AI Wars: CRM Strikes Back!

Print Friendly
Share on LinkedIn0Tweet about this on Twitter0Share on Facebook0

SEE LAST PAGE OF THIS REPORT Paul Sagawa / Artur Pylak


psagawa@ /


September 22, 2016

AI Wars: CRM Strikes Back!

CRM CEO Marc Benioff looked to steal a bit of the Oracle World thunder by teasing its planned “Einstein” AI offering as its rival kicked off its annual confab. CRM, which has been building its AI talent base through recent M&A, will apply deep learning as a platform to make its applications more customized to the specific needs of its users and able to drive more effective decisions and results. By aggressively addressing the potential of AI for enterprise software, CRM is stepping ahead of most of its rivals, including ORCL and SAP, and should gain advantage as a result. The one major exception is MSFT, who’s own AI resources are well beyond those of CRM, including a dramatically deeper roster of experienced talent and a superior hyperscale data center infrastructure. We believe both companies will be able to use AI as a weapon to help wrest market share from the rest of the enterprise software industry.

  • CRM is taking AI seriously. CRM only really committed to AI in 2014, when it acquired RelateIQ and incorporated its relationship intelligence technology into its core application. Since then, it has stepped up its investment, culminating in its April 2016 acquisition of deep learning think tank MetaMind and the elevation of that company’s founder to its own Chief Scientist post. In May, Benioff formally announced an AI-first strategy for his company, and just this week, revealed early details for the company’s Einstein AI platform which will be formally introduced at CRM’s Dreamforce event in San Francisco next month. Einstein is a set of deep learning functions that will be embedded into CRM’s SaaS applications and that will be available to partners on the Salesforce cloud.
  • AI success depends on 3 factors. While the building blocks of deep learning systems are simple, the most powerful models are anything but. With thousands of building block algorithms combined by feedback loops into multiple layers designed to learn as they iterate through huge collections of data, outstanding AI solutions must be carefully built and adjusted for optimal results. This takes experienced talent, big data and powerful processing platforms. While CRM trails the top AI players, like Google and Microsoft, on these factors, it has made progress in differentiating itself from other rivals.
  • CRM focusing on aqui-hires. CRM’s venture capital operation, one of the most ambitious in Silicon Valley, has a portfolio of more than 150 companies, with a strong interest in funding talented AI scientists. The company has spent more than $500M taking some of these investments, including MetaMind, Implisit, Tempo AI, PredictionIO and RelateIQ, in house, more for the talent than for the products. The development teams from these deals built the foundation for Einstein AI and for AI functionality that has been built into CRM’s applications, evidence of real competence in integrating acquired talent into the company’s engineering organization.
  • CRM moving up the list for AI talent. CRM claims to have 175 data scientists working on its AI-oriented projects, although just 5 of them show up in our data base of experts that have cited at least 1,000 in papers published on AI. Leading these, CRM Chief Scientist Richard Socher is considered a rock star amongst his peers, having racked up over 9,000 citations despite having just earned his Stanford PhD in 2014. Amongst established enterprise software companies, CRM sits well behind MSFT, which has a whopping 141 scientists with 1,000 citations, and IBM, which has 83. Still, its AI prowess clearly stands ahead of both ORCL and SAP, neither of whom has even a single scientist with 1,000 citations.
  • Borrowing data from customers. After talent, the next most important prerequisite for leading-edge deep learning systems is data. Unlike FB and GOOGL, enterprise focused companies like CRM do not usually have permission to combine their customers’ data or to use it for their own purposes. This makes Larry Ellison’s claim from Tuesday’s OracleWorld keynote that his company controls one of the 2 largest consumer data bases in the world disingenuous at best. Within those constraint, CRM, which unlike ORCL hosts all of its customer data on its own servers, has a strong position to build AIs to help those customers gain insights from their own data, and could, perhaps, convince them to allow use of anonymized data to develop more powerful tools.
  • CRM’s infrastructure remains a concern. CRM’s extraordinary sales application is bundled with hosting on hardware and software infrastructure that is subscale, inflexible and inefficient compared with the leading hyperscale operators, AMZN, MSFT and GOOGL. This could prove a modest liability to CRM in developing AI solutions to compete with those potential rivals. Still, compared with its classic competition, ORCL and SAP, CRM’s data center platform could be an asset.
  • AI will be a weapon for CRM. We believe that Einstein will help CRM continue its outstanding growth at the expense of incumbent enterprise ERP vendors who lack coherent cloud transition strategies, much less proficiency with AI. ORCL and SAP lack a critical mass of experienced AI talent and have aimed their M&A at established SaaS application companies big enough to help drive growth, rather than the sort of pre-revenue aqui-hires needed to quickly build a roster of deep learning scientists. With most of the customer data ensconced inside of private data centers, these companies will be disadvantaged in building AI capabilities into their applications. ORCL talks a big infrastructure game, but its CAPEX spending levels are inadequate for building web-scale data centers. Ditto for SAP.
  • Longer term, MSFT looms as a formidable competitor. While CRM has the AI advantage of most of its rivals, MSFT is the clear #2 industry player in deep learning after GOOGL, and has aggressive plans to use AI to go after the same enterprise opportunities. CRM can use its AI to enhance its market leading SaaS customer relationship management apps and hold MSFT at bay for the time being, but long term, its talent deficit and significant data center infrastructure disadvantage will need to be addressed.

