market, but for your best customers in a B2B market, the predictive marketing system might instead be configured to notify a sales representative to make a personal call to the customer and try to rectify the situation before it happens.

Churn can be detected by analyzing factors like:

• Demographic data
• Digital footprint (website/email behavior, etc.)
• Purchase history and payment patterns
• Social media sentiment analysis
• Product usage patterns, if measurable
• Customer support statistics

To detect churn, we simply want to know how likely each customer is to defect.

To detect churn, we simply want to know how likely each customer is to defect.

Classification algorithms like decision trees can be used for this, as they can classify a customer into one of two categories, based on their behavior and that of other customers.

In our case, we interpret the two classes as “likely to defect” and “not likely to defect”. We can then act upon these insights accordingly to keep the customer longer.

Many commercial vendors offer churn prediction tools, for example Optimove and DeepSense. But there are many more. Omni-channel customer data platforms (CDPs) benefit from this functionality as well.

Part of predicting customer churn is to learn how to act upon such insights.

Part of predicting customer churn is to learn how to act upon such insights.

Customer retention is the mix of activities you can take to reduce churn and retain as many customers as you can through customer care, purchase incentives, loyalty programs, or other means.

Retention is an ongoing process throughout the lifetime of the customer relationship and should not be bolted on at the end as a solution to an otherwise failing customer experience.

Having said that, retention and loyalty programs may help keep your customers, or reactivate them once they have defected.

Are you interested in marketing automation, artificial intelligence, or chatbot development? Contact us for more information!

AI-powered marketing automation is coming into almost every part of our industry. Content strategy is no exception, and so in this blog post, I will outline what is already possible to do with commercial tools.

But let’s start with a quick overview of the topic.

Content strategy is about measuring what content provides a meaningful ROI and marketing value, enabling content managers to double-down on producing content that generates leads and customers, and scale back on content that doesn’t provide good results.

Content strategy is about measuring what content provides a meaningful ROI and marketing value, enabling content managers to double-down on producing content that generates leads and customers, and scale back on content that doesn’t provide good results.

Search engines like Google are now focused on providing search results of high quality. Their search algorithms are becoming good at finding high quality content, and it is becoming increasingly hard to trick them (this is called black hat SEO).

To this end, content strategy tools have started to emerge.

For example, HubSpot measures domain and page authority (these are SEO metrics) and use artificial intelligence to recommend topics on which to produce content. They promote the concept of content clusters, where supporting content (often a number of blog posts) links to a piece of long-form pillar content to help push its authority in the eyes of search engines.

AI and machine learning can be used to help improve the quality of the content as well, either for better SEO ranking or to provide readers with exactly the content they are looking for (and are thus more likely to engage with).

Software tools can use AI to model a topic and build content briefs that show exactly how to write to systematically cover a topic in the best way.

For example, what are the relevant concepts to mention in the content? What are the questions to answer? How should content be written to make it rank in search engines?

 One company offering a solution for this is MarketMuse. Their tool enables the creation of better articles, increases topic coverage across a website, observes and analyzes the competitive landscape, and discovers content improvement priorities.

In this blog post, I discussed their tool with Aki Balogh, the company’s co-founder and CEO. He describes his product in this way:

“By using the MarketMuse methodology, marketers can build more comprehensive content and establish authority on topics relevant to their business, resulting in enhanced organic website traffic, improved thought leadership and improved rates of customer conversion.” 

marketers can build more comprehensive content and establish authority on topics relevant to their business, resulting in enhanced organic website traffic, improved thought leadership and improved rates of customer conversion.

MarketMuse is interesting because it scours massive amounts of web content to look for the coverage around a focus topic. It builds a topic model that helps a marketer write like a subject matter expert. This enables the creation of high-quality content that ranks in search engines.

In effect, the platform helps connect the dots between what the content marketers want to write about and what the users want to know. This helps improve engagement, increases search rankings, and develops thought leadership.

Large corporations with many employees can have different problems, including adhering to a consistent voice or brand guidelines. By now, it may come as no surprise to learn that AI can gather brand and audience goals and help ensure your content complies with those goals.

After all, most corporations want to give a consistent message.

Such tools use AI for content scoring and brand compliance checking, enabling all different pieces of content to be coherent across offices or content producers.

Other content strategy tools help with auditing your content, understanding how people behave in relation to your content, benchmarking your competitors’ content, and generating data-driven content ideas.

Some content strategy tools optimize what and how you should write, as well as where and when it should be published for the best return on investment. Several tools on the market address such problems, for example Acrolinx, BrightEdge, Concured, and Atomic Reach.

Are you interested in marketing automation, artificial intelligence, or chatbot development? Contact us for more information!