I recently experienced a little health-tech déjà vu.

I attended an IBM Executive Briefing at which one IBM speaker said, when considering how ripe the healthcare industry is for tech innovation, “The healthcare industry accounts for $9 trillion in spending in the US. Thirty percent of that $9 trillion is waste. So we see an enormous opportunity for tech to eliminate that waste and capture some of that revenue.”

So here’s the déjà vu: In 1996 I attended a conference at which Jim Clark, the founder of Silicon Graphics and one of the co-founders of Netscape, along with with Marc Andreessen, spoke about the vision for his new company, a healthcare start-up called Healtheon. He said, “We’ve targeted the 30% waste in the $9 trillion healthcare sector. If we can capture just a sliver of that revenue, we will make a lot of money.”

Apart from demonstrating that I have been attending health-tech conferences for far too long, the eerie similarity in those identical economic calculations—separated by 19 years—reveals something even more illuminating: just how naïve otherwise-rational technologists can be when they train their sights on healthcare.

Jim Clark’s Healtheon, which proclaimed the vision of radically simplifying the entire healthcare experience through superior IT (and exploding as the next Internet supernova in the process), wound up merging with WebMD shortly after this speech, and then quietly decayed in the cold darkness of fizzled phenoms. The dream of monetizing the inefficiencies in healthcare was built on the false premise that the “30% of waste” was money that people in the marketplace agreed was “waste” in the first place and, in the second place, they didn’t enjoy making that money and wouldn’t savagely defend their livelihoods.

IT purists looked at the paper-based system of medical records, patient billing, insurance claims and the like and concluded, quite sensibly, that this was no way to run a modern healthcare system. And they were right, of course.

It was (and is) massively inefficient. But, as we have learned through the initiatives to automate large-scale systems like the Internal Revenue Service, Veterans Affairs, the Department of Motor Vehicles and the American voting system, driving out inefficiencies takes many years and iterations. The entrenched interests running the prevailing system can be wily, politically astute and incredibly effective at making reforms ineffective, particularly those being championed by private-sector entities that are interested in relatively short-term financial returns. Witness Google’s abandoned effort to solve the electronic health record problem.

The high-tech visionaries were also radically naïve in their understanding of the byzantine complexity of the healthcare market, where the customer (the patient) is not the decision maker (the doctor) and the decision maker is not the payer (the insurance company). In this multidimensional dynamic, normal economic incentives simply do not apply.

In fast-forwarding nearly two decades, from 1996 to 2015, we can see some changes. The health economic dynamic has shifted, as patients are now more economically involved in decision making. Too, doctors have relatively less authority and payers enjoy dramatically more authority. But if IBMers think it’s going to be easy to monetize $3 trillion in healthcare “waste,” then I suggest they call Jim Clark, most likely out on his super-yacht Comanche. (While Healtheon didn’t achieve its ambitions, Silicon Graphics and Netscape worked out pretty well for him.)

But is the time right for IBM to do for healthcare what Jim Clark could not? Based on the evidence of a single day, the answer is … unclear. The company does have a few exciting things planned. There’s IBM Watson, which—besides being the cognitive computing Q&A machine that in 2011 beat the best humans (and Ken Jennings) at Jeopardy!—is now helping patients and doctors answer even more important queries.

“I’ll take ‘Hubris’ for $800, Alex.”

So where’s Watson? Is it the answer (or at least an answer) to the problems of streamlining healthcare and improving patient outcomes?

Phrased in the form of a question, my assessment is, “What is a cognitive computing platform that may speed diagnosis for doctors, help companies recruit for clinical trials and remove uncertainty for patients but can’t do some very important things pharma marketers could use today?”

I’ll unbundle these one at a time. But first, a bit of an explanation of what Watson is—and what is isn’t.

Watson, despite its portrayal on Jeopardy!, isn’t really a single velvet-draped computer, although it is built on scads of IBM’s massively parallel POWER7 processors and an array of algorithms known as DeepQA technology. It’s an open cloud-based question-answering platform that uses natural language processing to retrieve information from an enormous library of memory-held knowledge, then applies reasoning algorithms to determine how likely the answer to a question is to being true.

Or, in plain English: You can ask Watson a particular question in your own words, and it can sort through basically all the available knowledge on that topic in the known world and give you the most likely answer … within seconds.

Watson for Oncologists, Patients and Clinical Trials

In healthcare, this cognitive-computing engine has been deployed most tangibly in the form of an oncology adviser. Originally developed with the Memorial Sloan Kettering Cancer Center (reportedly after someone at MSKCC saw Watson humiliate Jennings and gave IBM a call about devising a nobler use for the technology), the Oncology Expert Adviser enables doctors to analyze a particular patient’s medical record to identify potential evidence-based treatment options based on an unimaginably broad and deeply established pool of the latest medical knowledge. The Oncology Expert Adviser is like having Dr. House on steroids (but without the Vicodin and the chronic crankiness) advising you on your treatment plan.

Soon, IBM says, Watson will be used by one particular pharma company to create a website where, once a drug is prescribed, patients can ask questions in their own words and the website will provide answers in near real-time. (I’m curious about how this will be positioned so as not to appear that the pharma sponsor is offering actual medical advice to patients.)

A third application of Watson lies in speeding the efficiency of clinical-trial screenings by including all the relevant inclusion/exclusion criteria for a trial, matching them against a database of electronic health records covering 100 million US patient lives and thereby identifying the best patients for the trial. Again, in seconds.

But what about, let’s say, marketing?

All of this is promising, exciting and a potential boon for healthcare. But it seems as though there must be dozens (or hundreds) of other equally powerful applications for the cognition engine. (By the way, it isn’t called a “cognition engine,” but I like this term and especially its nod to Charles Babbage’s Analytical Engine, which was one of the first designs for a working computer, back in 1837.)

As someone interested in helping companies more appropriately and effectively market their brands, I wonder if there isn’t something else we could do with Watson. Here’s one idea: Why not use Watson’s cognitive computing capability to match the plethora of streams of media data with the profiles of appropriate patient types and have Watson help refine (or even conjure) the best plan for reaching these patients?

Maybe Watson can help us up our game in programmatic buying for healthcare, since programmatic demands considerably more complex data in healthcare than in other categories. (Think about combining data sets spanning disease prevalence or insurance coverage—even sales rep access—as well as first-party data sets like those coming out of social media. It gets mind-numbing pretty quickly.) And then consider the value of being able to deliver more refined, even personalized, messages to patients who will be more likely to get the right product and improve their health.

It won’t make anyone $3 trillion, but it sure seems like a smart use of the cognition engine. Wouldn’t you agree, Watson?

Bill Drummy is founder and CEO of Heartbeat Ideas.