Hitmetrix - User behavior analytics & recording

The Dreamforce AI Dream

The Zen monks of Plum Village monastery had staked out a small space on the green lawn between the hulking bunkers of Moscone North and South, the expo halls at the epicenter of Dreamforce ’16. They were steering cross-legged peace-seekers down the path to mindfulness. As I passed by, I liked to imagine that the meditating attendees were seeking solace after three days of trying to wrap their heads around Artificial Intelligence—overwhelmingly the topic of the week.

And yet Salesforce had made it all look such fun. With a touch of movie magic, the brands two co-founders Marc Benioff and Parker Harris could be seen on the keynote stage chatting with a cartoon representation of the German genius. A human (I presume) dressed in an Einstein costume cavorted about the Dreamforce site. This was AI for everyone, democratized AI, designed to make the lives of all Salesforce users easier. 

If only I didn’t have a phrase used by Andres Reiner, CEO and President of PROS, stuck in my head: “The devil is in the details.”

AI for Everyone

Reiner, along with PROS’ CMO Patrick Schneidau, were just two of the people I sat down with at Dreamforce last week to try to get some perspective on what Salesforce Einstein actually means.  I also met Mayur Anadkat, VP of Product Marketing of Five9, the cloud contact center software vendor; Kraig Swensrud, former Salesforce CMO and now CMO of email marketing solutions vendor Campaign Monitor; Corinne Sklar, CMO of IBM’s Bluewolf; Shashi Upadhyay of AI-driven predictive analytics brand Lattice Engines; and Michael Plante, VP of Demand Gen at AI-based lead management platform InsideSales. I also spoke by phone with Sam Boonin, VP of Product Strategy with customer service and support platform Zendesk.

That’s some brains trust—and to cut to the chase, there was a clear consensus that Salesforce’s big AI move is exciting, daring, and an overall plus for the marketing tech space. And yet there were the details, and the devil was still in them.

Genius at Your Fingertips

Movie buffs would immediately have associated the Benioff-Harris main stage performance with Gene Kelly dancing with a cartoon Jerry the Mouse in the musical “Anchors Aweigh.” But at some point, with AI, you have to get serious. Modern artificial intelligence began not with Einstein, but with the work of a British genius, Alan Turing, who proposed some seventy years ago that a computational machine could, in principle, simulate any formal reasoning process—and therefore that, if it’s possible to describe human intelligence in purely formal terms, it’s also possible to artificially simulate it. 

In today’s marketing technology terms, however, AI refers to a much more nebulous and hard-to-define set of capacities. And just about everyone explains it by referring to the Amazon recommendation algorithm or the Uber app, or the familiar triumvirate of virtual personal assistants, Siri, Cortana and Alexa. Basically (very basically), AI as it’s understood in the marketing tech world is machine use of algorithms to formulate smart outputs in response to human inputs. The algorithms go to work crunching data—usually, the more the better—representing previous similar interactions, as well as all kinds of relevant contextual factors.

Old school AI meant the algorithms blindly repeating the same processes until a human showed up to tweak them. For some time now, however, there’s been the presumption that AI systems will incorporate a feedback loop so that the algorithms can correct and improve themselves without human intervention. That’s machine learning, and it’s why AI and machine learning are terms now used almost interchangeably.

Easy to understand for book or hotel recommendations: But where Salesforce plans to ring value out of AI is  by having it baked into the customer success platform so that marketers or sales reps get real-time recommendations for next best actions, from optmizing campaigns to calling a prospect about a whitepaper they just downloaded. Truthfully, Dreamforce didn’t see Salesforce unveil any major AI developments since those we described a few weeks ago. But it was the connecting thread in the keynotes, the mantra being: “It’s smart, because it’s powered by Einstein.”

Celebrating Genius

The AI mavens and brand reps I spoke to praised Einstein—and Salesforce’s savvy—to the heavens (many are Salesforce partners, of course, but not exclusively so). “We’re very excited about Einstein,” said Sklar, despite her position in the C-suite of an IBM company: “Einstein—and Watson,” she said, “are going to be key platforms.” 

Executives at brands which have already invested in AI felt that Einstein will raise the profile and value of their offerings. Zendesk, Boonin told me, has been investing in AI for a couple of years, using it to power its Satisfaction Prediction which predicts outcomes of customer support interactions. “Thank you, Salesforce, for bringing this to the fore,” he said. “They’ve done a great job of weaving Einstein into existing functionality across their cloud, and they’re democratizing for AI the everyman. Of course, we’ve seen this [making sophisticated innovations readily accessible] for twenty to thirty years in enterprise tech, so this is straight out of the enterprise software playbook.” 

Swensrud, with his twin perspective as a former Salesforce CMO and current CMO at a point solution brand, emphasized the need to get beyond the buzz.  “I’ve been part of creating and showcasing these buzzwords. At Dreamforce in 2006, we said ‘Welcome to the cloud.’ Now that has become absolutely meaningful, but the words get overloaded—cloud this, cloud that, when all it means is accessing apps over the internet rather than from the server in the closet. We’ve had many of these: social, mobile. Now it’s AI, and there were probably 25,000 people sitting in the Benioff keynote and wondering what it is.” They’re looking for companies like Salesforce to explain it to them.

