Machine Learning for Marketing Agencies – Current and Future Applications

Ayn de Jesus

Ayn serves as AI Analyst at Emerj - covering artificial intelligence use-cases and trends across industries. She previously held various roles at Accenture.

Machine Learning for Marketing Agencies - Current and Future Applications

A collaborative report by Moore Stephens and WARC estimated the size of the martech industry around $34.3 billion dollars in 2018. It follows that AI would find its way into the marketing world. Marketing experts agree that AI will have a significant impact on the marketing world in the coming years. As of now, numerous companies claim to assist marketers in aspects of their roles from garnering insights from dashboards to automating spend on ad networks.

We researched the space with the intention of providing examples of companies that claim to offer AI solutions for marketers with the express goal of figuring out how those solutions actually use AI to meet KPIs such as increases in clicks, leads, and sales. As it turns out, martech AI companies are as a whole largely opaque with regards to the software they offer.

Most of the companies we found had a large button on their websites labeled “Request a Demo.” These companies were not transparent as to how their software works and what resources they might require, such as a team of in-house data scientists or a certain sum of historical customer data. We regret that we were unable to find this information, but this is unfortunately common in the world of B2B AI applications.

Additionally, many of the companies we cover in this article lack videos demonstrating their products and, in some cases, case studies reporting on whether their software generated success for business clients.

We caution businesses to be wary of companies that lack transparency with regards to their AI product. Before getting on the phone with these companies, it might be a good idea to read our guide to separating the AI hype from reality in order to know what to look for in a company claiming to do AI.

That said, below we outline AI martech companies that claim to be solutions for the following:

  • Lead generation
  • Segmentation
  • Copy/creative

Again, we caution readers to go into this article as skeptics and consider the lack of transparency in this space before dedicating time in their next quarterly meeting to AI in martech.

Lead Generation


Albert is an AI-powered digital marketing software that claims to help businesses generate new leads. The company claims its software sifts through a business’ backlog of their customers’ data, such as their geographic location and their on-site search history. Then, Albert searches for Internet users with similar profiles across the web and markets to them using the creative the business provides it.

The company also claims that Albert can determine which creative is performing the best and then pivot to using that creative across various channels. We were unable to find exactly how Albert does this.

Albert does not have a demonstration video available showing how its software works. The videos it does provide focus on marketing the software.

In 2016, Harley Davidson NYC used Albert to increase digital sales. According to the case study, Albert discovered “trending data,” likely related to motorcycle sales at the time, and used this in conjunction with previously collected customer data to digitally market to potential new customers. Again, the information regarding how exactly Albert did this is unavailable. Harley Davidson NYC attributed 40% of its motorcycle sales to Albert in its first six months using the software.

Albert’s company website also lists Cosabella, Natori, Dole and Gallery Furniture as past clients.

CTO Tomer Naveh holds an MS.c. in computer science from the Hebrew University of Jerusalem. Unfortunately, we could not find evidence that anyone working for Albert has a significant background in AI. The company uses the term “data researcher” to refer to its employees, but no employees listed on LinkedIn have the phrases “machine learning” or “artificial intelligence” in their titles.


PredictiveBid is a bidding software that claims to be able to market to potential new leads on Amazon and Google Ad Network. The actual bidding process involved marketers selecting each keyword and choosing the amount they are willing to pay each time a customer searches one of those keywords and clicks their company’s advertisement. PredictiveBid claims to automate this process.

As more people click on marketing creative, PredictiveBid claims that the software learns which attributes of the creative work best for generating the click. Was the click garnered because the ad targeted people in a certain geographic location? Did the ad’s time of day influence whether or not people clicked it? Or was it the keywords behind the ad?  PredictiveBid claims its software determines the likely driving factor behind the click and then applies this knowledge to future advertisements in order to drive up sales.

We could not find a demonstration video available for the software, nor any explicit case studies claiming success with the software. However, we can infer a “case study” (minus the statistics) based on how we understand the software to work.

If a coffee retailer wants to post ads on Google to target buyers in North America, the company can choose from hundreds of industry-specific keywords that Google Ads offers which describe the company. These keywords could perhaps be relevant to the customer’s location or the types of coffee the retailer sells.

Instead of manually bidding for each keyword on Google, the coffee retailer could use PredictiveBid’s software to determine a bid price for each keyword, a result of the software having been trained on billions of data points from Google ranging from searches to clicks.

PredictiveBid’s technology also has the capability to measure keyword or adgroup performance and to continuously segment the audiences it sets the coffee retailer’s ads to target as its database grows by finding patterns in customer behavior, such as which ads they click and when.

Meanwhile, a potential buyer in the U.S. can search for a coffee supplier in the coffee retailer’s location on Google and find the retailer’s ad.

We were afforded context to the process behind PredictiveBid in an interview between our CEO and PredictiveBid CEO Lior Tasman.

CTO Liel Villa holds an MS in Statistics from Tel Aviv University and was head of Algorithms for nearly 2 years at Jellyfish.

