By Devaki Raj
Enterprises of all sizes rely on visual intelligence: Think of the skilled factory worker who spots a specific product defect a mile away, or the safety officer who understands when working conditions look unsafe.
Subscribe to the Crunchbase Daily
Today, organizations are at an inflection point. The sheer increase in volume of data and sources of visual information in the digital age presents a challenge: There is simply too much information to process and too few eyes to make sense of it.
What organizations require is a scalable way to institutionalize the expertise distributed across their skilled workforce.
This is why I believe computer vision has the potential to transform the way business is conducted today. Visual information is everywhere in global business, and computer vision provides the tools to empower an organization to unlock the value of that data and transform it into real ROI. In fact, experts predict that computer vision is already on track to become a $50 billion industry in the next five years.
Over the last 10 years, as organizations consolidate their data investments, we have seen conversations at companies evolve from analyzing “big data” to realizing value from that data through the integration of AI.
However, efforts to roll out various types of AI across the enterprise — including computer vision — have not always met with success. Just last year, Venturebeat estimated that only 13 percent of data science projects end up in production.
So why are the vast majority of these projects stuck in R&D when there are real business outcomes organizations can realize with these technologies today?
The challenge is that companies thus far approach AI as a tool for a centralized group of highly trained experts. This is an understandable approach: Effective deep-learning powered computer vision requires data science of the highest quality.
Many companies then go to great lengths to recruit top-notch data science talent. As someone who recruits data scientists herself, I can tell you the process is far from easy.
Still, even after retaining the best data scientists, AI projects at these organizations struggle to leave R&D.
Let me be clear: It isn’t the data scientists’ fault that these projects rarely make it into production. It’s a failure of the data-science-centric paradigm itself.
That paradigm typically ends up working like this: Business leaders receive pressure from executives to roll out AI solutions. Operational directors have a laundry list of problems they’re trying to solve, so the data scientists are called in to help “sort it all out.”
But, data science teams are almost always overburdened — they have project requests as diverse as modernizing financial software to fortifying security operations. It is not possible to manage such disparate projects from one centralized location all at once. Even given infinite resources, these teams lack the essential subject matter expertise to solve each individual business problem — knowledge possessed by those operating on the front lines.
At CrowdAI, we firmly believe that the best AI happens as close to the problem as possible, with direct involvement from all the relevant stakeholders: subject matter experts; data scientists; operational leaders; and frontline workers. That is why we insist on our AI being code free for users from start to finish, as all of these stakeholders must share a part for the successful deployment of vision AI.
Code-free tooling, however, is not a panacea. To truly democratize AI, model development needs to be approachable to both data scientists and non-scientists alike. This is why at CrowdAI, we’ve built a platform that brings together technical and nontechnical users with point-and-click interfaces that allow users of all technical capabilities to apply their particular expertise to model success.
Think of the power of online tax preparation software: What was once solely the domain of highly trained specialists is now available to everyone with an internet connection. Tools like TurboTax haven’t removed the need for accountants — quite the opposite. Now, everyday citizens can contextualize their personal financial circumstances and collaborate with experts to focus on the most difficult or unique cases.
In every case, democratizing AI development across the enterprise ensures that vision AI products move swiftly out of R&D and become force multipliers for organizations to radically reimagine the way they do business.
Devaki Raj is CEO and co-founder of San Francisco-based CrowdAI. Formerly the analytics lead and a product manager at Google, Raj earned a BA and MA in biology and statistics from Balliol College at University of Oxford, and her MSc in applied statistics from the University of Oxford.
Stay up to date with recent funding rounds, acquisitions, and more with the Crunchbase Daily.