What is Happening?
Adding to its portfolio of advanced analytics/deep learning/machine learning capabilities, Salesforce has acquired startup MetaMind, which was launched in 2014 with $8 million in funding from Salesforce CEO Marc Benioff and Khosla Ventures. Among previous AI/machine-learning-oriented acquisitions, Salesforce picked up startup PredictionIO in March of this year, along with Tempo AI in mid-2015. While Salesforce has confirmed the agreement, it has revealed neither financial details nor whether MetaMind’s existing staff will be kept.
What MetaMind’s deep-learning platform can do is interesting; it typically has been used to predict outcomes based on non-traditional, unstructured volumes of data (often text and images) in order to predict outcomes of operations.
What makes it of compelling interest to us is what it can, or should, be used for in the context of Salesforce’s own business and the business of its customers and partners. This acquisition spotlights the real, rapidly-growing, daily business benefits of deep learning and machine learning technologies for IT providers and for their customers and clients.
Why is it Happening?
A new blog post on the acquisition from MetaMind CEO Richard Socher states that Salesforce plans to use its technology to “further automate and personalize customer support, marketing automation, and many other business processes. [MetaMind will] extend Salesforce’s data science capabilities by embedding deep learning within the Salesforce platform.”
It has been becoming obvious for a few years now that traditional enterprise business and IT data analytics is considered passé, despite the fact that the vast majority of business and IT operations still cannot access, process, analyze, and apply the vast majority of data already available. In most cases, data is isolated in functionally-oriented silos, and existing systems, operations, and management orientations are not designed, let alone optimized, to utilize the analytics in decision-making. As a result, what most business IT providers are doing is focused on task/function/process optimization and automation – and we see a practically unlimited global market for these types of capabilities through at least 2020.
That being said, various forms of “advanced analytics” are already the hot topic in business and IT leadership circles, due in large part to the acceptance and expectance of the trend toward Digital Business. Boiled to its essence, Digital Business builds on context-sensitive utilization of any relevant data in any form of decision-making/action-taking operation by any entity. Because the scope and pace of business are increasingly complex and less predictable in many ways, and because data is comprised of traditional and nontraditional types (e.g., numbers vs. images, sounds, text, colors, smells), simply mining and processing huge quantities of “big data” is no longer considered enough.
The next generation of almost any type or size of business will require increasingly-automated, neural/cognitive-IT-powered abilities to identify, qualify, analyze, and then act upon all types of data in all types of situations, correctly and in context. That is where MetaMind and others become part of the business and IT environment, especially as more requirements emerge to “learn” both input data (e.g., text, language, images, numbers, and more) and suitable actions, in context. This requires varying combinations of “machine learning,” including the following:
- Supervised learning is used to classify something, and/or to predict something based on what we know and can apply to a situation, categorizations, operational structures, and so on. It’s like interpreting the images on the flashcards by reading the names of the objects printed on the face of the cards.
- Unsupervised learning is closer to how humans learn from visual stimuli; we categorize visual clues based on similar attributes such as shape, size, color, etc. with the output validating a hypothesis of what we think we see, or enabling us to see something new: it’s like interpreting the image on the cards without knowing the names of the objects.
Net Impact
For Salesforce directly, this acquisition is pretty much as summarized by MetaMind CEO Socher’s blog post: Salesforce plans to use its technology to “further automate and personalize customer support, marketing automation, and many other business processes. [MetaMind will] extend Salesforce’s data science capabilities by embedding deep learning within the Salesforce platform.”
That type of application should enable substantial improvements in Salesforce’s own efficiencies. But of course what Salesforce learns from its own use, adaptation of, and innovation with MetaMind should translate into substantially advanced capabilities in the Salesforce platform and associated platforms/offerings (e.g., Database.com, Force.com, Marketing.com). Most of Salesforce’s competitors need to partner for and license such capabilities; this should provide a leg up for Salesforce in most if not all of its current markets while enabling new revenues, possibly in more/new areas of Digital Business (some of which may yet need to be invented).
And of course Salesforce will have more data, and more ways to profit from it. We would be surprised if Salesforce did not use MetaMind and other acquired capabilities on its own customers, to figure out what customers are going to do, why, and when, and then get more money out of them.
We know that Salesforce is not alone in these types of intentions. Amazon, Facebook, Google, IBM, Microsoft, SAP and Twitter all have invested heavily in AI, machine learning, and deep learning initiatives to utilize structured and unstructured data, to create and envision more digital business offerings and opportunities. We should see more from traditional analytics-oriented IT firms as well. As one senior IT strategist noted in meetings this week, “If you don’t have machine learning and you are a provider of anything that uses analytics, this is an important checkbox.”
The emerging multi-dimensional awareness and interaction requirements of Digital Business suggest that businesses (and their IT providers) need a more robust and scalable fabric to integrate customer and other interactions, preferences, improved interaction, and everything else that happens or might happen.
If we are correct in this assumption, then the future of business will be enabled and shaped by (1) business IT providers delivering solutions business users and analysts can use without hand-stitching of data scientists / machine language programmers; and (2) adaptable combinations of machine-learning-driven data gathering, analytics, and decision-making/action, from IT and Finance process automation to core business systems operation.
Beyond Salesforce, IT leaders can anticipate further acquisitions of new cognitive / machine learning startups over the next several years by large business IT application providers, seeking to continue to expand into new territory, acquire hard-to-find talent, take action to learn more about their customers, deliver new capabilities to customers, and reposition offerings to deliver on the promises of digital business. For today, we still need scientists and machine language programmers to validate much of what is done and seen – tools cannot (yet) save us from making a bad hypothesis and then spinning our data to match it. But they will.