We believe that a future partnership or combination with one of the AI/IaaS powerhouses – AMZN, GOOGL or even MSFT itself – is an attractive possibility.

Einstein on the Beach

CRM CEO Marc Benioff seems to see the AI-generated writing on the wall. The enterprise applications of the future will embed deep learning-derived capabilities to adapt themselves to the specific needs of customers, to derive actionable insights from the flow of data and to empower users with intuitive, personalized interfaces that effortlessly connect them to those insights. To that end, CRM has invested – acquiring deep learning talent through hiring and acquisitions – to build Einstein, an AI platform to power these capabilities for its SaaS delivered applications and to enable its hosted partners to tap into the same. The company will formally present Einstein to its customers and partners at its Dreamforce event in SF next week, and expects it to be a major competitive advantage in its quest to expand its sales management software dominance into other aspects of the enterprise.

Versus most of its traditional competitors, this is a great opportunity. ERP leaders SAP and ORCL, already well behind CRM in realizing the vast benefits of the SaaS model, are also poorly positioned for the AI revolution. ORCL has only one and SAP doesn’t have a single scientist that has been recognized by the AI community with as many as 1,000 academic citations, while CRM has 5, led by wunderkind Chief Science Officer Richard Socher, who has racked up 9,000 despite having inked his Stanford PhD just 3 years ago (Exhibit 1). CRM has made 7 AI-focused acquisitions since 2014, has investments in many more AI companies through its aggressive venture group, and claims 175 engineers hard at work on Einstein (Exhibit 2). CRM also has a leg up on those rivals in its access to hosted customer data and its data center infrastructure.

Of course, ORCL and SAP talk a big game about the cloud, but stitching together a quilt of acquired cloud applications haven’t given either of them a coherent platform for realizing the functional synergy that CRM can lever from its well-designed suite of apps. During his recent OracleWorld keynote, founder Larry Ellison hyped the billions of consumer records contained in his company’s database software without acknowledging that almost all of them were held in many thousands of private data centers, and that even if ORCL had the ability to access them (which it doesn’t), it would be impossible for it to draw meaningful insights from information across different structured databases (Exhibit 3). SAP’s cloud aspirations are less fantastic, but still will be difficult to pull off given this lack of coherency. CRM can use its aggressive investment in AI as one more weapon to separate its offerings from the big two.

The elephant in the room is MSFT. The advantages that CRM holds over ORCL and SAP, MSFT holds over CRM. The Azure cloud platform is lower cost, higher performance, and far more scalable than CRM’s somewhat tired architecture. CRM may have 3 bona fide AI experts, but MSFT has 141 – second only to GOOGL. CRM has the dominant position in salesforce management software and will use AI to lever toward ERP, but MSFT has the dominant position in enterprise productivity software and intends to use its even greater mastery of AI to lever that franchise toward ERP from the opposite direction. While we see ample opportunity for both companies to erode SAP and ORCL’s strongholds, eventually, a confrontation awaits. Of course, Benioff understands this – last year he not so quietly discussed the potential for a merger between the companies. 500-pound IaaS gorillas AMZN and GOOGL stand as potential allies for the fight, or even as future acquirers, so there are strategic options. On balance, we would be buyers.

Exh 1: AI Citations by Company

Exh 2: SalesForce AI Acquisitions, 2014-16

Exh 3: Factors for Success in Artificial Intelligence

Exh 4: Net Plant Property and Equipment, 2010-2015

Exh 5: Capex Spending, 2010-2015

Exh 6: The SSR AI Heatmap

Print Friendly