As it readies the 50th release of its software suite, Salesforce says it is on course to be the fastest technology company to reach $10bn in revenue. It is slated to reach just north of $8bn in the current fiscal year, which runs through January 31, 2017.
Its European business is thriving, according to the company, and the UK – its biggest country market outside the US – is driving much of its business. Meanwhile, back in the proverbial lab, the company has many people working on machine learning and other techniques to make Salesforce smarter, and to take advantage of its combination of massive data sets and powerful cloud-computing infrastructure.
Salesforce Lightning is the UI platform the company launched in 2015. It comprises Lightning App Builder and Components. In addition to being a UI platform, the company has also started branding products built on this platform as Lightning products, such as Service Cloud Lightning. In a similar vein, the company released Field Service Lightning in March, which is designed to connect agents, dispatchers and all mobile employees, to schedule and dispatch work and management jobs in real time on any device.
Among the more than 150 new features upcoming in the Summer 16 release are Lightning Microsoft Continuum, a way to make the desktop-to-mobile experience seamless, File for Connect for Box to enable files in Box to be accessed from Salesforce desktop and mobile apps, Wave Notifications & Maps (automatic alerts and geospatial mapping in the Wave Analytics product), and Omni-channel Supervisor for real-time views and operational alerts. Summer 2016 ships in June.
Salesforce has quite a few staff with varied and interesting backgrounds, some in what is known as its search and data science team. Search probably doesn’t jump out as something Salesforce does considerably better than any of its competitors, but the company is working on changing that, and is doing so by leveraging machine learning.
Gary Flake is the CTO of the group. He has almost 30 years’ experience in machine learning, and was the chief science officer of Overture. That company effectively invented the paid-search model, and was bought by Yahoo, where he went on to start its research labs before moving to Microsoft and then Salesforce. Scott Rickard is the VP of data science, and worked with Flake in the past. Speaking at the company’s recent EMEA analyst summit, Rickard said he believes search is a major opportunity for Salesforce to differentiate itself.
Although Salesforce cannot access its customers’ data, or even the search queries (a point it is at pains to emphasize), it can draw inferences from user behavior once the user has seen the data. A theoretical example might be the ability for a CEO to predict revenue in any period based on the likely close rates of deals currently in the pipeline. This could be inferred over time from patterns around opportunities and leads, for example.
For search, Salesforce uses an implementation of Apache Solr, coupled with machine learning. Besides its Solr implementation being the biggest there is, Salesforce reckons its indexes are larger than that of Google’s Web search engine, simply because there are more objects in Salesforce than there are Web pages.
Salesforce already uses a variety of machine-learning algorithms, including Deep Learning and Random Forest, but also, as Rickard emphasizes, lots of math. The company is keen to push beyond rule-based AI (acknowledging that Random Forest is essentially rule-based), because it sees that as the old way of doing things. Google’s AlphaGo, after all, didn’t know the rules of Go. It learned to play Go by playing Go thousands of times before its recent triumph over the best human player in the world.
The company continues to make acquisitions that are machine-learning focused, such as PredictionIO in February 2016, MinHash in December 2015, and just recently, MetaMind. The last launched in December 2014, with $8m in initial funding from Khosla Ventures and Salesforce chairman and CEO Marc Benioff, so it’s a company Salesforce clearly knows well. MetaMind has been focusing its Deep Learning smarts on natural-language processing, database prediction and computer vision, and we expect that to continue within Salesforce.
Sometimes the results of those acquisitions end up in the search and data science group, while others, such as PredictionIO, end up in specific product clouds. The company now has hundreds of data scientists, including some neuroscientists. We expect to hear more about the fruits of their labor in the next couple of years, although we don’t expect it to be spearheading the company’s marketing message any time soon.