The story of how a team of brilliant minds convinced us they have the AI skills needed to allow large companies to make a leap forward and overcome the challenges of deploying and scaling AI.
As some of you may have read Thursday, Grove Ventures co-led a $15M Round A in BeyondMinds, alongside private investor Marius Nacht. We are thrilled to partner with Rotem, Roey and BeyondMinds’ brilliant team, who are developing a universally applicable and easily adaptable AI platform.
In recent years, we have been closely following Artificial Intelligence (AI) developments. When we met BeyondMinds, we were convinced that their ideas will shape the future of AI adoption for large companies. This has led us to write one of our largest ‘first checks’ to date.
In a world awash with buzzwords about Artificial Intelligence, finally, we felt that we’ve met every VC’s dream team: founders with vision, a deep understanding and know-how, and a great idea that actually can solve a growing pain of a lot of large companies. As I will show later on in this article, the vast majority of the market is struggling with implementing, deploying and maintaining AI solutions in a cost-effective manner.
BeyondsMinds’ unique way of thinking about the AI space allowed them to build a company like no other. The process of overcoming pre-conceived skepticism wasn’t easy (luckily, I had become acquainted with the company a few months prior to them seeking funding). During this demanding process, I was convinced to write Grove Ventures’ largest first check to date. Hopefully, by the end of this article, I will be able to share my reasoning.
For Some, Scaling AI Might Be Harder than Expected
The pace of enterprise AI adoption over the past few years is rapid. IDC’s Worldwide Artificial Intelligence Systems Spending Guide from 2019 stated that Enterprise AI Software is worth $18 billion in 2020 and will be worth $44 billion in 2024, a 24% CAGR. According to Grand View Research, the Artificial Intelligence market size will be worth $733.7 Billion By 2027. Though the numbers look impressive in terms of growth, most companies have not yet succeeded in their AI implementation strategies.
It seems that AI is the future of software, and that remarkable progress has been made in the field with problems that can and are already being solved based on data. All tech giants, be they Google, Facebook or Microsoft, invest heavily in their AI capabilities, build AI research labs, and hire the best academic talents. VC money is also pouring towards startups: Venture funding in AI companies had reached a mind-blowing $61 billion from 2010 through the first quarter of 2020.
Enterprises, on their end, understand that they can harness advanced AI capabilities to their benefit. Failure to scale AI means not being able to get a competitive edge, which puts their businesses at risk. However, many are still struggling to move beyond the proof of concept stage. Only one out of eight AI solutions makes it to production. A portion of the solutions that pass the deployment hurdle successfully, still result in negative ROI due to high maintenance costs, or issues such as data drift, the need to meet regulation and provide solutions with explainability, confidence, and bias mitigation, as well as continuous monitoring to stabilize AI systems to ensure increased value after deployment. According to a study by MIT Sloan in collaboration with BCG, only 1 out of 10 projects achieve profitability with AI. These all cause enterprises to cancel projects, sometimes even deep into the process.
Clever enterprises understand that a successful AI project relies on three things: (1) Empowering the right champion; (2) Designing the “correct” AI system that can meet production requirements; and (3) The right implementation within existing organization workflows. Only a combination of all three, will mean that their AI project will be deemed successful.
In the past few years, we have seen lots of reports from Gartner to McKinsey trying the analyze this slow adoption and systematic failure to achieve profitability with AI. While these reports focus on the management aspects of barriers, BeyondMinds is focusing on the technology-production barriers that hold the adoption back. While most (~70%) of companies are piloting with AI and have broken the management/organizational challenges, only 10% have broken the technical/production barriers. BeyondMinds has developed a unique holistic proposition for companies to accelerate AI adoption and break these challenging barriers.
AI Companies: The Horizontal Model vs the Vertical Model
AI startups are on a race to build a ‘magic machine’ that can cater to enterprises and help them use the power of AI for their specific needs at scale. Two types of startups emerged:
- Horizontal AI startups, who build data science platforms and aim to solve AI infrastructure challenges that most sectors will face as they deploy and scale AI solutions. Many great companies as Databricks, DataRobot, C3, H20.ai (and Grove Ventures portfolio company Deep-Checks) are defining this domain.
- Vertically integrated AI startups, on their part, aim to bring AI solutions and solve a specific need of a specific market. They become subject-matter experts, understand the types of data and workflows, understand the organizations and the players in the space in which they operate in as well as their needs and prioritizations, and lastly, they understand how to best integrate with unique systems and create an end-to-end solution for this business vector. Grove Ventures’ portfolio company Navina, for instance, is doing exactly that: it replaces chaotic patients’ data for primary care physicians with intuitive Patient Portraits, based on multi-sourced data and complex AI capabilities that analyze and organize the most important medical information, then serves it to the doctor in an intuitive way that saves time and improves the quality of treatment.
Introducing BeyondMinds’ Multi-Vertical, Modular AI Technology – From Insurtech to Manufacturing
BeyondMinds came to us with a different approach: They want to build a large AI organization and research lab so that they can create a technology stack that handles the core barriers in bringing AI to production. A technology that can be easily adapted to provide AI solutions to varied problems. Quite a formidable task.
