Looking into B2B AI Startups
Since the advent of the Age of Automation, companies have been racing to implement machines that help maximize how much labor one worker can perform. However, within the last decade the challenge has shifted to how much labor zero workers can perform. What the previous century’s engineers could only imagine is now made possible by Artificial Intelligence (AI).
Every startup founder knows that in order to vertically disrupt an industry, the incoming company must be able to offer either a superior product or a lower price, with the ideal scenario involving both. Equally obvious is that achieving and maintaining this competitive edge requires some kind of efficiency advantage. That is to say, lower costs allow the company more leeway for reinvestment or simply to offer the product at a lower price. Innovations in production, customer acquisition, or supply chain and distribution are some of the more common areas where startups can gain an advantage, but the underlying enabler (and the very definition of efficiency) is labor-saving technology, also called automation.
Traditionally, the labor-saving aspect of automation was concentrated in trivial tasks. Now, with the development of AI, we are seeing machines improve efficiency in increasingly complex processes. Given the powerful yet versatile disruptive capabilities of AI, businesses of all sizes and industries are demanding B2B AI solutions for their niche efficiency needs.
What’s more, Sweden is uniquely positioned to spearhead the next stages of the AI revolution. Generous government spending on social safety nets such as unemployment benefits encourages entrepreneurs to take on the risk of researching and developing new AI applications. Of course, these incentives also benefit startups in other non-AI fields (which is why Sweden is one of the world’s most entrepreneur-friendly environments), but AI innovators gain a secondary source of support: public opinion. Again in part due to comprehensive welfare, Swedish AI entrepreneurs are blessed with a domestic population that supports rather than fears technological change and workplace automation.
The important consequence isn’t limited to a positive public reputation; the Swedish government is also on board, which entails favorable policy measures. Sweden’s official AI strategy includes both funding and infrastructure for startups pursuing AI projects through the Vinnova innovation agency in addition to fostering collaboration between university-sponsored research and the private sector. Less concretely, the overall outlook is highly supportive of sustainable integration of AI into the economy but simultaneously cautious of the ethical implications of potential physical harm caused by technologies such as self-driving cars or surgery robots. Thus far, Sweden has formed the Committee for Technological Innovation and Ethics, as well as a few other initiatives, but has not passed formal legislation regarding these concerns.
Swedish policy instruments by budget. Each black square represents one instrument, for a total of 7 policy initiatives. (Source: OECD) see full list of initiatives here
One point of uncertainty, however, is the recent European Commission's plans to impose strict regulations on “high risk” AI applications. In agreement with Sweden’s strategy report, the EC identifies transport and healthcare as areas of high risk, also including criminal justice on the list. The good news is that there doesn’t seem to be any intention of outright banning the technologies; instead, the goal is to establish oversight and accountability measures to ensure continued public support, which could be jeopardized in the event of a large accident. Loss of faith could touch off another AI winter, which in the 1980’s brought massive funding cuts to the industry.
In any case, where there is public support, there will be venture capital funds. Last year, a total of $4.9 billion USD was invested into European AI companies, surpassing all other technology subcategories, with the majority of funds coming from the United Kingdom, France, and Germany. Although this figure pales in comparison to the United States and China, this represents a consistent trend of massive growth, more than tripling the $1.5 billion USD raised in 2017.
European VC activity in artificial intelligence and machine learning (Source: Pitchbook)
As the figures imply, the AI and ML industry as a whole is experiencing some impressive growth globally. At the same time, however, not all AI companies are created equal. This relatively new technology especially thrives on innovation, resulting in a diverse spread of products and business strategies. Popular D2C applications include smart home appliances such as refrigerators, air conditioners, and the iconic Roomba vacuum. But AI is destined for greater things than mere housework; companies in many different industries are already beginning to rely on AI to enhance their products and services. Furthermore, individual consumers currently lack both the need and infrastructure to demand most of the more complex AI systems, hence this article’s focus on vertical B2B applications.
Agriculture— In order to keep pace with the world’s rapid population growth, producers in the agricultural sector are eager to purchase technologies to boost efficiency. Thankfully, new AI predictive simulations are able to make predictions on future crop yields in addition to monitoring the health and ripeness of existing crops. The agricultural robots subset includes drones deployed for planting, harvesting, and weed/pest control and is expected to reach a global market size of $11.58 billion USD by 2026.
Entertainment— Most subscription-based video and music streaming platforms such as Netflix and Pandora already employ AI to make recommendations using data on viewing habits. Computer vision can also automate content monitoring, scanning for inappropriate scenes or logos and copyrighted audio. Publishers and broadcasters have experimented with AI input on what kind of content viewers want to see at a given time of day, as well as automated reporting on sports and financial statistics. After an AI program almost won the Hoshi Shinichi Literary Award, the prospects for automatically-generated creative content, from e-books to music to cinematics, have never looked brighter.
