Skip to content

Our Privacy Statement & Cookie Policy

All Thomson Reuters websites use cookies to improve your online experience. They were placed on your computer when you launched this website. You can change your cookie settings through your browser.

AI Experts

HPE’s head of AI: Educating your employees (and yourself) on the benefits of AI

Tomorrow belongs to those who heed the call to get smart about artificial intelligence, says Seyed Mirsepassi of Hewlett Packard Enterprise.

With all the buzz these days about artificial intelligence, it is tempting to fall into traps of existential crisis. Putting aside the science fiction-driven scenarios of robots run amok, there are more practical considerations an AI-empowered tomorrow will mean to the commercial landscape. Business leaders may worry that their enterprises will drift away from their core competencies, forcing them to become more like AI tech companies or risk being left behind; and amidst their workforces, employees may grow increasingly anxious that they will be replaced by swarms of data scientists, algorithms and bots.

Not so fast, says an AI expert who has been working in the digital space for roughly two decades. To get his take on how organizations really need to approach the burgeoning AI revolution, we connected with Seyed Mirsepassi, Executive and Head of Artificial Intelligence for Hewlett Packard Enterprise. Seyed’s resume includes a post as the founder and CEO of Neugenetics (an AI platform and solutions provider) as well as positions with IBM’s Watson team, GE Digital and Autodesk.

According to Seyed, while it is certainly advisable to increase your knowledge of what AI is and how it works, you (and your business) don’t have to know all there is to know about AI; rather, be expert at your particular domain and simply learn how to let AI compliment your operations.

“When I look at the projects that I’ve been involved in or that I know intimately, the successful ones were rarely the result of one company having such deep technical AI knowledge and people that other companies didn’t. I’m not saying that is necessarily true in all cases, but for the most part it is business and domain knowledge (and not AI knowledge) that has initially been a critical part of the project.”Seyed Mirsepassi, Hewlett Packard Enterprise


ANSWERS: How do you see AI impacting the human capital needs and business strategies of organizations? How are those going to evolve over the next decade?

MIRSEPASSI: One of the things I’ve seen that mid-size and even large companies are going through is trying to achieve a balance of AI skillset within the company and externally. Companies are struggling with human capital questions around AI development; if you want to hire data scientists, they’re very expensive and hard to find. Lots of companies seem to be asking, “Do I need to have an army of data scientists? I’m not an AI company. I’m a manufacturing company, a healthcare company, etc.”

I think for the most part the answer is no, but there is a balance because eventually AI is going to be a core part of every part of the business. You would want to own and keep the intellectual property associated with it internally, as this has the potential for being a competitive advantage for your company. You want to have some control over technology that is created internally. I think there is going to be a balance between what you have internally and what you are augmenting it with externally.

Now, one thing that is alarming to me right now is the disparity between executive interest in taking action to educate and retrain their workforces on AI, versus the actual appetite of the employees themselves to be educated. I was reading several surveys that showed leadership interest in educating their teams on AI matters was only around 30-35%, whereas most of the people (almost 70%) who serve those companies do want to be educated on this. It’s very lopsided.

In the next five years, they will all need to know more about AI. This doesn’t mean that everybody needs to become a data scientist but it is a new tool, much in the way at one time we didn’t use the internet or even a computer before, but now we do.

There is obviously going to be a transition in many areas, especially areas like manufacturing, where you need a different set of skills to work within an environment where AI and robots are part of the team. It becomes more and more a collaborative environment. What students are going to be doing in the workforce is going to be different than yesterday. I think that companies who are going to be leading this area 10-15 years from now are the ones who are taking action today to educate or retrain their workforce.

ANSWERS: As companies become more invested in an AI future, do you think we’re going to see more data sharing partnerships between organizations to feed their AI engines?

MIRSEPASSI: I would like to say yes. I’m an optimist, although the practice up to now certainly has been limited in terms of success. I’ve talked to many, many CEOs and executives over the last few years, and they all say, “Everybody tells me that data is the new oil, the new currency, and that I need to not share it with anyone.” I used to work for a company where I was an employee of the company; I was trying to access data from another group within the company itself, and that was difficult to do.

I think we’re still at that stage where some of the reluctance towards sharing is warranted, but some of it is not warranted; some of the data can actually be shared. Where I’ve seen some success in this area is when there is a third party company that’s providing some level of AI expertise, and they bring several industry players together and provide a platform where data can be shared. I think in the future, we will see this model more and more.

As we are all becoming more educated about data and governance and security, I think you will see a greater awareness that with better quality and more diverse data, you can have better models and better insight. Essentially from a business needs perspective it would be another element that forces companies to come together and do collaborations. I think those things need to come together.

In the short term I’m not as optimistic, but I think in the more mid to long-term I am more optimistic because we’ve seen it in other areas where new technologies initially didn’t have enough traction. The need and desire for companies to come together and build these partnerships to share data that all benefit from will increase.

