What Do Venture Capitalists Look For In An AI Deal?

Proprietary, non-replicable data and the ability to predict across systems. JL

Tom Taulli reports in Forbes:

"The most logical area for AI to have an impact in the world of enterprise software will be to sit on top of multiple systems of record where it can act as a system of prediction. I am excited about sales conversion, the most effective marketing channels, the probability that a credit card transaction is fraudulent, and whether a particular website visitor might be a malicious bot.” One of the biggest challenges for startups—in order to succeed, you need a data set that can’t easily be replicated. "Diligence comes down to who has proprietary data. Does this startup have domain-specific data or could Google or Amazon sweep in and replicate what they do?’

Recently SoftBank Group launched its latest fund, called Vision Fund 2, which has $108 billion in assets. The focus: It’s primarily on making investments in AI (Artificial Intelligence). No doubt, the fund will have a huge impact on the industry.

But of course, VCs are not the only ones ramping up their investments. So are mega tech companies. For example, Microsoft has announced a $1 billion equity stake in OpenAI (here’s a post I wrote for Forbes.com on the deal).
The irony is that—until a decade ago—AI was mostly a backwater, which had suffered several winters.  But with the surge in Big Data, new innovations in academic research and the improvements in GPUs, the technology has become a real force.
“I’m convinced that the current AI revolution will be the largest technology trend driving innovation in enterprise software over the next decade,” said Jeremy Kaufmann, who is a Principal at ScaleVP. “The magnitude of this trend will be at least as large as what we observed with cloud software overtaking on-prem software over the last two decades, which created over 200 billion dollars in value. While certain subfields like autonomous driving may be overhyped with irrational expectations on timing, I would argue that progress in the discipline has actually exceeded expectations since the deep learning revolution in 2012.”
Given all this, there are many AI startups springing up to capitalize on the megatrend. So then what are some of the factors to improve the odds of getting funding?
Well, in the AI wave, things may be different from what we saw with the cloud revolution.

“Unlike the shift from on-prem software to SaaS, though, progress in AI will not rewrite every business application,” said Kaufmann. “For example, SaaS companies like Salesforce and Workday were able to get big by fundamentally eating old-school on-prem vendors like Oracle and ADP. In this AI revolution, however, do not expect a new startup to displace an incumbent by offering an ‘AI-first’ version of Salesforce or Workday, as AI does not typically replace a core business system of record. Rather, the most logical area for AI to have an impact in the world of enterprise software will be to sit on top of multiple systems of record where it can act as a system of prediction. I am excited about conversations around the likelihood of sales conversion, the most effective marketing channels, the probability that a given credit card transaction is fraudulent, and whether a particular website visitor might be a malicious bot.”
Data, The Team And Business Focus
An AI startup will also need a rock-solid data strategy, which allows for training of the models. Ideally this would mean having a proprietary source.
“At the highest level, a lot of our diligence comes down to who has proprietary data,” said Kaufmann. “In every deal, we ask, ‘Does this startup have domain-specific understanding and data or could a Google or Amazon sweep in and replicate what they do?’ It’s one of the biggest challenges for startups—in order to succeed, you need a data set that can’t easily be replicated. Without the resources of a larger, more established company, that’s a very big challenge—but it can be achieved with a variety of hacks, including ‘selling workflow first, AI second,’ scraping publicly available data for a minimum viable product, incentivizing your customers to share their data with you in return for a price discount, or partnering with the relevant institutions in the field who possess the key data in question.”
And even when you have robust data, there remain other challenges, such as with tagging and labeling.  There are also inherent problems with bias, which can lead to unintended consequences.
All this means that—as with any venture opportunity—the team is paramount. “We look at the academic backgrounds,” said Rama Sekhar, who is a partner at Norwest Venture Partners. “We also like a team that has worked with models at scale, say from companies like Google, Amazon or Apple.”
But for AI startups, there can often be too much of a focus on the technology, which could stymie the progress of the startup. “Some of the red flags for me are when a pitch does not define a market target clearly or there is not a differentiation,” said David Blumberg, who is the Founder and Managing Partner of Blumberg Capital. “Failed startups are usually not because of the technology but instead from a lack of product-market fit.”