AWS Head of Innovation for SMBs, Ben Schreiner reminds business leaders to focus on data and problem solving when making decisions around generative AI.
Generative artificial intelligence is a hot topic, but many of the things it can do seem very similar to yesterday’s predictive algorithms or machine learning. We interviewed Ben Schreiner, head of innovation for small and medium businesses at Amazon Web Services, who says today’s generative AI isn’t magic; SMB purchasers should look at it with the full context of AI’s weaknesses and its impact on people. However, generative AI does offer use cases that weren’t previously possible.
This interview has been edited for length and clarity.
Jump to:
Megan Crouse: How is generative AI different from the type of machine learning that we had five years ago or longer than that? How is it the same?
Ben Schreiner: Generative AI is not magic — it’s math. What we’re seeing in the market is generative AI hype has captured people’s imagination and is fostering a conversation around innovating that we weren’t having before.
SEE: Generative AI has reached the peak of Gartner’s Hype Cycle, where expectations are inflated. (TechRepublic)
When the economic downturn happened, most people were focused on saving money and costs. This generative AI news cycle has had small and medium business leaders talking more about innovation, maybe in the same conversation as cost savings. It has allowed us to have that conversation (about innovation).
Most of the use cases end up being things that have existed for quite some time. What I’m most excited about is we’re having that innovation conversation whether you’re using the latest large language model to do actual generative stuff or you’re leveraging AI that has existed for five or 10 years.
It really doesn’t matter. We just want our customers to leverage it, because that’s where innovation happens for their business.
Megan Crouse: What questions should business leaders ask when deciding to use generative AI or a generative AI-enhanced service?
Ben Schreiner: The number one question I have to ask is where is the data? What data was used to train this model? Everybody’s learning very quickly, and most of the customers we talk to understand that the model is only as good as the data that it has. Understanding that is really important. Understand who owns that data, where it came from and how much of your own data you need to put into the model or augment the model (with) in order to get out real answers that are valuable. That balancing act is a very important one for business executives to understand. Where is the model?
We want to bring the model to your data, not the other way around. So our approach to AI and generative AI is to allow our customers to have their own instances of models that they can modify and enhance with their own data, but all protected within their own environment and their own security controls where no one else has access to that information.
Priority number two is making sure you’re partnered with an organization or a partner that’s going to be with you for the long haul and has the expertise. We have a bunch of third-party partners that make either new models available or that have experts that can help some of these companies that don’t have data scientists on staff.
Then just learn. Learn as much as you can as fast as you can, because this (generative AI) is changing almost hourly.
Megan Crouse: Two concerns I often see people bring up with generative AI are copyright, specifically generative AI being trained on copyrighted works, and hallucinations. How do you address those problems?
Ben Schreiner: I think everyone needs to go in with eyes wide open, right? The machine is only as good as the data. You have to understand what data is in there. And AWS is trying very hard in our own models.
We make sure that we know where that data is and that we’re not creating a liability or a potential risk for those customers. We have our own Titan models. Then you have all of the open source models that are coming out, and we intend to have the best models available. We don’t believe it will be a one-size fits all, or that one model will rule them all.
But I do think executives need to understand the source of the model’s data itself.
Regulations are going to trail (behind businesses). You’re seeing lawsuits now being filed trying to protect some of that (copyrighted) information.
Megan Crouse: In what ways do business leaders in small and medium businesses need to invest in people before they invest in AI? And what questions should they be asking themselves about how adopting generative AI might change the way they invest not only in tech but also in supporting their own people?
Ben Schreiner: I think all small and medium businesses should be people-first. (People are) your biggest assets, and the tools and technology really are only going to ever be as good as the people who leverage them. In regards to investing in your people and investing in their training, earlier this month, we (AWS) released seven new AI-oriented training classes. We intend to help people learn as fast as possible and make it as easy as possible for folks to leverage this technology.
SEE: Hiring kit: Prompt engineer (TechRepublic Premium)
Not every business is going to be able to afford or attract a data scientist. How do we make it so you can still benefit from some of these technologies and not be kept out of the market, kept out of this revolution, because you can’t get a data scientist on staff?
Megan Crouse: Is there anything else you would like to add?
Ben Schreiner: I want to highlight the concept of generative business intelligence. We are helping a lot of small and medium businesses aggregate their data. That’s kind of priority number one.
You aggregate your data, ideally in AWS, and layer on business intelligence on top of that. So think about reporting, but add the generative component to reporting and being able to use natural language to, for example, tell me the product I sold the most of that has the highest gross margin for the summer months and compare that year over year.
I’d like to be able to verbally ask that of the tool and have it spit out a chart for the data that I need. That is very, very compelling because now I don’t need a database administrator that’s doing SQL queries and creating advanced pie charts for me. I can have the tool, and can have the intelligence embedded inside of it, and be able to ask it things.
The next level of generative BI is to actually write the story of the data that it’s seeing. It comes up with paragraphs for a summary or an executive summary of the data. And I’m not spending time generating that — I just edit it to meet my needs. So I’m excited about that because all small and medium businesses have data, and most of them are not maximizing the value of that data.
