A recent article in the Financial Times argued -- fairly -- that despite the billions of dollars poured into “AI” companies, investors have, on the whole, not seen returns consistent with the hype. There are exceptions of course, but, by and large, the promise(s) appear to have not been met, as of yet. The argument was not simply a lamentation, however, with the author suggesting that the next wave of focused AI solutions might indeed generate better results and returns.
Such a sentiment is not uncommon in technology. In order to garner investment, entrepreneurs employ hyperbolic language to excite potential investors and the business press follows this lead in order to ensure that they don’t miss out on the appearance of prescience. So, out of the gates, there is much promise and little delivered and when this gap is revealed, negativity enters the scene.
While sentiments are cyclical, so too are the benefits gained from particular technologies or methodologies. When technologies are introduced, they are “sold” as integral to the “future-proofing” of the buying business. As such, the seller and the buyer create a social-contract of sorts that suggests that the time-horizon of measurement is long; it is understood that “sand-boxing” and “experimentation” require risk-taking and a techno-version of FOMO (Fear Of Missing Out.)
After the initial phase of excitement, the technologies have to be operationalized, they have to fit into the processes and culture of the organization. When technologies -- even AI -- are in the process of being operationalized, the theoretical joys give way to the practical difficulties of “making stuff work.” This always takes longer than even the initial social contract suggests and as organizations feel financial and other pressures on their tech investments, they look for technology ROI. When they neither see it nor know how to measure it in their own context, they push back on vendors, and internecine conflict between the “dreamers” and the “realists” drains organizational energy.
Most organizations push through this phase in, well, phases. When technologies are finally operationalized, the organization moves into the “routinization” phase wherein the technologies adopted become, in a sense, invisible. They are just “part of the air.”
For most, AI is still in the “social contract” phase. Dreamers dream of what could be and vendors happily slake those dreams with sugarplums. For progressive organizations, however, the dreamy phase is over, and operationalizing AI is upon them. Here, the key to success is “vertical AI,” which is to say that AI technologies have to be trained on very specific, contextual data sets and must be used in the context of the matrix of constraints and imperative specific to the vertical industry or sub-industry.
Indeed, vertical AI cannot be “built” in a vacuum without deep industry expertise. Further, vertical AI must be built in direct collaboration with its ultimate consumers and not in a hermetic box, no matter how well-funded that box might be.
The great AI companies of the present and future will have to heed these admonitions. They’ll have to sell the hype but then work directly with customers to operationalize and routinize the technologies. This will happen only if they build AI frameworks and “fill in the blanks” in active and dynamic collaboration with customers deeply embedded in particular industries or verticals.
Investors in AI must insist on this or keep their financial powder dry.
For more on AI, check out these recent articles.