putilov_denis - stock.adobe.com
As more organizations deploy AI projects and products, many enterprises are looking beyond surface-level AI. They're looking to take their AI projects to the next step.
One way to do that is by pursuing an AI-first data strategy, according to Forrester Research. This means creating machine learning and data models that are designed with an AI mindset instead of trying to use already created data to fit their AI.
Phases of an AI-first strategy include delivering and deploying data for scale, testing and training models to create trust, and discovering and sourcing data that represents the business model of the enterprise.
In this Q&A, Michele Goetz, a Forrester analyst, explains what an AI-first data mindset means for enterprises and some of the challenges enterprises face in their attempt to build models using an AI-first data strategy.
What does an AI-first data mindset or strategy mean?
Michele Goetz: An AI-first mindset means that not only are you applying sophisticated methods and algorithms or data science to understand what's going on with your business, with your customers, or gaining insights, but that knowledge also is embedded and being put to work into the experiences, the products or the value propositions that you're offering to market.
AI is really the activation of data and insights in the way that you do business. It's not just information or insights on a screen or a dashboard. It's the driving engine behind the way your business works and the way that your digital experiences are generated.
What are the signs of enterprises employing this AI-first data strategy?
Goetz: To be an insight-driven business, are you able to take advantage of your data? Then you're a beginner. Are you able to generate the insights that you need to make decisions? You're an intermediate. Do you have an AI mindset and your business running on AI and running on algorithms? That's an advanced state.
Generally, we're finding that organizations are still a bit in the getting started space with AI because they're having to better develop not only the data science capabilities, but they have to address the data platforms in order to help data scientists utilize the data that they should be using to come up with the insights that they need.
Then certainly, how you deploy those models out into your digital ecosystem is still generally challenging, or at least challenging to do at scale.
How can enterprises surmount challenges preventing them from using an AI-first data strategy?
Goetz: Enterprises spend a lot of time modernizing their data platform by moving to the cloud and generating data lakes and consumption layers.
That usually isn't enough because the data scientists and the business subject matter experts who are developing and designing these models and AI experiences don't know what data they need. They don't always know what they must do to transform or prepare the data. They really need to have a platform that isn't just a place to go for the data, but also includes the capabilities to help them work with the data themselves. They have the expertise and the knowledge about the business and what the model needs to do.
Michele GoetzAnalyst, Forrester
It's easier for them to just be able to work with the data themselves than it is to pick up Slack and get in touch with a data engineer again, to do a whole bunch of further preparation and modeling for the data. So, ensuring that you're addressing the data experience for the data scientists and consumers as much as just modernizing a platform where they can build and train models is going to be really important.
The second thing I look at is how collaborative is the process to create and train or build and train on machine learning or deep learning model?
[If a data scientist is] sitting at their desk creating models, then generally I know that they're very much a beginner. This is in terms of the sophistication of their data science, and also in terms of the value or the literacy that their organization has around what AI can provide for collaborative purposes.
As you see pods being created between data scientists and the subject matter experts where they're continuously working together, they're iterating the training and they're validating in highly collaborative ways. These stakeholders in the business understand what's happening in the model and are producing the right results. Then they move forward and they're being a bit more strategic in the way that they want to go after AI. The last phase [is] looking at this through the lens of not only data science and insights, but also a developer community. Along with that you have data engineers, machine learning engineers and application developers all involved to ensure that the models that are built can be pushed into the business applications and platforms to run the business and provide experiences and value to the market.
How will this trend of using an AI-driven mindset develop?
Goetz: You're seeing the trends not only manifest in the principle of strategic sourcing but there's also now roles in organizations such as a data librarian or a data curator. They could be sitting within the analytics or data science organization, or they're sitting within the data governance organizations in the chief data officer's office, particularly in places like banking and investing.
Certainly, the hedge funds have been doing this for quite some time, but it's also rolling out into general investing and wealth management. You're seeing this within life sciences and pharmaceutical companies. There's a real recognition that your ability to gather as much information and knowledge as possible, is only going to make your AI capabilities better.
Editor's note: This interview has been edited for clarity and conciseness.