Tom Taulli

According to data from PitchBook, the funding for AI deals has continued its furious pace. In the latest quarter, the amount invested came to a record $31.6 billion. Note that there were 11 deals the closed more than $500 million.

Granted, plenty of these startups will fade away or even go bust. But of course, some will ultimately disrupt industries and change the landscape of the global economy.

“To be disrupted, you have to believe the AI is going to make 10x better recommendations than what’s available today,” said Eric Vishria, who is a General Partner at Benchmark. “I think that is likely to happen in really complex, high dimensional spaces, where there are so many intermingled factors at play that finding correlations via standard analytical techniques is really difficult.”

So then what are some of the industries that are vulnerable to AI disruption? Well, let’s see where some of the top VCs are investing today:

Software Development: There have been advances in DevOps and IDEs. Yet software development remains labor intensive. And it does not help that its extremely difficult to recruit qualified developers.

But AI can make a big difference. “Advancements in state-of-the-art natural language processing algorithms could revolutionize software development, initially by significantly reducing the ‘boilerplate’ code that software developers write today and in the long-run by writing entire applications with little assistance from humans,” said Nnamdi Iregbulem, who is a Partner at Lightspeed Venture Partners.

Consider the use of GPT-3, which is a neural network that trains models to create content. “Products like GitHub Copilot, which are also based on GPT-3, will also disrupt software development,” said Jai Das, who is the President and Partner at Sapphire Ventures.

Cybersecurity: This is one of the biggest software markets. But the technologies really need retooling. After all, there continues to be more and more breaches and hacks. 

“Cybersecurity is likely to turn into an AI-vs-AI game very soon,” said Deepak Jeevankumar, who is a Managing Director at Dell Technologies Capital. “Sophisticated attackers are already using AI and bots to get over defenses.”

Construction: This is a massive industry and will continue to grow, as the global population continues to increase. Yet construction has seen relatively small amounts of IT investment. But AI could be a game changer.

“An incremental 1% increase in efficiency can mean millions of dollars in cost savings,” said Shawn Carolan, who is a Managing Partner at Menlo Ventures. “There are many companies, like Openspace.ai, doing transformative work using AI in the construction space. Openspace leverages AI and machine vision to essentially become a photographic memory for job sites. It automatically uploads and stitches together images of a job site so that customers can do a virtual walk-through and monitor the project at any time.”

Talent Management: HR has generally lagged with innovation. The fact is that many of the processes are manual and inefficient.

But AI can certainly be a solution. In fact, AI startups like Eightfold.ai have been able to post substantial growth in the HR category. In June, the company announced funding of $220 million, which was led by the SoftBank Vision Fund 2. 

“Every single company is talking about talent as a key priority, and the companies that embrace AI to find better candidates faster, cheaper, at scale, they have a true competitive advantage,” said Kirthiga Reddy, who is a Partner at SoftBank. “Understanding how to use AI to amplify the interactions in the talent lifecycle is a differentiator and advantage for these businesses.”

Drug Discovery:  The development of the Covid-19 vaccines—from companies like Pfizer, Moderna and BioNTech—has highlighted the power of innovation in the healthcare industry. But despite this, there is still much be done. The fact is that drug development is costly and time-consuming. 

“It’s becoming impossible to process these large datasets without using the latest AI/ML technologies,” said Dusan Perovic, who is a partner at Two Sigma Ventures. “Companies that are early adopters of these data science tools and thereby are able to analyze larger datasets are going to make faster progress than companies that rely on older data analytics tools.”

AI is becoming strategic for many companies across the world. The technology can be transformative for just about any part of a business. 

But AI is not easy to implement. Even top-notch companies have challenges and failures.

So what can be done? Well, one strategy is to provide AI education to the workforce. 

“If more people are AI literate and can start to participate and contribute to the process, more problems–both big and small–across the organization can be tackled,” said David Sweenor, who is the Senior Director of Product Marketing at Alteryx. “We call this the ‘Democratization of AI and Analytics.’ A team of 100, 1,000, or 5,000 working on different problems in their areas of expertise certainly will have a bigger impact than if left in the hands of a few.”

Just look at Levi Strauss & Co. Last year the company implemented a full portfolio of enterprise training programs—for all employees at all levels—focused on data and AI for business applications. For example, there is the Machine Learning Bootcamp, which is an eight-week program for learning Python coding, neural networks and machine learning—with an emphasis on real-world scenarios. 

“Our goal is to democratize this skill set and embed data scientists and machine learning practitioners throughout the organization,” said Louis DeCesari, who is the Global Head of Data, Analytics, and AI at Levi Strauss & Co. “In order to achieve our vision of becoming the world’s best digital apparel company, we need to integrate digital into all areas of the enterprise.”

