AI cheat sheet: The basics of Einstein, Albert, Watson and friends

Artificial Intelligence is unparalleled at planning campaigns.
Artificial Intelligence is overrated and cannot match human intuition.
Artificial Intelligence makes teams work more effectively.
Artificlal Intelligence will lose us all our jobs.

The constant avalanche of opinions, theories, and suppositions as to the future of AI in marketing is truly endless. Yet it is undeniable that its adoption by the industry is increasing as much as its evolution is unceasing.

However, this rapid expansion has led to a dizzying plethora of platforms that can be painfully difficult to keep track of. To help, Campaign has put together an in-depth guide to the biggest kids in the machine-learning playground and what is coming down the line.

Assisting with their opinions on the matter are:

• Or Shani, CEO of Albert
• Eric Thain, managing director and head of digital for China for Lewis Global Communications
• Zaheer Nooruddin, head of digital transformation at Shiseido

THE TOP 5

Meet the most talked-about marketing-related AIs.

1. Einstein

Nooruddin describes Einstein as, “The AI platform within Salesforce CRM, developing the promise of algorithmic optimization of Salesforce’s Marketing Cloud.”

As an artificial intelligence, Einstein isn’t a self-sufficient powerhouse in itself. Its potency derives from acting as a string that ties together the series of products that make up Salesforce’s ever-growing Customer Success Platform.

As Thain explains, “Einstein isn’t a product so much as a set of intelligence functionality that underlies the entire Salesforce platform, and while the types of functionality that it’s enabling now are somewhat limited, the idea is to provide a base on top of which the company can continue to add new capabilities into the future.”

Since its announcement in 2016, Salesforce has launched two main components for Einstein: Einstein Vision (which uses image recognition for functions like visual search, brand detection, and product identification), and Einstein Language (which uses natural language processing to measure elements such as customer intent and sentiment).

These tools work in tandem with the Salesforce product range (such as the Sales and Marketing clouds) to provide analysis and strategic assistance based on supplied criteria. Thanks to Salesforce's partnership with IBM, findings can be cross-referenced with the IBM Watson system (see below) for even greater targeting.

“Einstein is gaining a lot of traction as part of Salesforce solution," Thain says. "Salesforce has been around as an important data ingesting and CRM partner for many brands so the addition of Einstein will be beneficial to that offering. It can function very well as a step to power up the optimization of such data throughout the sales funnel in a meaningful way.”

Breaking it down to basics, Salesforce has set on gathering several different AI systems and providing the capability to have them work together, with Einstein acting somewhat as a go-between. Some of those AI systems are homegrown, and others have been made available to Salesforce through licensing and partnerships.

The biggest win touted by Salesforce is that of US Bank, which doubled its wealth customer conversion rate, reportedly through using the Einstein-based Salesforce system. Another would be Shazam, which had a 752% ROI on using Einstein, saving the time of its analysts by 15%.

Nooruddin opines that Einstein, “is primarily, for now, being developed as the brain behind Salesforce’s Marketing and CRM suites, to drive better results with the usage of these tools, and as a way to maintain Salesforce’s market leadership in the space of CRM personalization-based analytics and predictive marketing campaign automation.”

The formation of a single superbrain AI hasn’t been the ultimate development goal; building up the versatility of the Salesforce platform is. And as more functions are added to the collective, the attractiveness of the platform to marketing teams will only grow with it.

2. Albert

While other AI platforms provide universal features for a wide range of business types, Albert has been specifically crafted for the marketing industry. Furthermore, Albert doesn’t act as a single tool to assist brand and agency staff in any single aspect of their roles, as much as it takes over their workload entirely.

“Albert is a fully autonomous digital marketer," Shani tells Campaign. "He executes, optimizes and analyzes cross-channel, cross-device, digital campaigns and makes ongoing adjustments to reach the marketer's goal, in real-time, 24/7.”

Created in 2010, the platform is intended to handle the entirety of individual tasks that go into the running of a campaign. Once it has been fed a sufficient amount of data and given proper access, Albert can plan, enact, and update a campaign completely independently.

“Because he is autonomous, he takes decisions on his own based on his learnings, operating at a pace and scale that is not humanly possible," Shani explains. "He is essentially an off-the-shelf application for full-funnel, full lifecycle, digital marketing campaign execution and analysis.” 

Albert’s multitasking skills remove the need for long hours toiling on research, analysis, and social media planning, but as impressive as that sounds, some in the industry are understandably wary. One chief concern is that that they’re being sold their own career tombstones. Promises of massively reduced workloads for teams can easily be interpreted as code for equally massive layoffs.