ANSWERS: In what key business areas do you see AI improving revenue performance in the near future?

MIRSEPASSI: It depends on the industry, but what I’ve seen that can provide immediate benefits are areas where you do not need to invest heavily in a lot of infrastructure or new technology because companies want to stay away from that. One example would be conversational AI (i.e. chatbots). There are small companies that have been able to create a solution around conversational AI, and they then approach companies with the pitch, “If you have a corpus of data somewhere in your help desk, call centers, etc., we can quickly ingest your data, analyze it and build an intelligent interface (chatbot) to converse with your customers through a SaaS model.” What this means is that the company doesn’t have to do much in terms of support, development and infrastructure; the chatbot provider can very quickly look at that data and then translate that into customer insight.

You can also use AI to do some initial monitoring and diagnostics (and even predictive capabilities) in more industrial and manufacturing areas, allowing you to very quickly gain some insight without a lot of risk. I think those are areas that companies should explore because they can build off what other companies and industries are doing at a much lower risk than trying to reinvent the wheel or go into a new area where there’s a higher level of risk.

Generally speaking, roughly 70% to 80% of companies right now are at a very early stage in AI adoption. They’re trying to figure out what are the business areas they want to focus on. For the most part, I would say that looking at gaining operational issue efficiencies would be the least risky area to focus on. (An exception may be if you’re in retail and you’re being forced into digital transformation in how you serve your customers; this is a big area.) I think that for large manufacturing and industrial projects, AI can help make your operations more efficient and with less unplanned downtime. I think those are going to be some of the areas where you’re going to be much more successful in deploying AI to see business value right now.

ANSWERS: How can organizations ensure their investments into AI/machine learning pay off and provide an advantage over the competition who might also be exploring the same technology?

MIRSEPASSI: When I look at the projects that I’ve been involved in or that I know intimately, the successful ones were rarely the result of one company having such deep technical AI knowledge and people that other companies didn’t. I’m not saying that is necessarily true in all cases, but for the most part it is business and domain knowledge (and not AI knowledge) that has initially been a critical part of the project.

A lot of times, AI projects are started by purely technical people. If you look at data scientists by and large, most are coming from a computer science background with not much domain knowledge. Also, as you get to certain domains, the expertise needs to be deeper. Usually, projects are not successful, from a business perspective, because there hasn’t been enough domain expertise built into the use case. Therefore, there is a higher percentage of failure in those projects. Then, your competitor (if they’re doing it in the right way by getting the entire team involved from the get-go and having executive sponsorship) can have a better result.

The other part of it is being able to bring in the right external expertise. A lot of times, companies want to do AI projects only with internal resources. But maybe they don’t have the right expertise internally to bring it all together. While there is a lot of science to behind AI and data science, there is also an art piece to how you build your model and make it fine-tuned so that it works with your entire system. If you don’t have the internal expertise to get your model to behave really well in an iterative way, it could work against you in your projects.

ANSWERS: What should artificial intelligence developers and researchers be tackling now that they may be largely missing out on?

MIRSEPASSI: There are many areas where developers could focus their attention; however, I would like to focus on these three areas:

One area of focus should be to build AI applications that can solve a large set of problems without requiring a huge investment from the end customers. Right now, on one end of the spectrum you have what I call horizontal platforms, where large tech firms employ a lot of resources to build a set of AI tools that can be utilized by other engineers and data scientists to build a specific AI application. On the other end of the spectrum, you have small companies that are very smart and focused on one specific slice of a business problem, and they use AI to solve that for specific customers. In between those two extremes, I think there is a lot of untapped opportunity in the middle. I think developers would be wise to create AI solutions that aren’t too narrow or too broad, and which require very little work on the part of the end customer to deploy the AI solution and see value from it.

Another one is around what’s called “explainable AI.” As AI becomes more and more a part of the way we do business, I think a lot of enterprises are hesitant to get into deep learning because they feel that many of these use cases are black boxes, where they can’t explain why an AI produced certain outcomes. For instance, consider lending decisions – if an AI turns someone down for a loan, that person will want an explanation as to why they were denied. If we work more around new technologies where we can actually explain and provide transparency, I think there would be even more growth in this field.

The third area of focus should address the need to be able to work with less data as well as unlabeled data. In a lot of use cases, you just don’t have enough data. Labeled data is especially hard to come by. It’s costly, it sometimes requires an army of people to actually label the data, and it can be very difficult to do because you might need expertise to label the data accurately.


Learn more

In our new series, AI Experts, we interview thought leaders from a variety of disciplines — including technology executives, academics, robotics experts and policymakers — on what we might expect as the days race forward towards our AI tomorrow.

  • Facebook
  • Twitter
  • Linkedin
  • Google+
  • Email

More answers