The Cloud Experience Everywhere

3 Steps to Create Your Roadmap to AI


In my previous post, “A 5 minute Into to AI: What it Can (and Can’t) Do”, I looked at some common misperceptions around artificial intelligence, and sketched some of the truly exciting insights that the technology can deliver. It’s time to consider what AI can do for your business. Many organizations are eager to put their data to work, but they’re unsure where to start. I’ll offer some suggestions in this post, based on some of the methodologies that we use in HPE Pointnext’s Artificial Intelligence Transformation Workshop, which we announced in March.

Broadly speaking, you create the roadmap for your initiative in three steps. First, consider your business objectives, with an initial look at relevant data. Next, determine the outcomes you’re seeking and choose the analytics to support them. Then do a more in-depth assessment of the data sets, including ways to combine them for better results.


Let’s look at those steps in a bit more detail:

1. Select and analyze your potential use case.

Though it may be tempting to jump into researching the latest and greatest AI tools, it’s much better to start by looking at your company’s current opportunities and challenges to see where AI can make the biggest contribution. What are the overarching business strategies that you want to tie into? What initiatives are already in progress that AI might be able to support? Or what completely new initiative can you envision that would deliver significant value? The goal at this point is to quickly identify a promising use case, which in turn will help you identify the data sources you’ll need.

Spend some time considering the data sets that are available. Most organizations choose to focus initially on the data they’re most familiar with. In manufacturing, for example, that might be machine logs, audit logs, temperature records. For a financial services firm, it may be customer records and transaction data (treated with great caution, as I’ll explain in a moment). This approach is fine, but it’s a good idea to think more broadly about the data to support your use case. You may end up discovering valuable internal information that you’ve previously overlooked, or uncovering ways to enrich your data further with material from external sources.

2. Identify your desired outcomes and the right level of analytics.

Once you have your use case clearly defined, you’re ready to think about what you want the AI solution to do for you, and how it will deliver the outcomes you want. This will help you select the right level of analytics. Do you want reactive, proactive, or real-time outputs? Do you want to understand why certain critical events occur in a piece of equipment? Then you’ll want to investigate diagnostic analytics solutions. At a higher level of analysis, do you want to know what actions you should take to head off possible equipment failure? Prescriptive analytics solutions can help (HPE Digital Prescriptive Maintenance Services is a great example, described in Hande Sahin Bahceci’s post Making Artificial Intelligence Enterprise-Ready: HPE Unveils New AI Solutions).

One note of caution here. If the data for your AI use case hasn’t been leveraged before, it’s worth asking why not. The reason may simply be a lack of interest, insight or expertise within the business, but it may also be a regulatory compliance issue. Any time you broaden the use of your data, you need to pay careful attention to usage rights. In the European Union, the General Data Protection Regulation (GDPR) has brought increased attention to this issue (and let’s remember, GDPR applies to all businesses, including multinationals, that hold data on EU citizens, regardless of domicile – see Lois Boliek’s post The 5 P’s of Data Protection.) There may also be service agreements, non-disclosure agreements, or other legal terms and conditions that constrain your use of data in new ways. Proceed with caution.

AI transformation workshop2.jpg

 3. Assess and analyze your required data sets.

Now you’re ready to take a closer look at the internal and/or external data sources you identified in step one. Describe their characteristics carefully. Is the volume of data adequate for your purposes? How is it labeled? What’s the refresh frequency? Will the data require cleaning?

An important part of this process is to consider aggregating data sets. Companies tend to think in terms of a single data consumption model, but often the most successful approach is to layer multiple data sources into the pipeline. It could be multiple AI algorithms, but it could also be adding traditional analytics or even just a traditional data filter. Prescriptive maintenance is a good example. In a wind turbine, internal sensors and temperature gauges can tell you whether excess vibration is building up within the unit, indicating the need for intervention. If you then combine that with video inspection of the external machinery, you might be able to detect a worn edge on one of the blades, which could be causing the vibration. Combining data streams in this way can often give you far more precise and useful readouts.

Your partner for AI transformation

At this point you’ve done all the spadework, and you’re ready to start assembling the development plan for your proof-of-value or pilot project, and eventually full deployment of the solution. It’s time to start building your timeline, plotting the various activities across it, and getting the SOWs lined up.

HPE can help you at every step of your journey to artificial intelligence. As leaders in the data science space, we can supply the dedicated expertise you need to get the best results from your AI investments. We can provide pre-trained models that can get you 80 percent of the way there, and help you complete the training on your own site, with your own data. The AI wave is one of the most exciting areas of digital transformation today, and we would love to help you harness it.

Learn more about HPE Pointnext here.

Related articles:

About the Author


Senior Technology leader and strategist. Delivering opportunities, value, and outcomes through applied technology and artificial intelligence.