They want to be a different kind of startup: Their solutions address AI challenges that concern multiple vectors. BeyondMinds can stabilize AI solutions in production where the data is dynamic and noisy, and deploy trust, monitoring, and feedback technologies to increase value over time. Their technologies are built in a modular way and are being tested in advance in extreme real-world scenarios. Their architecture minimizes the expert manual labor in the data science and deployment processes, and this enables them to scale like a software business, rather than like a service company. The scalable success of enterprise AI, they understand, does not reside within Data science excellence, but within a combination of operational and technological excellence.
BeyondMinds want to cater to specific common needs in multiple business verticals. It has built the first enterprise AI system that is universally applicable and easily adaptable. They deliver hyper-customized, production-grade deployments that enable sophisticated companies to overcome the massive failure rate in AI adoption and rapidly implement ROI-positive transformations. Though it is only the beginning of their journey, early customers and demand allowed BeyondMinds to create typical applications including: defect detection and predictive maintenance in manufacturing, fraud detection, insurance claim automation, risk assessment and underwriting in the financial services industry, and numerous additional use cases where AI is traditionally applied to solve core business problems.
In this, BeyondMinds combines elements of both software and services. They rely on powerful code, that can be tailored to repeatable needs in scale. Highly modular capabilities are at the heart of this software. These capabilities allow the software to perform complex tasks like generating natural language or interpreting images, as well as easily adapt to complex requirements. Unlike a lot of companies that require significant interactions with their clients to deliver, much like a services business, BeyondMinds managed to automate the process so well, that the deployment and scaling phases are as quick and easy across sectors and industries. The problems stay similar, the clients change. They are building a true
“AI as a Service”.
Creating Substantial Value with AI: BeyondMinds’ Approach to Scalable AI
A lot of AI projects nowadays suffer from bottlenecks in the manual work of data cleaning or model training. BeyondMinds’ approach is that if you want to be able to scale as an AI company, your core AI tech should be scalable. Therefore, they built a unique, highly-modular AI platform that allows them to reduce a significant part of the human labor and still reach state-of-the-art AI results. Here are some examples of the challenges that BeyondMinds solves to allow companies to move from AI POCs to Production at scale: Using their modular approach, they are able to implement AI systems that suit a variety of needs, and help their clients close the gap of moving forward from Proof-of-Concept to Production. To allow mass AI adoption, they are looking at AI models as one piece in a puzzle, and treat AI as a complete system, to create increased value over time. Their team understands that one critical issue is that AI Models work well in POCs and academia if the data is clean with a stable environment and under specific settings. In Production, however, these models hardly ever succeed. Handling noisy and highly dynamic data in production, working with a limited amount of available data for training, high maintenance costs, regulation for mission-critical systems, monitoring, security, etc. — all these can cause severe issues — and therefore must be an inherent part of the solution in production, which they are able to offer. This technology enables businesses to not only move beyond pilots and proofs of concept, but also be successful at achieving scale. In this way, the business transformation they achieve in their AI projects becomes extremely powerful.
Wish We Had Their AI Platform for the Due Diligence
After bootstrapping in its first year with contracts from big companies, BeyondMinds received the support of Prof. Ehud Weinstein and Dr. Ofir Shalvi, serial entrepreneurs who previously achieved huge success in deeptech with companies they’ve created, backed and supported.
All our concerns have been refuted throughout the DD process. We were afraid that they are merely a services company. We thought that they will not be able to scale their model. We hesitated that we found yet another startup who perhaps is over-confident in its ability to do that which no one else could.
In reality, they could. We saw that the company was able to deliver a product that works well for varied customers and creates value to all who came in contact with it, without the involvement of experts in the implementation and scaling phases. Some of these customers included super-savvy corporations with renowned internal data science teams. We witnessed how the company created automation for manual data engineering tasks (including labeling, cleaning data, and implementing expert insights without multiple consultations). We were no longer skeptical when we saw that their solutions were deployed and were working properly in a record time. A startup that built a modular AI platform for varied markets and needs, with the capabilities that we initially thought only Google’s DeepMind or OpenAI could achieve. An AI platform that lets them always stay up to date with the latest technological and academic innovations in a field that constantly develops. A step forward towards a truly Explainable AI system. An internal research hub, that keeps the models on par with state-of-the-art publications and achievements.
We also learned that BeyondMinds managed to attract large, sophisticated customers have impressive internal AI teams of their own, such as Microsoft, Samsung and KLA. This achievement is especially impressive for a startup at this early stage. We contacted some of their clients, got feedback, validated their claims, understood the use cases and how the BeyondMinds platform can be used.
Why We Invested?
To sum up, Rotem, Roey and the BeyondMinds team rose above dozens of other startups that we’ve met since they had a different approach to building a scalable AI organization that can address the core enterprise AI needs and identifies the most relevant and repeatable issues across industries.
In a very short time, the company already managed to gain extraordinary early success with some of the world’s leading customers. Recently, it also announced a series of executive appointments.
With an experienced, eager and exceptional leadership team, building a business alongside a research-hub that attracts the best minds in the field, we are positive that BeyondMinds can execute its go-to-market plans and become a great success story that enables large companies to generate significant value from AI.
This article originally appeared on the Grove Ventures Medium page. Visit Grove Ventures’ Medium for more columns and insights.
