I’ve received some inquiries about where spending on artificial intelligence and cognitive technologies occurs in our tech market numbers (see, for example, my “US Tech Market Outlook For 2017 And 2018: Mostly Sunny, With Clouds And Chance Of Rain” report). The short answer is that we include them in our data on business intelligence (BI) and analytics, though, so far, spending on these technologies is still small — probably less than a billion dollars for 2017.
But even as artificial intelligence spending grows, it is likely to remain small in terms of visibility, because artificial intelligence solutions are likely to be functions in existing software products and not something that firms buy directly. Put another way, the biggest buyers of AI will probably be software, services, and hardware vendors that use AI to help their products and services work better.
There is precedence for this pattern in the BI and analytics market. My Forrester colleague, Boris Evelson, has been collecting data from the leading BI vendors in terms of the percentage of their revenues that they get from end customers versus from OEMs (original equipment manufacturers). On average, about 10% of these vendors’ revenues come from sales to OEMs. And that could well be understated, because vendors such as IBM, Microsoft, Oracle, or SAP don’t provide data on the explicit (or, more likely, implicit) value of the analytics products that are used in their applications.
With AI, these proportions are likely to be even larger. Among the ePurchasing application vendors that I cover, almost all of them are talking about using artificial intelligence in their solutions, most often in the form of machine learning for capturing and making sense of contracts, invoices, purchase orders, RFIs, and other ePurchasing documents. In some cases, these AI functions that vendors have added are homegrown, created by their own data scientists. But just as many ePurchasing vendors have included BI tools such as BusinessObjectives, Qlik, and Tableau, so too are they likely to private-label and use AI tools from other vendors. For example, in a recently announced deal between IBM and SAP Ariba, IBM is effectively closing down its Emptoris ePurchasing product suite (which competes with SAP Ariba’s) in return for SAP Ariba licensing and using IBM’s Watson AI tools.
Deploying AI in this fashion makes a lot of sense. Employees and users who can benefit from the insights and efficiencies of AI can do so in the context of the applications that they are already using for their work activities, rather than having to master a new AI tool. My prediction is that other AI leaders will follow in the footsteps of IBM and primarily sell their AI tools and services to other vendors, with direct sales a significant but still secondary channel. That will certainly present challenges to those of us who track the size and growth of the tech market. To avoid double-counting (more than we do today for BI), we will need to adjust downward the reported or estimated revenues of AI vendors for their sales to vendors of process applications. But for most tech buyers, getting AI built into the applications they already use will be a better choice than buying an AI tool or service that they will then have to customize to address a business problem.