Earlier this week a London VC firm, MMC Ventures, released a report titled “The State of AI: Divergence 2019.” The document is 145 pages long and contains a wealth of fascinating insights about the adoption of artificial intelligence across Europe. The report attracted the attention of major media outlets, including the Financial Times, Forbes, The Verge, and CNBC, and despite the report’s broad scope, the headlines all drew attention to a single finding on page 99: Based on their analysis of 2,830 European AI startups, 40% do not actually use AI in a way that is “material” to their business.
The media’s singular focus on this particular point illustrates one of the fundamental challenges in the nascent age of AI in the enterprise. MMC notes that “companies prefer to buy, not build, AI. Nearly half of companies favor buying AI solutions from third parties.”
This dynamic explains the explosion of AI enterprise software startups, not just in Europe but worldwide. The downside, according to MMC, is that almost half of the vendors promising AI magic are pretenders not contenders. Like the Wizard of Oz, this group would urge you to pay no attention to the man behind the curtain.
The inevitable shakeout has not yet occurred because of the extraordinary demand for AI solutions. MMC calls it “the fastest paradigm shift in technology history. In the course of three years, the proportion of enterprises with AI initiatives will have grown from one in 25 to one in three. Adoption has been enabled by the prior paradigm shift to cloud computing, the availability of plug-and-play AI services from global technology vendors and a thriving ecosystem of AI-led software suppliers.”
When evaluating potential AI enterprise software vendors, we recommend following the approach famously taken by Ronald Reagan in nuclear disarmament negotiations with the Soviet Union: Trust, but verify. We believe the following criteria are critical in choosing a potential partner:
- Check Credentials: MMC notes that “demand for AI talent has doubled in 24 months. There is a gulf between demand and supply, with two roles available for every AI professional.” However, despite the fact that “Data Scientist” is the “best job in America” for the last four years in a row according to Glassdoor, according to MMC “the pool of AI talent remains small. AI demands advanced competencies in mathematics, statistics and programming; AI developers are seven times more likely to have a doctoral degree than other developers.” Building and deploying profitable AI models requires a deep bench of top-notch talent.
- What’s in Your Model? Although inevitably described using the umbrella term “AI”, the source of explosive growth is due to major advances in machine learning, which is a subset of AI. MMC also points out that “machine learning” is another umbrella term with over a dozen different approaches. Every solution has an architecture behind it. Ask vendors to explain what machine learning techniques they currently use in their models and what they see as the critical elements in their data science roadmap. Is it simply a mash-up of tried-and-true techniques such as random forests and Bayesian networks, or are they working at the frontier of reinforcement learning? Do they have a proprietary approach, have they filed any patents? Are there systems in place to monitor model performance and data fidelity? And perhaps most importantly, can they explain how the model made a particular decision? Beware of junk data science
- Show Me the Money: if enterprises successfully identify the valid signals from the cacophony of AI marketing noise, MMC declares “AI has numerous, tangible use cases today that are enabling corporate revenue growth and cost savings in existing sectors.” Any qualified AI software startup will have positive results they can share from current customers or pilot projects, but AI is not about incremental improvement. AI done right is revolutionary, revealing paths to better, faster decisions that create profitable, sustainable growth. Don’t settle for anything less.