Don't pick on Watson. IBM didn't mention Jefferies analyst James Kisner by name, but it was clear CFO Martin Schroeter was trying to knock down the analyst's critique of Watson point by point.
Following IBM's second quarter earnings, Schroeter talked about Watson becoming easier to implement and finding more use cases in enterprises. In a 53-page report, Kisner noted that Watson may lose its cognitive computing position because it requires services and consultants to implement.
Here's a look at the Watson debate blow by blow.
Is Watson a bear to implement?
Kisner's bottom line on Watson went like this:
Our checks suggest that while IBM offers one of the more mature cognitive computing platforms today, the hefty services component of many AI deployments will be a hindrance to adoption. We also believe IBM appears outgunned in the war for AI talent and will likely see increasing competition. Finally, our analysis suggests that the returns on IBM's investments aren't likely to be above the cost of capital.
We're not just scaling our business into new areas but also improving deployment. For example, we implemented Watson for Oncology at Baheal Pharmaceutical Group in less than a month. This quarter, as I mentioned earlier, we expanded with Baheal into genomics. We also announced the new collaboration with Hackensack Meridian Health, a prominent U.S. provider, to combine Watson for oncology with their real-world data to help oncologists improve cancer treatment and reduce costs.
Now there's a good reason IBM is noting health-care wins. One of Kisner's core arguments was that the University of Texas' review of a Watson deployment at the M.D. Anderson Cancer Center highlighted the issues with deploying Watson. The special review is a must read.
Our checks suggest that IBM's Watson platform remains one of the most complete off-the-shelf platforms available on the marketplace. However, many new engagements require significant consulting work to gather and curate data. Our checks suggest that Watson is a finicky eater when it comes to data enterprises can feed it - in other words, IBM has very exacting standards for data preparation. The halt of and cost overruns in the MD Anderson engagement with Watson epitomize our concerns here.
IBM launched the Watson research project in 2006, won Jeopardy in 2011 and then embarked on health care. Watson for financial services arrived in 2012 and now there's a set of industry-specific efforts. The pitch is that Watson will augment human intelligence in multiple areas.
Schroeter said that Watson is finding use cases in government as well as financial services. And the IBM Services Platform with Watson will also bolster adoption. "We saw Watson deployments continue to expand globally. The cognitive opportunity is a global one. It's not centered in New York or Boston or Silicon Valley. We can't just look and listen in those places," said Schroeter.
Can Watson handle competition?
Kisner noted that there are plenty of areas where Watson can be competitive, but industries can stitch together more APIs to other AI tools without the consultants and time Watson requires. Kisner noted:
One of the most competitive elements of a Cognitive Software Platform is Machine Learning. Of course not all competitors can offer Machine-Learning-as-a-Services (MLaaS) like Google, Microsoft, IBM, and AWS can, but there are a growing number of open source deep learning frameworks on the market.
This proliferation of frameworks, AI startups and scaling competition means Watson is going to face price competition. Everywhere Watson plays--natural language processing, generating evidence-based hypothesis, learning and processing--the cognitive system faces a number of competitors.
We also present below the AI/Machine-Learning related APIs available from IBM, Microsoft, Google, and Amazon - it appears customers have a lot of choices. This is already having an impact on pricing - for example, IBM dropped the price of Watson Conversation over 70% from $0.0089 to $0.0025 per API query per month in October of 2016. Other companies like Microsoft, Oracle and SAP also have significant machine learning efforts underway and may be more credible threats in the Enterprise near-term.
Schroeter didn't address Watson pricing directly, but the API calls are ramping. He said:
We also had strong growth in our Watson platform, which underpins our enterprise AI strategy as we build scale. Conversation API usage and the number of active users are up strong double digits quarter-to-quarter as we help clients embed cognitive into their workflows...The value Watson provides depends on the complexity of the conversation. So the number of API goals can vary, but we have the potential to deliver value from every digital conversation. This will scale as more of today's 270 billion customer support phone calls shift to digital interactions.
Will Watson deliver a return for IBM?
This question is almost impossible to answer or predict. Why? IBM's Watson brings revenue to multiple businesses--cloud, consulting and software. Watson may also open up new markets.
Kisner made a few projections and came up with the following:
And a more detailed model.
Now IBM isn't likely to outline that detail and growth may be hard to track given Watson's capabilities were largely built by acquisitions. Kisner's detailed model includes the larger purchases such as Truven.
In the second quarter, IBMs cognitive unit saw revenue of $4.6 billion, down 1 percent from a year ago. Gross margin was also down.
Yet, Watson's impact probably has some impact on IBM's other business. After all, IBM recently created a cognitive hardware unit. However, revenue declines are the norm. The one sure bet is that cognitive and artificial intelligence will be a big business. The years ahead will determine where Watson ultimately stands.