Good afternoon! This week, we're focused on the hot enterprise AI market, so we asked the experts to tell us about signals they look for when deciding whether a particular AI use case will be successful or not within a business process. Questions or comments? Send us a note at email@example.com
CTO at Scale AI
In the fullness of time, everything will be transformed by AI. For enterprises, AI technology needs to meet a certain speed and accuracy to be operationally viable.
Traditionally extracting high-quality text from complex documents, like those used in logistics or financial services, has been extremely difficult. At Scale, we believe the combination of transformers, transfer learning and self-supervised learning has led to a breakthrough. We're now seeing very fast, high-accuracy extraction work in operationally intense environments.
In the industrial robotics space, co-bots – industrial humanoid robots working alongside humans – get a fair bit of interest and hype. Eventually, the generality of these solutions will be transformative, but for the foreseeable future, the economics will continue to favor deploying use case-specific robotics that are safety-rated for the specific task.
The power of AI in the enterprise is undeniable. AI is poised to transform virtually every industry, and there is tremendous potential when it comes to bringing powerful AI processing to the edge.
Today we're seeing a shift in the enterprise from a centralized AI model to a more distributed model. More data is being generated at the edge than ever before. Data compliance laws and low latency requirements for data-generating IoT applications like autonomous vehicles mean it is no longer viable for all AI processing to occur solely in a centralized public cloud. Cost is also another factor – with AI at the edge, you don't have to backhaul large data sets to be processed at the core. Since AI and machine learning can require access to data spread across clouds, data centers, brokers and edge devices, it is increasingly important to host AI infrastructure at a location with high speed and secure connectivity to these data sources.
Many IT leaders are weary of what AI will actually be able to deliver in terms of quantifiable business results, and they want to deploy AI solutions in a way that can create value quickly. As hybrid cloud, software and APIs continue to advance, deploying AI solutions in a streamlined way will help businesses across many industries realize the full potential of AI.
Gartner predicts that AI is still in its two- to five-year hype cycle.
So, we are not yet at the point where robots are going to take over the world. However, computer scientist Andrew Ng boldly calls "AI as the new electricity." Every industry — education, retail, manufacturing, energy, health care, technology — all have great opportunities that will transform their landscape.
With that, I would not call out one specific use case as having the most significant potential, but instead, recommend a focus on an "Augmented Intelligence." You need not splash huge investments to explore brand new innovations, but rather look at your most challenging business process that can benefit from "transfer learning of ML models" — reuse already-trained and readily available ML models for common business problems. This compelling yet straightforward approach of recycling models will enable us to automate the most demanding business use cases like demand pricing, customized product recommendations, personalized marketing, fraud detection, email spam filtering, demand forecasting, pattern recognition, social listening and many more.
At HPE, we are very bullish about enhancing our omnichannel customer experience through a single unified data layer as we pivot into our "As a Service" transformation journey. We're equally excited about our journey on enhancing our integrated demand pricing and inventory management using telemetry data.
For that, we need to start with a robust data-first mindset strategy that will enable business-use data foresight to lead the outcome!
AI and enterprise software are intrinsically linked in a way that gives SaaS an amazing advantage over legacy software. Data confers power in AI, and SaaS companies produce tons of data that gives them visibility into how their users interact with their software. That data can be used in many ways, whether it's AI that uncovers customer patterns revealing why a set of users is more productive, drives a higher quota or is more likely to churn, or data that helps refine their operational approach.
One of our newest portfolio companies, Vivun, is a great example of this process in action. The company's presales platform uses AI to create a "Hero score," that helps presales and sales teams understand the likelihood that a deal will close and decide what to address. The score uses signals from the platform's users to optimize a predictive model that gets smarter over time with more data. The company was built from the ground up, focusing on what I refer to as "AI-first SaaS." It's in the DNA of the company and the core ingredient of their secret sauce, not just a marketing tagline.
For every company like Vivun, there are dozens that bolt on AI after the product has gone to market. These commonly cursory applications of AI drive the feeling that AI is overhyped in the software market. Any VC investing in this space has to filter that out.
Many software companies claim to do something with AI, but we're only interested in those where AI is instrumental to the product's capability. They have the power to unlock a new category in a powerful, distinctive way. Investing in AI as a marketing message is a fool's errand.
The times I've seen AI have the most impact have been the most ordinary circumstances. The truth is, there are still so many things humans do that could be better aided by computers. Think about looking across many numbers, for example in sales opportunities. There's valuable information there, but people should be looking at a summary of that data, not trying to read and understand every data point of every opportunity. AI works best when you find the tasks that people have to do every day to get to more interesting work. Take out the "have to do" tasks and let people get right at what people do best: connecting, reasoning, synthesizing.
When I've seen AI fail, it's when it's applied to an imagined scenario that could be interesting one day but isn't what people are doing now. The potential in AI today comes in the most ordinary tasks, and usually the ones that are frustrating, repetitive and time-consuming.
Today, we've seen AI deployed across a wide spectrum of use cases to solve business problems — from gleaning new insights into human behavior to identifying and responding to cyber threats while helping to guide discovery of COVID-19 vaccines.
AI has increasingly been integrated into the fabric of business. As an example, during the pandemic, intelligent bots were put to work for major airlines to speed the processing of customer cancellations, automatically creating an e-voucher option giving customers faster credit. Intelligent automation has also enabled hospitals to manage oxygen levels for patients on ventilators. And in banking, AI enabled one financial institution to complete nine years of work in just two weeks.
However, while AI substantially transforms organizations, there are still barriers to scaling its benefits and a realization of the need to always keep humans at the forefront of any decision making. Automation Anywhere has brought focus and intent to our mission by creating human-first governance in the form of a cross-company, global Responsible AI Council. The council aims to guide how our company brings inclusivity, accessibility, change management, upskilling and reskilling to the human workforce as they transition to automated processes —enabling them to leverage these state-of-the-art AI technologies that will help unleash even more human potential.
I believe the greatest potential for AI continues to lie within the workforce itself. The pandemic highlighted the importance of creating efficiencies to keep organizations resilient and productive while ensuring supply chains remained undisrupted. Intelligent automation, which includes AI and robotic processes automation, will continue to play a significant role to automate the mundane, unleash human creativity and ultimately make work more human for billions of knowledge workers.
Kevin McAllister ( @k__mcallister) is a Research Editor at Protocol, leading the development of Braintrust. Prior to joining the team, he was a rankings data reporter at The Wall Street Journal, where he oversaw structured data projects for the Journal's strategy team.