One AI company that claims to collect customer data from all over the web to generate new customer leads is, a search engine powered by machine learning and natural language processing (NLP). The company claims to use B2B customer data to create what it calls a relationship graph that connects B2B companies with companies who may be interested in their product or service.

Node worked under the radar for two years after its establishment and claims to have amassed over a half billion profiles of people and companies over that time. These profiles were gathered from various sources on the web, including websites, social posts, and LinkedIn profiles. Node claims that their system can update profiles in real time.

In the 4-minute video interview below with CNBC anchors, CEO Falon Fatemi explains that Node could suggest the right company or contact person to which a business should market.

Falon claims that Node can suggest how to communicate to that company’s representative when the sales or marketing professional connects with them for a sales call by revealing common connections between the marketers and the company rep on the other end. These connections could include the school both people attended or the shared companies at which they have worked.

We were unable to find information on how Node suggests these connections, but based on information gathered from various sources, we have inferred that the voice-activated search engine uses natural language processing technology when the user asks questions such as “What company will be interested in my product?” Node seems to then search its database and respond by suggesting a prospective business customer, the specific contact within that organization, and the reason behind the recommendation.

Falon is noted for having once been Google’s youngest employee, starting her tenure there at age 19. During that time. Louis Monier is Chief Scientist at Node and holds a PhD in mathematics and computer science from the Paris-Sud University.

To date, Node has received a total $23.3 of funding from investors, including Mark Cuban. The company has yet to release any case studies but claims to have worked with Zoomdata, Pagerduty, Full Contact, Information Builders, Conversica, Blue Jeans, among other clients, according to its website.



Optimove’s Chief Data Scientist Yohai Sabag says AI-driven segmentation could potentially help increase customer engagement by grouping individuals with similar behaviors and preferences and sending them personalized content that is more relevant and timely to them.

Optimove’s marketing optimization bot called Optibot uses machine learning algorithms to automate the segmentation process. The company claims that the software searches and collects customer data sourced from a business’ raw data and Optimove campaign response data. The software then finds patterns in that customer data, groups customers with similar attributes, and finally recommends customer segments.

The company claims the system also finds patterns in the data that can predict which current customers are likely to become inactive, suggesting those customers as another segment. This could allow marketers to implement steps to re-engage customers and prevent attrition.

In the 7-minute video demo below, CEO Pini Yakuel, at around 1:39, demonstrates Optibot’s features. He explains how customers become more responsive to campaigns as the machine learns about behavior changes. He also shows how A/B testing is set-up, and how the self-optimized solution mixes and matches possible campaigns.

Gaming company BetBright worked with Optimove to deliver specific email messages to more than 30 distinct player segments. BetBright claims the future values of sports bettors and casino players increased by 30% and 9% respectively after sending the targeted emails. Additionally, the attrition rates for sports bettors and casino players decreased by 7% and 10% respectively, and the company saw a 25% increase in the number of repeat depositors.

Adore Me, Spacebar Media, Sisal and Freshly are also listed on Optimove’s website as past clients.

Pini holds an MS in industrial engineering and management from Tel Aviv University. Yohai Sabag holds an MS in Information Technology from Ben Guiron University. Optimove has raised $20 million in its first round of fundraising.


Cortex, an AI-driven marketing automation tool created by Retention Science, claims to analyze behavioral, transactional, and demographic data to segment customers with similar attributes. The segments help clients create more personalized messages that could potentially move the customers to make purchases.

The software applies machine learning to forecast the likelihood customers will purchase or unsubscribe by finding patterns in their email activity, page visits, session durations, and browsing behavior. Although we searched for it, information on how the user interacts with the software, how the software works with that input, and its explicit output seem to be unavailable.

Dollar Shave Club wanted to reduce customer attrition by reactivating those who had canceled subscriptions. Retention Science helped the company plan a campaign by determining the email intervals and frequency and the personalized email messaging. The application helped the client execute the campaign by first sifting through the transactional, behavioral, and demographic data. The software then offered email templates, self-correcting subject lines, and automated email mechanisms. As a result, the company claims that the attrition rate decreased by 60% and the re-optin rate increased fivefold.

Aside from Dollar Shave, Retention Science also claims Target, Draper James, and Hollar as clients.

As Director of Data Science, Vedant Dhandhania leads projects related to building large-scale recommendation engines and retention statistics. He also oversees statistical evaluations and data visualization within the company. He holds an MS in electrical engineering and has previously worked at Apple.


Persado Pro Email

Some AI-powered companies claim to have applications that assist marketers in the generation of marketing creative. One such company, Persado Pro Email claims to offer an application that generates email subject lines using natural language processing that appeals to emotion.

CEO Alex Vratskides claims the system was trained by feeding it billions of online impressions. Finding patterns in these impressions, Persado claims to have found that “emotions drive 60 to 70% of consumer response.” This could, in theory, increase email open rate, possibly leading to more conversions.

