PHOTO: ERIC MILLETTE/SALESFORCE
An announcement by Salesforce.com this week that it would embed its artificial intelligence technology “Einstein” into its software for salespeople caps a three-year, $700 million AI push by the workforce enabler.
Salesforce isn’t the first, or the only entrant. But it’s the first commonly recognizable name, and it has the scale to push wide adoption in the business place. Now, that day seems closer.
Currently, artificial intelligence and its precursors, machine learning and data mining, face common objections: They’re not simple to do, there’s nothing off the shelf, and hiring data scientists and building big computing facilities takes too much time and money. This chips away at those.
It turns the corner from early commercial applications of AI (robo advisers, virtual assistants, and image recognition) to something deeper for the enterprise: Actually employing machine learning in functional areas as a new way to deal with the wealth of data that companies are amassing.
That affects not just IT departments, which have to make room to store and access the information, but Chief Marketing Officers, sales executives, human resources leaders and more who will actually use it. In fact, as AI creeps into functional areas, strategic purchasing decisions will involve more players than just CIOs and CTOs. That spells big shifts, and it’s a big deal.
Gartner predicts that by the end of 2018, 25% of durable goods manufacturers will use data generated by AI to power customer-facing sales, billing, and service offerings.
That sounds suspiciously like what Salesforce just announced as it joins a field that includes Aviso (also planning a presence this year at Dreamforce), NG DATA and, among others, inContact, which purchased Attensity. (Salesforce purchased RelateIQ.)
The products are proliferating. Some form of artificial intelligence is making itself felt in functions from sales to marketing, fraud detection, HR and recruiting, and intelligent tools. Some use supervised learning from data mining, others unsupervised machine learning, while a few newer experiments and applications are creeping into actual artificial intelligence.
But don’t be intimidated.
From an enterprise level, companies should make table-stakes investments in data management and cloud storage, so they have the room and the optimal formats to store and access data when needed.
Executives should get conversant at the difference between predictive analysis, which uses simple algorithms to make hypotheses, to neural networks, where machines build the algorithms and re-weight the importance of interconnections on the fly. (A “deep” neural network just layers on more types of information.)
My colleague Naveed Asem at Slalom explained to me the way our data analytics specialists meticulously build capabilities from quick wins in data mining to machine learning to complex algorithms. It sounded reasonable. My customer engagement colleague Mike Jortberg works with clients all the time at implementing the frontline-facing applications that make all this accessible. I know he’s excited by these developments.
We may be sliding down into what Gartner calls the “Trough of Disillusionment” – that valley of irritation between our fondest hopes and actual productivity. But put these developments into context: AI’s baby steps are already somewhat ubiquitous (Siri and Google Now might be in your pocket). Separate advances in data availability, new processing hardware, more powerful neural network algorithms, social network analysis and the quantified self have all progressed rapidly, laying the groundwork for organizational adoption.
Actual product innovation may be low at the enterprise level, but the most significant barriers in the way now include data architecture, governance, and integration – hardly insurmountable.
There may be a heap of vendors hyping their products and services (897 of them, according to a recent VentureScanner analysis), but the signs are clear. The pieces are aligning.
What’s left is careful planning.
A good place to start based on Gartner’s and my own analysis may be seeing what’s available and within budget to enhance your existing business practices, consider whether those enhanced processes can lower the cost or time-to-market for “low-hanging fruit” innovations, consider contracting proof-of-concept applications, partnering or acquiring a provider (depending on how well known your needs are), and invest in startups or open innovation to keep your antennae in the future marketplace.
The next innovation horizons are staggering in their promise: Next will be AI-enabled industries, as diverse as drone deliveries or automated financial services. Then come entirely new industries, such as driverless cars.
But the first steps will take place in your own enterprise. And the first new surprise might be plug-and-play AI.
— James Janega leads the Market Insights team at Slalom’s Chicago office