Employment concerns aside, Albert is useful for a cross-functional integration and optimization throughout a business functional processes and departments, Lewis's Thain says. "It executes complicated, multi-channel and multi-step business processes, from start to finish, without ongoing input from human colleagues."

On issue for Albert regardless of its capabilities: You’d be hard pressed to find a project lead who’d feel completely comfortable leaving their livelihood entirely in the hands of a piece of software. The chief fear being that without the oversight of human intuition it could fall prey to the same ad fraud bots that plague automated systems.

Shani has a response: “Albert was built with a mission—to reach a brand’s KPI(s)," he says. "In doing so, he aims to avoid fraud. He was built with custom ad fraud capabilities baked into him. He is able to calculate thousands of variables in real-time in order to swiftly identify and weed out fraudulent traffic sources as they appear along the campaign journey, without need for regular human supervision. He quickly processes and stores any and all information regarding sources, applying what he has learned to future purchasing decisions.  This helps to ensure current and future campaign safety, while also ensuring investment goes to bot-free traffic sources.”

Albert can also be taught to recognise context and be given “blacklists” of topics to avoid placing ads next to potentially dicey content.

Until recently, all this came at a rather hefty cost. Anecdotes from the marketing chattersphere put Albert’s asking rate at a sweat-inducing 15 to 22% of a company’s media spend. The team has since switched to a full SaaS model based on flat monthly fees, with payment tiers starting at $5000 per month.

“We changed our pricing model in the fall of 2017 as we realized the percentage of media model no longer serves the industry well," Shani says. "It also does nothing to convey the real value cross-channel AI solutions bring to the table.”

In spite of the price tag, several brands (including Harley Davidson, Evisu, Cosabella and Dole) have embraced Albert, crediting it for generating revenue that easily offset the price.

In one such case, Australian experience retailer Red Balloon testified that Albert was able to identify underutilized keywords and make recommendations that reduced the brand’s acquisition costs by more than 25%, contributing to a total cost savings of 40%.

“Our point of view,” says Shani, “is that marketers need better orchestration tools that condense the time it takes to launch, execute, analyze and ultimately, react/act again on the insights and recommendations from prior campaign learnings.”

3. Watson

IBM’s Watson has come a long way since its origins in the mid 00s. Originally developed as a spiritual successor to chess master Deep Blue, the Watson supercomputer was built as a passion project to prove an AI with natural language and learning abilities could beat humans at Jeopardy. Which it did, in 2011.

The system now drives myriad non-gameshow-related functions for clients in sectors such as healthcare, fashion design and medical research. it also drives the Watson Marketing platform.

“Watson offers algorithms for a variety of industries that can be leveraged to build custom solutions, such as wading through mass amounts of data for marketers so they can make more informed business decisions,” Shani explains.

In essence, a Watson system can be designed to do whatever you have the inclination and planning to make it do. From campaign automation and media planning to handling EDM distribution.

As Nooruddin puts it, “Think of the Watson platform as more of an AI application toolkit that allows businesses and organizations to build their own applications on top of. These could be applications for marketing, but they could just as well be other business applications built with artificial intelligence technology.”

However, Watson Marketing has been provided with a pre-built toolset for essential tasks:

• Customer-journey pattern analysis.
• Cognitive analytics and insights, which can create user profiles to log and predict behaviour.
• Website personalisation, which tailors experience based on existing behaviours.
• Cognitive tagging of visual content via image recognition, which saves time for campaign creators searching for assets.

Watson is even linked up to IBM’s The Weather Company, which means it can plan and execute customer engagement based on localised weather conditions. So, in theory, a clothing brand could push gloves and jumpers during a cold snap, or sunglasses and shorts during a heatwave.

"Watson’s AI solution is very robust and in my opinion, one of the most versatile and effective through the line," Thain says. "It can offer end-to-end AI solutions, and coupled with its effective business integration services, it is able to produce meaningful and real results to companies.”

Happy customers include eyewear manufacturer Luxottica, which stated that use of Watson Marketing decreased operating costs for its marketing team by 50% while causing a 500% increase in its campaigning. ING bank found that Watson tripled customer responses while decreasing costs by 35%.

“Applications built using Watson platform can then be integrated into business systems to unleash the potential of AI in business," says Nooruddin. "Or, Watson can be used, more simply, to build applications as useful as customer-service chatbots.” 

