Episode Summary: In some ways, investors in AI have to do a lot of what we do at TechEmergence, which is sort through marketing fluff and determine what’s actually working and what’s more of a pipe dream, as well as what’s coming up in the next five years that seems inevitable and what’s more likely to flop. In this episode we’re joined by Li Jiang, a venture capitalist with GSV Capital whom I was connected with through Bootstrap Labs as a pre-event interview — we’ll both be at Bootstrap Labs’ Applied AI event in San Francisco on May 11. This week, Jiang speaks about the current areas of AI that he sees driving business process automations, as well as what technologies he believes will make a long-term impact in terms of automation. His insights on where AI automations are generating cost savings and increased efficiency, as well as what roles might be completely replaced or significantly augmented by AI, are useful nuggets for companies who are thinking through some of their own business processes and are eager to identify low-hanging fruit.
Expertise: Startups and investment management
Brief Recognition: Li Jiang is a venture capital specialist, Vice President at GSV Asset Management and the chief evangelist at GSVlabs. Prior to GSV, Li served as an analyst at William Blair, a growth-focused investment banking firm where he completed both mergers and acquisitions and equity and debt financing for clients in the healthcare, consumer goods, and business services sectors. He has also held roles at Morgan Stanley in global capital markets and Goldman Sachs in investment management. Li founded a student storage and shipping company and grew it into Northwestern University’s largest student company.
Current Affiliations: Venture Capital Specialist, Vice President at GSV Asset Management and Chief Evangelist at GSVlabs
1 – Network effect may be the “ninth wonder” of the world.
Software vendors are collecting so much data from customers using their services, that it may not be necessary to collect as much data for programming needs in the future — and programming is a task that has been safely in the singular domain of humans. For example, today companies don’t need 5,000 people booking appointments and replying with a time available to learn how to program effective responses, because service providers are leveraging the network effect — tapping into mass data sets from other clients who have built the same type of response programs — and selling these extended benefits of being able to calibrate data to one particular activity. This exploding plume of data (a term that I picked up from my interview with Canvas Ventures’ Ben Narasin), which businesses are now able to tap into and sell is something that Jiang terms memorably as “the ninth wonder” of the world.
The following is a condensed version of the full audio interview, which is available in the above links on TechEmergence’s SoundCloud and iTunes stations.
(2:42) Where do you see AI delivering a tangible return on investment, beyond R&D and the “cool factor”, in the companies you’re looking at today?
Li Jiang: I look at it in a very simple framework.In every company there’s two sides of the equation: one side is the revenue side…and the other side is the expenses side — how do you drive down expenses and drive more efficiency. Right now, we see AI companies in basically every imaginable function…just on the revenue side, you see companies trying to work on higher conversion rates in sales and marketing, that’s an area where I think AI has a lot of promise…the company Brainify, they’re looking at all your data from your public sources online to understand your intent…and using that data to give to companies and brands, and when you go to the websites of those companies and brands, they’ll serve you their right products on their homepage.
(7:50) Anything else on the revenue side that you see working…what else is exciting you about what you see hitting the ground running today?
LJ: One area that I think is exciting and maybe scary is really products that are trying to replace human staff, like in sales for example, the entry-level sales person is called a FPR Sales Development Rep, and there’s a company— Kylie, as an example — that is building a product that could automate a lot of the conversation, the back and forth emails, the most common conversations, and I think that’s one area where it’s starting to emerge and there’s lots of promise; instead of having a thousand sales people doping that work, you can do 90% of it with AI and the 10% that can’t be handled can go to a human rep, so all of a sudden you have this 10x improvement or 10x efficiency…
(10:40) What are some of these tasks — is it scraping the web for the proper contacts and sending out a one-to-one style campaign, is this voice and phone related — what is the domain of application for Kiley?
LJ: Right now it’s mostly text and email, it’s actually smart enough to understand photo and images, but it’s not making sale calls as far as I understand…there is a human component, where a human can look at a pre-generated email and make changes before they send it out. There’s ways the system gets trained over time to become more and more like that person, because you get this feedback loop where if the standard message you’re sending is being changed by the human all the time with these edits, it learns that…
(16:42) What are the areas requiring some human calibration but really not PHD level unique insights every single time, text activities that can eventually fade out. It sounds like that’s something you see a lot of traction with today and are optimistic about in the future.
LJ: Yeah, that ‘s right, and even another quick example, when you project this out a couple years, is engineering. There’s two companies, Tero.Ai and Gigster, who are working on automating a lot of the engineering flow; you write code once for this one aspect of a product, you don’t have to build it again and again, and the way they’re approaching it right now is in an outsourcing model or a contract-worker model, where they’re going to big companies and saying if you use our services, we’ll build this app, we’ll build this product for you, and overtime their database of products and components gets so massive so that the next time someone says, hey I need to build this mobile app, instead of having a human write all this code, they say we already have 98% of this app done, throw it all together have a designer come and design the UI and they’re done…
(22:24) What sort of broad swaths of functionality do you feel most bullish about (in regards to) outright automation in the coming five years from an investor’s perspective?
LJ: I think five years is going to feel like a long time in AI because every week there’s some thresholds that are being hit, I think even insurance is a pretty obvious one. A lot of financial stuff, like being a trader, a public market investor, all the stuff that involves a lot of data and patterns that the human brain, we’re just not capable of, (our brain is) not doubling in capacity every 18 to 24 months, so we’re just not as equipped to process all that data. The same with marketing, same with customer support…there are ways you can optimize a machine to be much smarter and learn faster and the feedback cycle is so much faster than a human…
A couple of others — lawyers, a lot of the legal work that is basic level discovery or writing standardized templates, I think there’s some areas of law that will go away to machines and there are some areas that will not go away. Doing transactions, probably not because there’s so much emotional involvement…there’s an interesting company that started out of my alma mater, Northwestern, called Narrative Science, that writes financial reports, that’s basically taken over the job of journalists. That’s kind of a scary concept, there’s so much art involved in that profession supposedly, but content is being created by machines, which in some ways is almost sad, but it’s happening.