AI is changing enterprise SaaS profoundly, says Emergence Capital
The worst-kept secret about enterprise software is that it doesn't get used very much. Or at least, it doesn't get used nearly as much as you would think given that companies pay millions of dollars in seat licenses for the stuff. That's why a whole industry of startups has arisen such as People.ai to automate tasks like filling out forms in Salesforce.
Thus, the great challenge of enterprise software is engagement, how to get people to use it. Is it possible that some kind of artificial intelligence code, by giving constant feedback on signals in data, can solve that problem?
That's the bet being placed by venture capitalist Jake Saper, who is a partner with Emergence Capital, which has offices in San Francisco and nearby San Mateo, California. Saper has invested in a slew of startups that are using machine learning in ways he expects will change what it means for an employee to interact with software. They include Vymo, Guru, Textio, and Chorus.ai.
"Our hypothesis is that cloud software will become more and more personalized, it will become hyper-tailored to me as the employee," said Saper in a phone call with ZDNet. "And we call this coaching networks."
Being hyper-tailored is a way for the software to presumably offer a more personal engagement with an individual employee.
Also: Computer simulation enters the AI age: the Altair story
The coaching network, in Saper's view, is a thing that replaces the guild system of old, where you would apprentice to a master to learn a craft. In a modern, remote-work world, there's less and less of an opportunity for people to be next to one another in a physical space, or even in contact during a given time zone. "We have to find more scalable models to help people do jobs when they are more geographical dispersed," said Saper.
Enter the AI-driven coach, a piece of software that is informed about the best practices and is always awake and never takes a vacation and becomes more specific as it interacts with an employee.
Vymo is a good example. Its sales app lives on a smartphone and tracks movement of the sales rep when they're out and about. It will use a variety of machine learning approaches to surface recommendations, such as calling on the next-nearest prospect in a drive through a certain part of town.
"You have to build the product in such a way it drives really heavy usage," observes Saper of Vymo and the other programs. "Usage will determinate the success of these products, because usage begets more data which makes the product better and better."
"You have to build the product in such a way it drives really heavy usage," observes Emergence Capital partner Jake Saper of Vymo, Guru, and other "Cloud 3.0" programs. "Usage will determinate the success of these products, because usage begets more data which makes the product better and better."
The other programs also depend on user engagement to enhance the program's feedback. Guru makes "collaborative knowledge management," according to Saper.
"Within an organization there is lots and lots of knowledge in people's heads, and the idea behind Guru is that in order to gain access to all that knowledge, you dis-aggregate an organization's knowledge into cards, such as a card on how to process this refund, or a card to handle this customer objection," explained Saper. "The idea is that those cards are proactively surfaced to a sales rep or a customer support rep while they are having that convention, so that the knowledge they need finds them wherever they, are whenever they need it."
Saper likens Guru's program to "Clippy," the ill-fated Microsoft Windows utility that would show a little animated paper clip offering tips during your computer session. "Clippy sucked," Saper acknowledged. "It was cute and funny, but the depth of Clippy's knowledge was limited; imagine that but actually built in 2019, with all the machine learning we have available to us."
Chorus.ai is a program that plugs into a Zoom Video call and listens to the dialogue. Chorus can help coach a sales rep in real time on next best answer to give in response to a prospect's question. As an interesting side note, Emergence is also a backer of Zoom, and Saper thinks Zoom will become a very broad platform for new companies such as Chorus. "We believe there will be coaching network companies built ton top of Zoom."
Coaching networks, if they take off as Saper expects, mark a shift from the prior two waves of cloud computing, a shift that Saper feels is profound. The first wave was led by Salesforce, in which Emergence was the first investor. (Emergence focuses on early-stage investments.) The promise of Salesforce was to simply reduce cost by replacing physical data center investment with cloud-based code. It was radical twenty years ago and now it's the norm for software.
Also: AI makes inroads in life sciences in small but significant ways: Lantern Pharma's quest
Salesforce led to another cloud pioneer, Veeva Systems, which Emergence also bankrolled. Veeva is a suite of applications for life sciences companies. The insight there was that an application could be more valuable if it ran the core processes that are very specific to an industry, such as the compliance document management for clinical trials.
"I think that that era of software is commoditizing," said Saper of the previous two eras of cloud. Those tools are predominantly "workflow" tools. They set up a static way way for people to work, and that is becoming easier and easier to build. What then is the "moat" in the 3.0 cloud era of Vymo and Guru and Chorus, asks Saper. It is "the data network — not just the data, not unless it is constantly refreshed by the use of the product." The idea is that a product like Vymo, to the extent it is used, contains the insights about a sales person.
"Imagine a world where Vymo ends up succeeding wildly," Saper prompted. "It becomes the world's most insightful database of what makes a sales rep successful, this type of rep that responds to this type of prospect in this manner ends up increasing their close rate by 15%."
Such insights are what "create data moats in these products over time," said Saper. Once it's successful, no one wants to rip out a product because it contains amassed value. "If we are right that it's the next big app in software, if you can build your usage moat quickly enough, it will be very difficult for someone to come in and displace you," said Saper.
On the contrary, they may want to expand its usage.
Which leads Saper to another of his big ideas: The pricing of software will have to change for Cloud 3.0 programs. "To price in an age of machine learning, you want to be sure you are incentivizing usage," and that can't, he insisted, be the seat license model that is typical with Salesforce and the rest. "This is not something the industry has woken up to."
Also: 'We won't have the sexiest AI, but everything it says is true,' says Narrative Science
The first thing to do, Saper believes, is to "identify a usage-agnostic metric that grows as your customer grows, and that your product impacts meaningfully and measurably." He offers Textio, one of the portfolio companies, as an example. Its technology for "augmented writing" can help an HR manager produce better help wanted ads, the company claims, guiding them as they write the ads with prompts for what words to use.
"The way they price is quite clever," said Saper. It's not based on the number of recruiter seats you have, or the number of job posts, because that would all dissuade usage.
"Instead what they have done is identified this usage-agonistic metric that grows with the customer, and that metric is simply the number of people the customer plans to hire next year." In other words, buyers of Textio pay based on expected outcomes.
Saper thinks a lot of pricing will go this outcomes-based route. He expects it will make pricing more complex for startups to figure out, but ultimately fairer to the user. "It will be harder" for the vendor, he said. "You will actually have to think about your business" to come up with the right pricing. But then, "building companies is not easy" is Saper's way of summing that up.
It all comes back to usage. None of the data moats will happen without people spending time in the program. Which means the "land and expand" strategy of software firms, to find a champion in a department and then spread by enthusiastic co-workers' word of mouth, becomes more and more essential. The biggest danger for startups like Vymo and the rest is that someone buys them with a credit card, tries them out, and then tires of them and they lie fallow.
That suggests there will be a high bar for machine learning in SaaS: make it engaging, make it addictive, and make it worthwhile.