4. Vidora

The adage goes that past behaviour is the best indicator of future behaviour. Vidora leans heavily on that line of marketing wisdom, with its AI (especially with its Cortex system) charting previous user activity so as to predict a consumer’s activity.

Nooruddin explains that “Vidora deploys machine learning (ML) to ingest and process data within a business to automate and optimize business processes and provide advanced reporting in real-time back to the business.”

Vidora’s big selling point is in its potential to boost user retention by accurately predicting the lifetime value of a client’s users. Each user is assigned a unique profile, as the system meticulously logs their actions online. Data can be highly customized as to the client's needs. Vidora is then able to predict the odds of a customer returning to a brand’s site and/or services. As part of this, it would also be able to calculate the positive or negative reaction a user would have to any change in product or services. It also has tools to better customer engagement.

“While it certainly could be useful for any business, the most obvious use cases are in the heavy Industries, CPG and manufacturing sectors," Nooruddin says. "So, very useful to optimize various parts of the B2B business process by deploying artificial intelligence to identify efficiencies and drive better outcomes from existing processes.”

One of the company's success stories is Boing Boing, an online publisher. The site required better methods to hook readers onto articles and increase the length of time spent on the site. Vidora’s AI was able to choose the best headlines to attract each user, which improved clickthrough rates by 170%. While further onsite personalization, engagement increased by an impressive 250%.

This AI may seem less ambitious than some of the other entries on the list, but the simplicity and focus of a screwdriver may appeal to some brands and agencies over the more daunting Swiss-army knife approach that other AI systems are pitching.

As Thain puts it,  “Vidora’s AI proposition is simplifying AI adoption and integration to users and businesses. Its system automation functions probably takes precedence over its AI capabilities which may leave much to be seen.”

5. Marcel

Since Publicis cannonballed the pool at Cannes last year with its announcement that it would abstain from awards and invest that money (a rumoured $2 million) on an AI system called Marcel instead, information has been incredibly scant. Coming seemingly out of nowhere, the move attracted equal amounts of shock, ire, and confusion from members of the industry.

Shiseido's Nooruddin is far from jumping on the naysayer bandwagon though, positing that Publicis CEO Arthur Sadoun has been pragmatic to publically, “Throw down the gauntlet” with Marcel.

“It’s great to see that the advertising industry is addressing and investing in AI," Nooruddin says. "Many acknowledge that AI will severely disrupt the creative industry, i.e. advertising and media agencies in the years to come. Media planning and buying will no longer require human intervention and the process of planning, buying and optimization will be fully automated, and at scale. This poses an existential threat to the advertising and the media agency business model.”

During a Twitter AMA session, Saduon faced down his interrogators by claiming the company is, “...investing in a game-changing tool that will multiply creative opportunities and give anyone a chance to stand on the global scene.”

That’s all well and good, but no one has a solid idea what kind of beast Marcel will be. Sorting through the crumbs of information available has led to a consensus that the AI is expected to perform as some kind of ultimate worldwide team management tool and matchmaking service across the entire Publicis network.

Project leads will be able to assemble teams based on their skills, geographical location, knowledge base and experience. One could compare it to a dating service, where instead of comparing hobbies and turn ons, you’re choosing matches based on criteria like their grasp of the Malaysian baby formula market.

This deduction is backed up by the few rare public comments Saduon has made on the matter, including one Tweet stating “Everyone will have a rich profile with their skills, experience, super powers, passions. Marcel will find projects that match.”

Thain currently sits among the skeptics on the matter, believing Marcel is, “Not even an AI solution yet. At this stage, this is more a marketing campaign for Publicis rather than playing any form of business solution (let alone AI) to its clients. Publicis is only now engaging Microsoft to develop Marcel. What it will be remains to be seen.”

Indeed, over six months after the initial announcement came the surprise news in January that Marcel’s development would be a joint venture with Microsoft. A sudden turn of events that left many questioning if Publicis is scrambling to cope with a rash decision. The company’s official statement is that partnering up with a tech firm was “always planned.”

Nooruddin continues to be positive, believing that this move is only the start of agencies producing homebrewed AI remedies: “As to what Marcel will actually do…I’m not too sure yet," he says. "But it’s an interesting development nonetheless, and no doubt WPP, IPG, and Omnicom will soon have their own versions announced.”

HONOURABLE MENTIONS

Lewis's Thain breaks down a few AIs he believes the industry should be talking about more.

