At my firm, Wing, we recently completed a research report looking at the young companies pioneering AI-powered “data-first” applications, and there is a lot of important activity happening in this space. Collectively, these startups have raised $1.8 billion of venture capital to pursue their goal of disrupting the $150 billion-plus market for enterprise applications.
In the same way robotic exoskeletons greatly increase the strength of their human operators, these software equivalents can help workers achieve far more than they can with existing applications.
The range of areas being addressed by this new generation of software is growing fast. Gong and Chorus, for example, are startups using natural language processing and other AI technologies to analyze sales conversations, highlighting ways to make sales teams more effective. ClearMetal is applying machine learning to data about ship movements, port conditions, and other factors to help supply chain managers better predict the availability and cost of containers. Other young companies like Prevedere and TravelBank are leveraging AI technologies to help executives improve financial forecasts and expense management.
The original business software paradigm pioneered by companies such as Oracle and SAP focused on using software to boost efficiency by codifying workflows. The data generated by the applications were treated as an afterthought. If insights were needed to guide workers, analysts were summoned to extract them from the data using specialized tools. The Software-as-a-Service revolution greatly improved the way business software was delivered and distributed, but it didn’t fundamentally change this underlying approach.
The new applications do. They treat data as the priority — hence the term “data-first” — and they leverage various AI technologies to recommend actions that are delivered directly to users, often in real time. Incumbents like Oracle are now rushing to inject AI features into their software. But as so often in the history of technology, many of the most compelling innovations will come from startups unencumbered by legacy thinking.
Startups like Signifyd, which just raised $56 million of fresh funding to develop its anti-fraud application, are focusing on key online processes like fraud detection, ad targeting, and consumer offer optimization, where the sheer volume and velocity of decision-making is so high it’s impossible to have humans in the loop. Using AI technologies to automate all or most of the work makes sense here.
Elsewhere, as our report notes, the emphasis is more on augmentation than automation. This is especially true in sensitive areas like IT security and IT operations, where a tsunami of events and alerts risks overwhelming human operators. Here, AI-powered software is being used to identify situations that merit the attention of an employee and to automate responses to some issues. For example, Moogsoft (in which my firm is an investor) has enabled one of its financial services clients using its “AIOps” platform to boost the number of servers an IT operations team member can monitor effectively from just over 1,000 to over 10,000.
Data-first applications are also helping to increase employee productivity in other sensitive areas like customer support. DigitalGenius applies deep learning algorithms to historical customer service transcripts and can generate automated responses to new queries or suggest ones for call center agents to use. By analyzing inbound queries and ensuring they are routed to the right teams, DigitalGenius claims it can increase agent capacity by around 30 percent.
Putting the new software to the test
As a recent New York Times story trumpeted, hiring is up, wages are up, and jobless rates are at a 10-year low. As the labor market continues to tighten in America, companies need to find innovative new ways to boost the productivity of their employees.
Those that fail to adopt data-first applications will leave their employees at a distinct disadvantage. Those that want to be in the vanguard will need to evaluate next-generation software applications quickly. Some of the criteria against which to assess them include:
- The distinctiveness of their data sources. The type and quality of data used to train an application’s embedded AI helps determine how effective it will be. That’s why many startups are putting considerable effort into developing unique data resources. Citrine Informatics, which helps accelerate R&D and manufacturing processes by applying advanced AI to materials development, has created a significant database of its own by mining patents, research papers, technical reports, and other sources for information.
- Their capacity to process and analyze large volumes of data. Data-first applications need to be able to ingest very large volumes of data and deliver recommendations for action in real time. That’s why the applications are often built on fast and highly scalable data architectures that use open source processing frameworks like Apache Spark and data lakes like Hadoop.
- Their ability to generate valuable “synthetic” data. Data-first applications aren’t just data-driven; they’re data-driving too. As well as ingesting data for analysis, the results they produce become important inputs to other algorithms, which deliver additional value-added insights. HiQ Labs, a data-first HR application, is a good illustration of this virtuous data cycle in action. Its Keeper product applies algorithms to publicly available data about a company’s employees to assess who might be at risk of leaving. Some of the information it generates is a useful input to HiQ’s Skill Mapper solution, which helps HR teams better understand the skillsets of their workforces.
- The quality of their data science teams. Most executives won’t have a deep enough understanding of advanced AI to assess whether a particular machine learning algorithm is the right way to tackle a problem. But they can do thorough due diligence on the data science teams behind the data-first applications they are considering. A number of the founders of the startups we looked at come directly from universities like MIT and Carnegie Mellon, which have strong AI research programs, or previously worked at agencies like NASA and the NSA that are users of advanced AI.
Human + data-first machine
As they penetrate more companies and business functions, data-first applications will inevitably find themselves swept up in the debate about whether AI-driven automation is going to destroy large numbers of jobs.
Yet in most cases, the new applications’ role will be to automate basic tasks and help employees become more productive as they are freed up to focus on more value-added activities. This ability to act as a force multiplier for humans will become even more valuable as ageing populations place a greater burden on remaining workers to drive future economic growth.
As a society, we’ll need to think carefully about the implications of software-driven automation in certain sectors of the economy. But the direction of change is clear. The companies and executives that are most adept at building machine-assisted workforces will be the ones that come out on top in the emerging era of AI-driven competition.
Martin Giles is a Partner at Wing Venture Capital.