The Three People Rule
I have a rule where if three people send me the same article, or talk about the same news story, in the same week, it warrants me looking into.
This past week, a company called Persado, which does natural language processing/generation for advertisements, triggered the rule. The news was Chase, after working with Persado for over three years, signed a five year deal to use Persado’s technology to start writing more advertisements. The story ranked #1 on Adage for most of the week and triggered a lot of fanfare about computers taking over.
Persado fits the classic "AI for X"
AI for X is when a company picks a niche area, collects lots of data in that area, hires data scientists/machine learning engineers to apply AI/ML to the data, and then releases a product. In Persado's case, they created a dataset with over a million tagged phrases and words, hired data scientists/natural language processing specialists to create algorithms to create analyze and create new text, and voila you have a product. Persado's hope is their software outperforms humans at writing ad copy, which allows companies to ultimately save money in the long run.
AI for X has exploded
Funding for AI startups has exploded, setting a record with almost $7.5bn raised in Q2 2019. CB Insights pointed out investors are willing to fund anything related to AI for X including applying AI to fish farms and bull matchmaking. The party will continue as Softbank is about to start making investments with its second Vision fund, which is focused on AI, and has $108bn in capital.
With all of this money floating around, it got me thinking - where do AI for X compare to other “Technology for X” movements, like Uber for X, Blockchain for X, Internet for X?
Uber for X
There are many good analysis on why the Uber for X craze has not worked out. In a good Hackernoon post, the author argues Uber has been able to succeed because driving is a very commoditized skill and everyone needs it, making logistics somewhat more easy to handle. Uber for SAT tutoring is not a lucrative business because SAT tutors have specialized skills, and people don’t want SAT tutoring at all hours of the day. Thus, many Uber for X companies are destined to fail because the supply/demand just isn't there.
Internet for X
I don’t have to go into detail here because “Internet for X” worked for a lot of industries. Yes, a lot of companies that tried it in the late 1990s blew up, but everything is on the internet
How will “AI for X” play out?
AI for X will fall somewhere in the middle, more towards the Uber for X side. As companies build out their data science teams, they will not need to use multiple vendors to apply AI to every part of their business. Those with relevant AI/ML skills took advantage of the shortage of AI/ML talent and started AI for X companies. But as more AI/ML talent enters, the advantage will shrink.
How do AI companies have sustained success?
First, AI for X companies need to avoid becoming commoditized. Even if companies build out their data science teams, AI for X companies will still have data assets that companies need. AI for X companies will need to make a hard decision on whether to start selling their data assets or keeping it proprietary.
Second, AI for X companies need to show ROI in a clear and constant manner, so business stakeholders don't decide to build the product themselves. This will involve AI for X companies continuing to innovate and applying advances in AI to their products.
Finally, AI for X companies products need to be accessible for everyone. If these products and services are integrated into everyday workflows, it will be hard to end users to give the services up. As more companies democratize data access, AI access will most likely follow.
Otherwise you may be moving onto the next Technology for X trend