Pole-vaulting Unicorn Wins Olympic Gold
Reading time: 3 minutes.
Don’t believe in fairy tales! Consider this story:
“Today there is only a 4%-8% adoption rate of AI within the corporate world. Within the next 18-24 months we are likely to see this rate explode to 80%-90%.”Fortune’s Demystifying Artificial Intelligence in the Corporation (9 October 2019) citing senior IBM executive, Rob Thomas.
The first half of this quote — current state — is way below the vendor and industry analyst-driven consensus in the market. The second half is quite the opposite, way above the market consensus. I have never seen anyone, anywhere, credibly suggest an “overnight” change of that magnitude. What’s going on here? Let’s split this into its two constituent parts.
It’s not unusual to find survey results that claim 20 to almost 50 percent of enterprises are already investing in AI technologies. They’re typically suffering from serious sample bias and other research flaws.
The PWC Global CEO survey I cited in Only 5% of US enterprises are investing in serious AI-based business projects does not seem to suffer from the sample bias issues I discussed in Incredible Sample Bias and Deception in AI Market Penetration Reports. (Caution: Some readers might be misled if they treat future intentions cited by survey respondents as accurate measures of the future state.)
In the PWC survey, they observed that only 5 percent of US enterprises — and 10 percent of European enterprises — were investing in widescale production investments in AI-dependent projects. Pretty much identical to the IBM executive’s pronouncement.
Kudos to IBM and PWC!
The next two years
Will corporate investment in AI jump to 80 to 90 percent of all enterprises?
That rate of change is unprecedented. Unbelievable. Inconsistent with all I’ve seen in 45 years in this industry.
Consider the chart below. I use Rogers’ terms to describe the diffusion cycle, the period during which a technology goes from inception at the leftmost side of the chart to total adoption at the 80 to 90 percent level on the right. Rogers breaks it down into five stages, starting with Innovators and ending with Laggards, as depicted in the graphic below. The blue curve (a bell curve) represents sales volumes per unit time. The mustard yellow curve represents total market penetration. Not all technologies ever get into the Early Majority phase. Geoffrey Moore’s Crossing the Chasm talks about the substantial obstacles to moving into the Early Majority phase.
Strategic production application of AI technologies is in the Early Adopter phase (between 5 to 10 percent penetration). Experimentation (pilots, prototypes) is common in roughly 40 percent of enterprises.
It takes 20 years for technologies to diffuse from the Innovators‘ stage to the Laggards‘ stage — to hit 80 to 90 percent penetration — in enterprise use. Or longer!
There is still some risk that enterprise AI adoption will crash in Moore’s chasm and go into another “AI Winter.” I’d rate it as a one in three likelihood. Jump to 80 to 90 percent in two years?
An “AI Winter” is an alternative to consider
- This doesn’t mean everyone stops using AI technology!
- Let the heavens forbid we lose Siri, Alexa, Insight Engines, Convolutional Neural Nets, Deep Neural Nets, Generative Adversarial Nets, Word Embeddings and all the rest. We won’t.
- But we are no longer in the primordial glow of the AI big bang — the birth of a new AI universe — the feelings generated by the huge progress during the first half of this decade. Gains since 2015 have been more sparse. Leading AI researchers are seeking new models that have “common sense” and can “understand” because current ones are powerful but not very smart.
- A move towards the next “AI Winter” will be presaged by media hype dying down, negative stories becoming more visible, pressures to develop an “AI strategy” dissipating, and the feeding frenzy of exponential growth being replaced with more moderate year over year increases in use. Do you feel a chill wind blowing? Or it just a temporary draft?
- See AI about to crash (again). Get ready. As a follow up to that, I’ll devote an upcoming article discussing the likelihood of an “AI Winter,” as well as causes and consequences, so keep following my work.
In the current case, 80 to 90 percent adoption in 18 to 24 months would be an “overnight sensation.” I’ve seen predictions of sudden overnight success before. I have yet to see one proven accurate.
Eight-step AI Action Plan
Does that mean you should not be investing in AI technologies? No! Instead, go look at The eight-step AI Action Plan section in Is AI needed to cure this existential business crisis? (Part 2) and execute!
Don’t believe Polevaulting Unicorns will win Olympic Gold.
Disclaimer: This article represents my own opinion.