Finance— Trading stocks now feels less like gambling and more like checkers with the applications of AI in data mining. BlackRock Inc.’s Aladdin is capable of not only reading and processing numerical trends but also analyzing emotions in public responses to company reports. Goldman Sachs has a similar engine called Kensho to quickly identify and exploit security mispricing. Automated portfolio construction and management is also rising in popularity. Presently, the fast-paced day trading is almost entirely left to computers. Because the tasks in financial analysis are already well-defined, innovations in this industry will likely rely on quantum computing for speed and accuracy
Healthcare— While D2C personalized healthcare can offer consumers basic diagnostics, few households are likely to implement artificial neural networks for further treatment and consultation. Hospitals and other medical providers are hoping to automate a wide variety of processes including but not limited to computer vision detection of anomalies in CT scans, data mining medical records to determine trends, designing treatment plans, prediction of surgery success, monitoring spread of infectious disease, and even synthesizing drugs.
AI in Healthcare Market Size by Region in Billion USD (Source: Markets and Markets)
Understandably, healthcare is highly regulated, so some of the more ambitious applications may not see their heyday for the foreseeable future. Nevertheless, the industry is indeed moving steadily forward with innovations such as the IDx diabetic retinopathy diagnostic system receiving approval from the U.S. Food and Drug Administration in 2018.
Manufacturing— Faced with the classic “drones crashing into each other” problem, factories and distribution centers are demanding AI systems to enhance the ability of existing robots to cooperate with both human workers and other machinery. Other areas of interest are the application of computer vision to detect microscopic defects in small products such as circuit boards and predictive analysis of equipment maintenance needs. The AI-in-manufacturing market is expected to exceed $15 billion USD by 2025.
Non-vertical Miscellaneous— Nonspecific applications deserving of honorable mention include marketing, customer success management, and human resources, all of which can make great use of chatbots and predictive analysis of individual preferences. AI can also be used in any industry to forecast trends in market demand and supply chains. Lastly, companies may opt to protect their data and online infrastructure with cybersecurity AI, which can automatically detect and fix vulnerabilities. These horizontal technologies tend to be less concentrated due to their widespread general usage.
Finally, let’s take a look at a few Swedish AI startups using the B2B model:
Mapillary (2013)
Acquired by Facebook. Funding: €22.3 million (Seed 1.5M, A 8M, B 15M)
Mapillary uses computer vision to generate street-view imagery and other augmentations to standard satellite maps. Customers include various municipal government agencies, infrastructure construction companies, trucking companies, and navigation apps. Additionally, the OpenStreetMap function allows for collaboration between individual users. One of the current projects involves adding hospital information to help identify which locations are in need of staff or PPE at any given time.
PerceptiLabs (2017)
Funding: N/A (currently in Seed stage)
PerceptiLabs provides machine learning developers with a user-friendly interface to monitor and document the predictions, accuracy, and loss of the generative adversarial network. The target customers— developers, engineers, and data scientists— can also access the TensorFlow code to further adjust parameters.
Funnel (2014)
Funding: €61.6 million (Seed 3M, A 10M, B 54M)
While the Funnel package provides various market research data management tools, it also includes an AI analyst called Data Core to condense, format, and analyze the massive amounts of data. This application is functionally a financial tool but could be used by business managers in various industries.
Match2One (2014)
Funding: €3.4 million (Seed 2M, Early VC 1.5M)
Another AI application in finance, Match2One analyzes customer responses to the customer’s advertisements and gives advice in generating leads from prospecting and retargeting, as well as the overall campaign strategy. The machine learning engine Clara also predicts who would most likely purchase the user’s product and shows the ad to them.
By no means are either of these lists exhaustive. The most wondrous aspect of the AI industry is its rapid, often unpredictable development as entrepreneurs continue to bring to life what was previously dismissed as science-fiction. Efficiency is a universal language among all industries, and the countless different angles from which to approach the problem guarantees that AI will be the next frontier in the B2B space.
To be sure, both founders and investors ought to carefully watch the state of regulations and public opinion in order to avoid the dangers of outpacing either. While political briefings are certainly an important source of information, staying up to date on academic literature can provide advanced knowledge, since, in the absence of existing regulations, governmental agencies will rely heavily on expert opinion from universities and think tanks. In contemplating the trend of articles expressing caution towards certain applications such as big data and biometrics, the EC’s current move does not seem surprising. If founders are unsure of their own judgment on the appropriateness of their technology, they may want to consider submitting a proof of concept for review before further investing in R&D.
Fortunately, it would appear at present that AI technologies that do not directly affect the physical world (most applications do not, with the main exceptions being healthcare, transportation, and law enforcement) are fair game from a regulatory standpoint. There will always be risk involved, but now is the time for founders to embrace the risk and rise to the occasion. Now is the time for founders to say “Hello, World!”