Granted, corporate training programs can easily become a waste. This is especially the case when there is not enough buy-in at the senior levels of management.

It is also important to have a training program that is more than just a bunch of lectures. “You need to have outcomes-based training,” said Kathleen Featheringham, who is the Director of Artificial Intelligence Strategy at Booz Allen. “Focus on how AI can be used to push forward the mission of the organization, not just training for the sake of learning about AI.  Also, there should be roles-based training. There is no one-size-fits-all approach to training, and different personas within an organization will have different training needs.”

AI training can definitely be daunting because of the many topics and the complex concepts. In fact, it might be better to start with basic topics. 

“A statistics course can be very helpful,” said Wilson Pang, who is the Chief Technology Officer at Appen. “This will help employees understand how to interpret data and how to make sense of data. It will equip the company to make data driven decisions.”

There also should be coverage of how AI can go off the rails. “There needs to be training on ethics,” said Aswini Thota, who is a Principal Data Scientist at Bose Corporation. “Bad and biased data only exacerbate the issues with AI systems.”

For the most part, effective AI is a team sport. So it should really involve everyone in an organization. 

“The acceleration of AI adoption is inescapable—most of us experience AI on a daily basis whether we realize it or not,” said Alex Spinelli, who is the Chief Technology Officer at LivePerson. “The more companies educate employees about AI, the more opportunities they’ll provide to help them stay up-to-date as the economy increasingly depends on AI-inflected roles. At the same time, nurturing a workforce that’s ahead of the curve when it comes to understanding and managing AI will be invaluable to driving the company’s overall efficiency and productivity.”

Companies like C3.ai and Palantir have shown that selling AI technology can be quite lucrative. After all, these companies command significant market caps and are growing quickly.

Yet selling AI technology remains difficult. Customers often want customized solutions that are based on their unique data sets. There are also the issues of adoption. The fact is that many AI projects fail to go beyond the proof-of-concept phase.

Then what are some ways to sell AI technologies? Let’s take a look:

It’s Not About Platforms: Many AI vendors extol their “platforms” that can seemingly solve any problems. But this approach is usually off the mark. Let’s face it, there are already top platforms that have solid features and powerful ecosystems.

“Businesses should sell a solution to a problem,” said Muddu Sudhakar, who is the CEO and founder of Aisera. “Customers will buy a platform when multiple solutions are acquired and there are the right integrations. Customers don’t have free money sitting on the side just to invest in platforms.”

Sudhakar believes that an AI solution needs to show value within three to six months and there must be a return on investment within the first year.

Insights: It is often fuzzy as to what an AI system does. But for businesses, one of the most compelling aspects of this technology is about getting insights on tough questions. 

Take the example of eightfold.ai. Founded in 2016, the company is focused on leveraging AI for the talent management category.

“A lot of the problems that organizations are looking to solve are intimidating because they don’t always have a clear answer,” said Kamal Ahluwalia, who is the President of Eightfold.ai. “There’s no one second solution to ‘why can’t I hire the right people’ or ‘what skills do I need to be teaching my team so we’re ahead of the game in a few years?’ But the data is there, and it’s usually just not being looked at correctly, or it’s not feasible to do so for thousands or millions of times over. AI is all about applying that data, at scale.”

Avoid AI-Speak: AI is a complex topic. There are a myriad of terms like machine learning, hidden layers, deep learning, backpropagation and so on. Even people in the industry can have a difficult time explaining the concepts.

This is why it is critical to avoid the jargon when selling AI. “Rather than explaining the wonders of AI, it’s better to provide practical steps that you can take together with a customer to achieve the desired business results,” said Thomas Hansen, who is UiPath’s Chief Revenue Officer.

The Buyer Persona: AI is still in the early stages and many of the potential buyers are early adopters. This means that they have a strong understanding of their needs and a good sense of what’s available on the market. They are also more willing to use software that is 90% finished and then find ways to fill the gap, such as with custom coding or configuration. 

True, executives will still write the checks. But they will still rely heavily on the AI practitioners within the organization. 

“The AI buyer persona prefers a self-service and hands-on approach,” said Omed Habib, who is the Vice President of Marketing at Tonic.ai. “They don’t need to speak to a sales person and would rather sign up for an account immediately. They’re curious. They’re self learners. They love to tinker. So, the companies that have figured this out have created flawless self-service sign up for their software. They have excellent documentation. They have a vast library of videos to help enable their customers. They host a community forum for other users to share knowledge. In other words, they’re not just teaching you how to fish but they’re putting the best fishers in the same room to learn from each other.”

Data: Customers are understandably sensitive when providing access to their data to a third-party. This is why an AI vendor needs to have strong data policies.  

