By now we're used to machine learning (ML) and Artificial Intelligence (AI) making their way through the developer community and into the tool chains of some of the biggest tech companies, including Microsoft, Google and Facebook. What we're not yet used to, though, is sighting such technology very often in self-service analytics tools and Enterprise/line-of-business software applications.
That is really starting to change now, as announcements in the last few weeks from Domo and Absolutdata orbit around AI/ML specifically and another announcement from Alteryx ties in too, albeit a bit more peripherally. Taken together, these announcements provide anecdotal evidence that ML/AI can be both useful and accessible to business users.
Let's start with Domo, which in late March announced its "Mr. Roboto" AI service, at the company's annual Domopalooza event. While Domo can be classified as a cloud BI vendor, it rather sees itself as a SaaS business application provider that coalesces data from multiple sources to give you a comprehensive view of your business. And with that context in place, Domo has dispatched AI technology to raise alerts when it sees anomalies in that collective data pool -- in fact, that's the crux of what Mr. Roboto does, out of the box.
Domo says Mr. Roboto uses models built on K-means, linear regression, logistical regression and other algorithms to detect anomalies in business data and surface them quickly. A centralized location for monitoring such anomalies is Domo's Alert Center, though it will also utilize the service's "cards" to display any anomalies that could impact sales forecasts and product renewals. And while the models underlying this anomaly detection functionality are accessible to interested data scientists, business users benefit from the machine learning directly, without having to master, or even study, machine learning fundamentals and technologies.
Of course, this begs the question of whether business software can do more than report ML findings and force the user to interpret them and decide on responsive actions. What if, instead, the software interpreted the ML findings and then recommended appropriate action? Domo is intentionally leaving that opportunity to its partner ISVs who can use Mr. Roboto as a platform to add machine learning to their own Domo add-ons.
But Absolutdata, with its Navik AI platform launched just last week, is taking those dots and connecting them. The company is providing SaaS applications that use AI to optimize and help direct priorities and next actions for Sales and Marketing professionals. Its SalesAI, MarketingAI and ConceptAI products look at data in relevant applications in order to rank deals, campaigns and product launches and recommend next steps to the people in charge of pursuing and running them.
For examples, SalesAI avidly monitors CRM data (from both Salesforce and Microsoft Dynamics CRM), ranks which prospective deals are most likely to close, and recommends a salesperson's next move, be it a phone call, an in-person meeting or some other activity. Rather than giving the salesperson some raw insight on the data, SalesAI stacks it all up and presents the sales professional with Weekly Game Plan screen listing deals in ranked order.
Each deal is shown in a summarized format, including a column showing the recommended product to sell as well as an icon representing the prescribed next action to close, and can be drilled down upon for more detail. One piece of deal-level detail is a pitch "cheat sheet" that will enumerate the points a salesperson can make to help close the deal. Think of this as a list of important notes, an outline for part of an in-person meeting, or even a script for a phone call.
Each deal gets a "lead score," based on past history of the account, history of similar accounts and success paths of the top salespeople in the organization. The application also allows the salespeople to provide feedback if and when certain recommendations appear to be unwise or invalid. This allows the models' accuracy to improve. Managers can also override certain recommendations and specify their own. This is crucial to have when specific company priorities for a certain quarter or year need to be pursued; sales is not always algorithmic, after all.
Alteryx: why embed when you can extract?
This is all pretty cool, but what if the problem domain isn't sales and marketing, and what if predictions alone are helpful to have? In that case, enabling a business user to build predictive models and score her data with them could work really well.
Now, if every business user knew how to program in Python or R, we'd be all set here. Unfortunately that's not the case, so what can analytics companies do about it? Well, Alteryx, which recently went public, decided some time ago to add predictive analytics capabilities to its core data prep product. Alteryx did so essentially by building a GUI around many parts of the R programming language, especially those that relate to building, testing and storing models.
And Alteryx announced just yesterday that its server product is now available for cloud deployment on the Amazon AWS Marketplace. Customers can "bring their own license," effectively making AWS a deployment option, in lieu of installing on premises, or can opt for usage-based pricing, which starts at $8.02/hr for a server with 4 virtual cores. While the AWS Marketplace reveal was not a machine learning-based announcement per se, the fact that Alteryx has the R-based predictive analytics capabilities that it does, with a server that can now be deployed in the cloud, and the option of usage-based pricing, makes it quite ML-relevant.
For business users to get the predictive analytics work done, having the Alteryx GUI, and some taxonomy by task of R functions, really helps. But to take full advantage of this facility, business users can really benefit with some deeper knowledge of the underlying technology. Perhaps with that as motivation, Alteryx has seen fit to partner with Udacity to offer a Predictive Analytics for Business "Nanodegree". Together with the product capabilities and cloud deployment, this makes things interesting.
Clearly, Alteryx's approach to ML is much more DIY-based than a fully embedded strategy like Absolutedata's. While the latter seeks to abstract away the technology, Alteryx chooses to expose it. But providing an on-ramp for technical competency is a valid enablement approach as well, and the Alteryx product does make working at this level easier than writing R code from scratch.
Take the AI train
Taken together, especially clustered into a period of less than four weeks, these announcements show some real momentum for applied ML/AI in sales, marketing and other lines of business. It's still very early days of course, but there's more than a modicum of enthusiasm in the industry, and interest in marketplace, and that points to further advances to come, likely at a rapid clip.