Trapica: “Trapica is a marketing automation tool powered by AI that is similar to Albert. It has also found much success with the marketing campaigns for their clients, via autonomously finding the brands’ most valuable audience and automatically targets it. It does this by incorporating machine-learning and predictive analytics over the brands’ customer data points to understand the consumers, classify them in real-time and targets them with the most relevant campaigns.”

Cloverleaf: "Cloverleaf makes AI shelves that can gather insights about what shoppers are doing in front of the shelf itself, engage them and convert the sale. The artificial intelligence-powered digital signs slide over retailers' shelves to give marketers insights about who is standing in front of their store displays. Based on the demographics and sentiment that the sensors pick up, marketers can then test different campaigns and swap out creative on the fly. The result is precision marketing at moment of purchase which is very effective for sales conversion."

YiMian: "AI is progressing very quickly in China and many of the AI companies are offering solutions that are making big differences to brands’ bottom lines. Chinese AI driven Data analytics company YiMian helps business unlock the potential of their data by processing internal & external, multipoints & unstructured data across the whole business such as sales, operations, marketing, social media, e-commerce etc. By doing this, it connects the dots & deliver actionable consumer insights to brand that drives bottom lines ROI."

Tencent: "Chinese tech giants are also making big strides in terms of the AI race. In the marketing front, all are offering advanced AI marketing solutions to brands to better target consumers. Tencent and other Chinese internet companies like Alibaba are the world's biggest generators of data. If AI is an engine, data is the fuel. The most important denominator of AI is data and this gives them a unique head start to capitalize on producing efficient AI solutions. Tencent has big plans in AI, with their AI ambition dubbed ‘Let AI be everywhere’. With their focus & stronghold over media, gaming, music, sports and content; Tencent has many AI solutions such as Wechat AI’s voice recognition, translation & also allowing brands to build chatbots on the platform."

Taobao: “Taobao has been using AI for a couple of years now with much success. It uses AI to create personalized shopping experiences for consumers and targeted marketing for brands, through AI-powered features such as smart product search and recommendation, customer-service chatbots, image matching and personalized pages. For example, over 60 billion personalized shopping pages were generated on Taobao during Double 11, in its run to generate RMB168 billion of sales for Alibaba in a single day.”

THE ROAD TO WIDER ADOPTION

Even if there are scores of data and multiple case studies that praise marketing AIs to the highest heavens, there are still lingering doubts throughout the industry.

Some of it comes from a lack of information or understanding, but a large amount stems from fear of the perceived high costs and structural overhaul that come with a company embracing the technology.

Nooruddin believes that is due to change quite soon. “Once more use cases become publicly available, which remains a challenge due to the complexity and confidentiality of development that’s often involved," he says. "I predict that we are still two to three years away from wider adoption, although the pace of adoption could change quickly, depending on the market, socioeconomic factors, as well as how quickly these AI systems can prove themselves as truly viable alternatives to current business, marketing and media processes.”

Unfortunately, the inherently dense technical nature of the processes that go into making these systems work is also what makes them unappealing to managers and team leaders who want to know what they’re paying for.

And the lack of understanding is a two-way street, with developers not always having a strong enough grasp of the company they’re trying to pitch their tech to.

“Many AI companies push their solution before the product is really ready for adoption," Thain says. "They need to make sure that the technology is mature enough and can deliver real results to businesses. Failure will leave companies disappointed and wary, and the AI company red-faced.” AI companies must strive to understand how a brand functions, the challenges it faces, its customers and its sales funnel in order to be able to contribute in a meaningful way, he adds.

What is seemingly blocking true AI platforms from being omnipresent in the running of every agency is that lack of intel. Which, in a catch-22 fashion, is hard to come by with the current limitations on the number of companies using these products and being willing to talk about it. However, it is obviously inevitable that this will change as the mountain of evidence as to its capabilities grows.

ANY FINAL THOUGHTS?

“It’s still just the beginning. It’s important for businesses to invest time and money in these platforms as they can be applied to their own businesses. Forget the noise and focus on the potential for positive mid- to long-term impact.”
—Nooruddin

“Marketers need better orchestration tools that condense the time it takes to launch, execute, analyze and ultimately, react/act again on the insights and recommendations from prior campaign learnings. Introducing an AI 'layer' into a particular piece of the marketer’s campaign execution journey is one way to do it, but that doesn’t solve the problem of time-to-market/time-to-respond.”
—Shani

“At its core, all are automation and optimization tools. [Such tools are] able to release the potential of data which many brands have not been able to tap into due to the abundance of it, the cross-functional disintegration of it as well as the lack of clean or meaningful data.”
—Thain