To start using the application, users must first input the name of a topic or “experiment,” then select an industry and email type (promotional, welcome, cart abandonment, etc). The 3-minute video below demonstrates the process:

The user then chooses specific attributes to include in the subject lines that the software will generate, such as elements of scarcity, urgency, or discount. Once these details are submitted, the system generates 15 subject lines, each with a different focus. The company claims the system will determine the voice and tone of the subject lines by taking into account the attributes specified.

The company then claims their software determines which subject lines would resonate most with customers in a way that would compel them to open the email, as shown in the 2-minute video below

The company also claims style filters also allow the user to manually choose the language to match the brand’s voice and messaging. Other basic style features include the ability to add or remove symbols, adjust text case, and edit the proposed subject lines similar to a document.

Thus far, the company has generated $66 million in funding and claims retail, finance, and telecommunications companies as its clients. In a case study with an unnamed department store and Fortune 100 eCommerce site, Persado claims to have implemented a marketing campaign that involved using language that reflected emotions of achievement, exclusivity, and safety.

As a result, Persado claims to have helped the client reach a 57% increase in click-through rates compared with their previous rates. Worthy of note, however, is that Persado did not name their client despite the fact that they are supposedly a Fortune 100 eCommerce site. We suggest readers approach the case study with caution.

That said, Persado claims that Dell, Sears, Staples, and Starhub are among its clients.

Persado’s Chief Data Scientist Panagiotis Angelopoulos oversees the research and development of analytics and machine learning. Prior to Persado, he was head of statistics at Upstream Systems. He earned his PhD in statistics from the National Technical University of Athens.


Narrative Science also states that its Quill software can generate copy using the data businesses feed into it.

According to Chief Scientist Kristian Hammond, when presented with charts, interpreting the data is traditionally left to the humans. Quill’s job is to analyze and interpret those charts and provide the story behind the numbers in narrative form. Unfortunately, information regarding how Quill carries through on this process seems to be unavailable, nor have they put out any demonstration video.

Narrative Science claims Dominion Dealer Solutions, an auto dealership, worked with the company to generate consumer-ready automobile descriptions in real-time for the dealership’s website. Using data such as car make, model, mileage, and features, as well as marketing and customer information, Quill proceeded to create individual vehicle stories and descriptions for each car.

Dominion Dealer Solutions claims to have achieved a 20% lift in page views across new and used vehicles as a result of using Quill, a 50% lift in page views for used vehicles alone, and more than $1M in new revenue.

Hammond earned his PhD in computer science from Yale.


Automated Insights, acquired by Vista Equity Partners, developed Wordsmith, another NLP platform that claims to transform data into narrative.

According to Automated Insights, Wordsmith’s open platform allows it to be integrated with data visualization or business intelligence software such as Tableau, MicroStrategy, TIBCO Spotfire, and Qlik. The 3-minute video below demonstrates how Wordsmith generates written explanations of user data when combined with Tableau:

Automated Insights claims Wordsmith translates data on a dashboard into narrative text that business leaders can then read to garner insight into that data without having to make sense of the dashboard interface. This could speed up the time it takes to review sales and marketing data. As with other companies, Automated Insights does not make it clear what the user inputs into the system and how the system processes that input to achieve the narrative text.

One client, Vivint, claims to have used Wordsmith to create copy for 12,000 pages on its website. The case study claims that Wordsmith’s NLP system interpreted the client’s data and created human-sounding, personalized content.

Vivint’s marketing team calculated that it would take its writers more than two years to complete the project at a huge cost. With Wordsmith, Vivint claims the team was able to produce content for more than 30,000 pages within three months.

According to Vivint, adding personalized content helped the company’s SEO efforts and pushed the client’s pages to the top one or two results in search engines. Additionally, monthly sales that were a result of online searches increased fivefold compared with the same period year-on-year, the client said.

The company also lists, Nvidia, Yahoo Sports, and Anterra Wines Canada as clients.

Nicholas Haynes is the Director of Data Science at Automated Insights. He holds separate MS degrees in physics and applied mathematics.

Concluding Thoughts

As of right now, companies offering segmentation seem the most viable for businesses of various sizes while also specifically selling to marketing agencies. Those companies which claim to make suggestions for or seemingly outright automate the digital marketing process are both likely to be the most expensive and the most experimental. That said, Albert employs 226 people according to their LinkedIn profile.

Before deciding where to spend their budget, we suggest businesses shop around for AI martech companies. A lack of transparency usually indicates either a difficult integration process that could cost businesses much more than just the sticker price of the software or a relatively untested system that may not yet have had enough time to garner results for other businesses.

At this time, we believe businesses should remain skeptical of AI martech considering the results of the research that went into writing this article. It’s likely that at the time of writing, AI martech is too nascent not to require a large amount of human involvement in running the systems which purport to automate lead generation and ad network bidding.


Header Image Credit:

Stay Ahead of the AI Curve

Discover the critical AI trends and applications that separate winners from losers in the future of business.

Sign up for the 'AI Advantage' newsletter: