Getting Real With AI

Getting Real With AI

Getting Real With AI

Assessing the much-hyped artificial intelligence trend

You no doubt have heard about artificial intelligence (AI) and machine learning (ML) being the “next big thing in [insert industry here].” Whether you’re in AgTech, MarTech, EdTech, InsurTech or some other -Tech, disruption is here, if not already a foregone conclusion.

In healthcare, the massive potential of AI & ML has been widely written about. I won’t discuss this in much detail here, but check out thisthis, and this for additional background.

The hype around AI pervades so many tech news outlets and in so many of our inboxes, that “AI fatigue” has set in. The result? Some are already beginning to minimize AI technology as a table stake, or satirizing it as a universal disruptor (the excellent and entertaining CBInsights newsletter does this regularly).

The AI hype is sky high in the digital health space

At StartUp Health, we’ve seen a notable uptick in applicant companies highlighting that their technology leverages AI and machine learning, despite it remaining unclear how or why that is the case. Promises of what the technology can accomplish are lofty, but clarity on how it will be achieved is somewhat lacking.

The value entrepreneurs promise lies in the AI engine itself, rather than the use case being pursued. They conflate “AI” with value rather than it being part of a larger, valuable offering.

What’s more is they rely on AI & ML to describe their company, not their technology. There’s a belief out there among entrepreneurs that latching onto buzzwords like “AI” will make stakeholders pay attention, and that their technology is automatically needed and desired in the market. When pressed, it’s difficult to get entrepreneurs to explain the specific utility of the output their solutions generate?—?especially when data sources are unclear, not proprietary or data integrity is unknown.

On a panel at Health 2.0 a few weeks ago, Iana Dimkova of GE Ventures echoed this, saying sarcastically:

“If you’re an early stage company looking for funding, make sure your deck says you do AI. It doesn’t matter what you do, just make sure it’s on there and it automatically results in a 20 percent bump in valuation.”

At StartUp Health we don’t track valuations, so it’s tough to see what degree of truth there may actually be here. We do, however, track VC funding, and this serves as a decent proxy.

Along with funding amounts and investors, we track the characteristics of companies that are getting funded, categorizing them by sector, subsector, end user, specialty of focus, and technology type.

While we’ve only reported on sectors in our quarterly funding reports to date,we recognize there’s more nuance in the digital health space than there has ever been. Accordingly, we’ve begun to shift to capture an unprecedented amount of information about digital health funding trends.

With this shift, we have been able to track and analyze the funding trends for AI & ML companies like we haven’t been able to before. In parallel, we at StartUp Health have shifted in our thinking about how we view a company’s technology.

How we view a company’s technology

When we evaluate a company, we utilize a scorecard methodology that takes into account 8 Company Mindsets: Health Moonshot Story, Backable Team, Differentiated Brand, Increasing Traction, Scalable Business Model, Proprietary Technology, Valuable Partnerships, and Capital to Execute.

The 6th, Proprietary Technology, deals with the actual technical offering of the company, as you might have guessed, as it relates to the use case being pursued. Does the technology solve a challenge for a specific customer? What value can be ascribed to the technology, based on specific examples of its use? Has the company established barriers to entry so that it cannot be replicated? The strength of the answers to these core questions leads to a score on the Proprietary Technology Company Mindset, and companies who inspire confidence on all three score highly.

The flipside? Companies who are showcasing their technology for its own sake, appear to be a hammer looking for a nail, or have no competitive advantage, score poorly.

The Big Data/Analytics digital health sector is a great example of how Proprietary Technology, and our thinking around it, has evolved. AI & ML are not sectors themselves, nor are they company descriptors. They are technology types?—?you’ll see this reflected in future StartUp Health Insight reports.

From the StartUp Health Insights Q3 Funding Report?—?The Top 10 Most Active Sectors of 2017 YTD

The figure above, from the StartUp Health Insights Q3 Funding Report (check it out in full here if you haven’t already) highlights funding in the top 10 digital health sectors of 2017 year-to-date. In terms of dollars, Big Data / Analytics tops the list, after hovering in in the middle for the past few years. AI & ML are among the several technology types in this sector.

Big Data / Analytics companies are nearing $1.4B in funding in 2017 YTD. Source: StartUp Health Insights

We have seen a sensical arc of the type of Big Data / Analytics companies being funded. In 2015–2016, it was largely about collecting novel data in novel ways. Think 23andMe in 2015. Specific use cases had yet to be clarified, but we all knew there was potential here with genomic data coming into the mainstream.

In 2016, there were 5 $100M+ deals in companies dealing with biometric data acquisition (Thalmic LabsGrail, Guardant HealthiCarbonX, and Kernel). Though data collection was still the focus, you see use cases for the diagnosis and/or treatment of specific disease crystallizing.

In late 2016 and 2017, these companies and others are starting to toutapplying this data to solving problems. Grail is now talking about using its massive datasets to detect cancer early, before it’s a problem, for example. Notably, you have to scroll a ways down their website to see any reference to the actual technology being used (“intelligent models” and “deep learning” appear in paragraphs towards the middle-bottom)?—?it’s all about the use case. Take a look.

The companies getting funded, and getting traction, are not using AI & ML themselves as the selling focus. The focus is the differential value delivered by applications powered by AI in the background. These companies have modeled themselves around getting proprietary data first, then layering advanced technology on top to deduce patterns and glean meaningful insights. This is “the right way.” As Vinod Khosla has said, data is the rocket fuel of the AI engine, so the effectiveness of AI applications relies heavily on training datasets and the availability of useful data at all.

If you view other technologies through this paradigm, you’ll see parallels. Remember a few years ago when being “mobile” or “mobile-first” was the crown jewel of a pitch? A year later, and especially now, the mobile app itself isn’t impressive. The use case powered by the mobile app might be.

Where we go from here

The shift in thinking around Proprietary Technology will be reflected on StartUp Health Insights going forward as well. Starting in 2018, we will no longer be counting Big Data / Analytics as its own digital health sector, but we are breaking these companies down by tech types and the applications they are pursuing.

This is timely, since funding in the space is growing. AI & ML companies have accounted for ~12% of deals in 2017 (more than 1 out of 10 deals!), and ~8% of dollars invested. Encouragingly, we’re see funding in the space trending upwards. 4 weeks into Q4, we already have seen more than in both Q1 and Q2 2017.

AI & ML funding in 2017 YTD by deal count and dollars. Data is inclusive of the first 4 weeks of Q4 ’17 data. Source: StartUp Health Insights

Along with the hard numbers, we’re anecdotally seeing more companies who are going about this “the right way” coming into the mainstream.

This discussion as a whole brings to mind companies like Conversa Health , who are facilitating automated, personal patient conversations to engage patients and collect patient-generated health data. Data collected is incorporated into the course of care. Conversa leans on behavioral science to engage patients in the conversation (they have dozens of ways of saying “hello”, for example), encourage them to share data, and create regular touchpoints between patients and care teams. They incorporate AI & ML technology to identify red flags in communication and data patterns and to determine whether to escalate care teams’ involvement. Conversa closed their Series A in May 2017.