“It’s important that sales teams are educated on these features and understand the compliance and certifications that the software or organization holds,” said Sid Mistry, who is Appen’s Vice President of Marketing. “Company data is gold; you need to value and respect that.”

When it comes to presentations, Elon Musk rarely disappoints his audience. And this was certainly the case with this week’s AI Day. He made a variety of announcements, such as for a new 7 nm semiconductor, the Dojo supercomputer and innovations with computer vision. There were also some deep dives into deep learning. 

But perhaps the most interesting announcement was the Tesla Bot. This is on par with what we usually see on dazzling sci-fi movies. Yes, Musk apparently is building a humanoid robot.  It will be five feet, eight inches tall, weigh 125 pounds and have human-like hands. What will she/he/it do? Basically, the Tesla Bot will be our cyber slave, handling tedious and repetitive tasks.  For example, you can tell it to go to Chipotle and get a burrito—and it will happen. 

Sounds pretty good, huh? Definitely. 

But then again, over the years Musk has made some ambitious claims that have not been realized (remember his promises of fully autonomous cars or his robot taxi service?) And this could easily be the case again.

The funny thing is that Musk has a history of saying that AI could run amok and become an existential threat to humanity. Hey, he once tweeted: “If you’re not concerned about AI safety, you should be. Vastly more risk(y) than North Korea.”

But somehow, Musk thinks his version of AI will be just fine. We just have to trust him on this one, regardless of the federal preliminary investigation of Tesla’s Autopilot and the various lawsuits. Actually, even if the Tesla Bot somehow turns hostile, it will only be able to run five miles an hour.

Yet Musk’s Tesla has some big advantages to be successful in building the Tesla Bot. At the conference, he noted: “Our cars are semi-sentient robots on wheels. It kind of makes sense to put that on to a humanoid form. We’re also quite good at sensors and batteries and actuators so we think we’ll probably have a prototype some time next year that basically looks like this.”

But there is good amount of irony here. Musk’s demo of his Tesla Bot was actually a person in a white spandex outfit and a black mask. He or she looked more like a go-go dancer.

Now if the CEO of Ford or GM tried this stunt, he or she would have been laughed out of the room and mocked savagely on social media channels. It would be a downright embarrassment. 

The Impact

Even if Musk may be overly optimistic on the timeline for humanoid robots, it does seem like we could see some true breakthroughs during the next decade. And the impact on the world will be significant. “These systems could be used to aid human labor in hazardous areas like mining and manufacturing, reducing overall safety incidents and saving lives,” said Michael Levy, who is a Senior Analyst at Harbor Research.

But the bots could also mean having to rethink some of the core fundamentals of society. In other words, what will “work” really mean and what will become of capitalism?

“Elon Musk touched on it and I agree,” said Dr. Jesper Dramsch, who is a machine learning expert and works at the ECMWF. “A lot of physical work will be optional. Tasks like shelf-stocking may be completely obsolete. As a society this means we have to move away from the concept that we trade our time directly for income and seriously consider universal basic income and social structures that go beyond the scarcity economy of pure capitalism.”

Last December, Upstart launched its IPO and raised about $240 million. On the first-day of trading, the shares jumped 47%.

But this was just the beginning of the gains as the IPO would soon become one of the top for the past year. The return? About 800%.

Then again, the company is a high-growth fintech company that has effectively leveraged the power of AI. It’s focus is on partnering with banks to provide a much better way to score the risks and automate the tedious processes for issuing and managing consumer loans. 

The CEO and cofounder is Dave Girouard, who built the billion-dollar apps business for Google. He had also served as a Product Manager at Apple and an associate in Booz Allen’s Information Technology practice.

As for Upstart, Girouard’s main focus is to upend the banking industry’s reliance on the FICO score. 

“The Upstart system uses AI and machine learning models with 1,600 data points and 15 billion cells of data to improve accuracy in terms of identifying and measuring credit risks,” said Phat Le, who is an Associate at Harbor Research. “Some of the variables that Upstart considers are employment history, educational background, banking transactions, cost of living, and loan application interactions.”

For the most part, Upstart is reducing the inefficiency with the traditional FICO approach. After all, about 80% of Americans never default on their loans yet only 48% have access to loans at prime rate. The result is that good borrowers often pay premiums rates while many other borrowers get loans when they should not. 

Granted, when it comes to AI, there can certainly be major issues. There is the potential for bias and discrimination, such as when the data is skewed. Yet Upstart has made great strides in addressing the problems.

“In 2017, the company was the first to receive a No Action Letter from the Consumer Financial Protection Bureau (CFPB), which was renewed in November 2020,” said Mike Raines, who is the owner of Raines Insurance Group. “According to Upstart, ‘the purpose of such letters is to reduce potential regulatory uncertainty for innovative products that may offer significant consumer benefit.’”