Nevertheless, he agrees that AI is “where we’re all going in the not too distant future.” It’s valuable for predictive lead scoring, and it’s “important for the future of marketing: Chasing the dream of one-to-one marketing—a campaign I built for you, not for people who are sort of like you.” 

As someone whose brand is not involved in developing AI, Five9’s Anadkat had an interesting insight. In the evolution of media and channels, AI is another “new thing in the middle,” like, for example, geo-location. “Customer facing brands will use this as a differentiator,” he said. “We’re looking at use cases.” One example he gave was Deloitte’s Patient Connect healthcare platform, which leverages Five9’s cloud contact technology, and where there seems to be a clear role for AI in interpreting and making recommendations based on signals from wearables.

PROS has been “perfecting” an AI approach to pricing and sales optimization “for three decades,” Reiner told me; in recent years it’s been evolving its analytics platform to drive business improvements from marketing to the supply chain. “I’m ecstatic that everyone’s talking about AI, ” he told me. “For a long time, B2B didn’t believe they could really use data science.  My only concern is that Salesforce is making it seem like it’s automatic.”

In Einstein’s Shadow

Despite the widespread enthusiasm, I did wonder whether there was concern that users getting their AI dose from Einstein would feel less need of the expertise of AI-powered point solutions. 

Lattice uses AI to accelerate revenue by driving insights about readiness to engage or convert to apps embedded natively in other platforms (including Salesforce). “AI, the way we think about it,” said Upadhyay, “it’s meant to make people do their jobs better.” It’s like putting on an Iron Man suit, he said. “Sidelined by Einstein?” he mused. “We don’t think so.  You can provide intelligence in so many different ways.” Also, he said, Einstein only works on data inside Salesforce.  It’s like talking to someone who has “only reading five books.” Lattice’s B2B data cloud tracks 250 million businesses worldwide, essentially by reading the Internet.  “Salesforce’s approach is much more gentle,” Upadhyay said.  Also: “They’re pretty good partners.”  They haven’t killed marketing automation, he pointed out. Marketo is still doing well; and their acquisition of Radian6 didn’t kill Sprinklr. 

At InsideSales, Plante told me: “The long term vision is application of AI to a range of challenges—we started with sales.” This is another platform which “crowdsources” data from external sources to optimize the sales funnel (“To make AI work you have to have massive data, and B2B companies tend not to have that scale [from their own resources].” He also felt bullish about the way Salesforce is raising AI’s profile. “I don’t think there’s a downside,” he said. “It’s absolutely fantastic to have the validation of our strategy. I couldn’t have wished for a better outcome.” 

As for the limitations of restricting Einstein to Salesforce’s own data eco-system, a Salesforce spokesperson told me: “Einstein leverages all data in Salesforce—customer data; activity data from Chatter, email, calendar and ecommerce; social data streams such as tweets and images; and even IoT signals. We also have tons of data coming from Chatter and community users around articles, topics, experts and groups. Once data is brought into Salesforce, it is ready to be leveraged within Einstein’s intelligent data models, without any additional data preparation requirements.”

But the more I dug, the more there seemed to be some telling misgivings about leveraging Einstein for specific use cases.

One Size Doesn’t Fit All

One shared reservation about Salesforce’s Einstein strategy was that it seemed to make light of what PROS’ Reiner called “domain expertise.” PROS has a deep focus on certain industry verticals, including travel, insurance and food manufacture and production. “It’s still naïve, the way people are talking about AI. Algorithms only work if you specialize to the problem. Computers aren’t more advanced than people, and the devil is in the details.” Using raw—i.e. not specifically trained—algorithms can produce “horrendous results.” Without understanding a domain in depth, he told me, it’s hard to understand which data elements are truly relevant. Specialize around specific business problems—that was Reiner’s message. 

Lattice’s Upadhyay echoed the point. “Different use cases require different algorithms,” he said. “Generic AI won’t work. The way people buy a $100,000 product is different from the way they buy a $100 product, and the same algorithm can’t be used for both.” Sam Boonin pointed to a concrete learning experience at Zendesk. “We thought we could build a [machine learning] model across all our customers. That was wrong. Zendesk’s Satisfaction Prediction tool produces positive or negative predictions for the outcome of individual customer service cases. Algorithms needed to be specific to each individual account, and Zendesk has produced tens of thousands. Conversely, for its Automatic Answers tool, which connects responsive content to customer needs using deep learning, they found they could train a single model to work universally.

“Einstein is brilliant from a marketing viewpoint,” Boonin said, “but this stuff is hard: It’s very use case and customer specific.” He speculated that the way forward for Einstein would be through premium packs designed to meet the needs of identified customers. (He also expressed surprise that Salesforce is making a “platform play”—encouraging users to build their own machine learning-powered features on Salesforce. With Google, Microsoft and Amazon already in that space, Boonin thought Salesforce was “bringing a knife to a gunfight.”

However the Einstein initiative plays out, it’s clear that, just as the Plum Village monks were leading Dreamforce attendees on a path to mindfulness, we are all—like it or not—on a “path to the cognitive,” as Corinne Sklar put it. “Augmented intelligence,” as Bluewolf calls it, “is the most important trend in business.” Perhaps ironically, for a concept which started out as the dream of using computers to simulate human intelligence, the personalization potential of AI will allow businesses to “treat customers like human beings again.”

–Salesforce covered DMNTech’s expenses to attend Dreamforce ’16


Total
0
Shares
Related Posts