Keep in mind that one of Upstart’s banking partners has recently eliminated any minimum FICO requirement for its borrowers. And this is what Girouard had to say about this on his earnings call: “To us, this demonstrates both a commitment on behalf of this bank to a more inclusive lending program, as well as an increasing confidence in Upstart’s AI-powered model. While credit scores can be useful, hard cutoffs based on a three-digit number invented 30 years ago leaves far too many creditworthy Americans out in the cold.”

The Upstart strategy has certainly resulted in staggering growth. In the latest quarter, the revenues soared by 1,308% to $194 million and the transaction volume came to $2.80 billion, up 1,605%. The company was even able to generate a net profit of $37.3 million, up from a loss of $6.2 million in the prior year. 

To expand its addressable market, Upstart has acquired Prodigy, which has allowed the company to move into the lucrative auto lending space. Based on the latest earnings report from Upstart, the U.S. personal loan originations are about $84 billion and they are $635 billion for auto loans. 

But interestingly enough, Upstart really does not need to look further than these two categories anyway. As Girouard noted on the earnings call: “[W]e just see a lot of opportunity out there. We don’t think credit is a solved problem almost anywhere in terms of people getting rates that makes sense for them based on their true risk. So you will definitely see us move beyond personal loans and auto, but frankly, we have so much uncharted territory, even in those two categories, we’re not in a particular rush to do so.”

Data is certainly the fuel for AI. Yet there is a source of valuable data that usually does not garner much attention. It is from mainframe systems. They hold enormous amounts of data—which go back decades—for mission critical operations.

But then again, there are difficulties working with mainframes and AI. “The biggest challenge is the lack of compatibility of emerging technologies,” said Chida Sadayappan, who is the Cloud AI/ML Offering Leader at Deloitte Consulting LLP.

But the benefits of AI are too important to ignore. So what can be done? Well, one strategy is for leveraging cloud platforms outside of the mainframe environment. 

“New approaches to cloud migration replace the traditional ETL (extract, transform, load) approach with a more modern ELT (extract, load, transform) approach that moves mainframe formatted data directly to any object storage target before using the target platform to transform it for use in AI applications,” said Gil Peleg, who is the CEO of Model9. “This pioneering method adds mainframe data to data lakes quickly, easily, and securely so leaders can maximize the ROI of their cloud BI and analytics applications.”

But like any IT effort, there needs to be a clear-cut plan and the goals must be achievable. The reality is that AI efforts can take considerable time to generate ROI.

“Companies should be aware that these types of initiatives don’t always cut costs,” said Sudhir Kesavan, who is the Global Head of Cloud Transformation at  Wipro FullStride Cloud Services. “They may have the opposite effect, so having them be business-led can help overcome the challenge of business benefits seeming less tangible to start.”

IBM Mainframes and AI

The capabilities of mainframes have been evolving quickly. For example, IBM has been retooling its Z system for AI and this has involved the integration with many common open source platforms like Spark, PyTorch, Keras, and TensorFlow.

“We are enabling our clients to embed AI into their mission critical enterprise workloads and core business processes with minimal application changes and giving them the ability to score every transaction while meeting even the most stringent SLAs (Service Level Agreements),” said Elpida Tzortzatos, who is an IBM Fellow and the Chief Technology Officer of z/OS.

By generating the AI insights on Z, this allows for real-time responses at the point of interaction, which can be critical for applications like fraud detection. There is also a major security benefit because sensitive data is not moved.  

Leveraging AI For Mainframe Environments

The power of AI for mainframes does not have to be about creating projects. For example, there are emerging AIOps tools that help automate the systems. Some of the benefits include improved performance and availability, increased support speed for application releases and the DevOps process, and the proactive identification of issues. Such benefits can be essential since it is increasingly more difficult to attract qualified IT professionals.

According to a recent survey from Forrester and BMC, about 81% of the respondents indicated that they rely partially on manual processes when dealing with slowdowns and 75% said they use manual labor for diagnosing multisystem incidents. In other words, there is much room for improvement—and AI can be a major driver for this.

“Mainframe decision makers are becoming more aware than ever that the traditional way of handling mainframe operations will soon fall by the wayside,” said John McKenny, who is the Senior Vice President and General Manager of Intelligent Z Optimization and Transformation at BMC. “The demand for newer, faster digital services has placed increased pressure on data centers to keep up as new applications come online, the volume of data handled continually increases, and workloads become increasingly unpredictable. In today’s fast-paced digital economy, this creates a perfect storm of higher customer expectations, faster implementation of an increasing number of digital services, and a more tightly connected mainframe supported by a less-experienced workforce.”