CIOs - how are you balancing the pressure to adopt AI with the reality that your current data foundation and compute power can’t fully support it?
Sort by:
We are initiating a 'Data Platform 2.0' initiative to get our data where we will need it to be in support of pending/future AI initiatives. The current 1.0 platform was designed around prior operational, reporting and analytics desires some 20 years ago. We have tried to leverage it with minimal success.
One of the key success factors is to not just lift and shift, but to understand what we need to do to cleanse and normalize our data. The key is that as we move and clean, we patch the holes in the processes that allow the bad data to get in. If we do not fix our processes, we know that after a short period of time, our AI ability and more it's accuracy, will degrade.
At this time, we have not really addressed computer power. But acknowledge that we will.
John, you are absolutely right. The BI data is not very suitable as AI data (input). In most cases BI data set does not have the necessary context and variability required to train algorithms. Other key aspects you want to consider - Cleary defined process for output validation and output adjudication to ensure that you can avoid data drifting and produce higher quality results.
You can always find pockets of area where you have good and usable data for AI. Work with business to define a narrow scope of AI solution that can be built on those data domains or work with the teams on creating clean and quality data pockets. You do not need to go through a big data project and wait for it before you can do any AI projects. Once you have a solution that works on a particular area and proves the value of the AI, you can use that success and value prop to build out or broaden your data and AI initiative. While you are doing this, put in place the conceptual roadmap for the Data Strategy and iteratively build it out in conjunction with your roadmap for AI solutions.
Exactly how we have approached this situation, and has been working well for us
I recommend attacking AI after executing digital transformation in this order:
1. Agile
2. Public Cloud
3. DevOps
4. Data Strategy
5. SRE
6. AI
To get to AI sooner, you need not transform your entire organization, just the area that support your AI use case. The smaller the AI use case, the less digital transformation prep is required.
The absolute fastest thing to do is skip most of the aforementioned steps and just do data strategy followed by AI. Doing data first is a non-negotiable.
Balancing the pressure to adopt AI with the reality of limited data infrastructure and compute power is a juggling act many CIOs are navigating right now. Here's how we're tackling it:
Start Small, Aim Big
We don’t need to leap into the deep end with AI. Instead, we focus on smaller, achievable use cases that deliver value without overloading our systems. Things like automating routine tasks or enhancing customer insights are great starting points.
Fix the Foundations
AI can’t thrive on messy, siloed data. A big part of our focus has been cleaning, integrating, and organizing our data to create a solid base. Without that, even the most powerful AI can’t deliver results.
Leverage the Cloud
Rather than overhauling our on-premise infrastructure overnight, we’re leaning on cloud solutions. Platforms like Azure or AWS offer scalable AI capabilities that let us experiment and grow without committing to huge capital expenses upfront.
Bring Everyone Along
It’s not just about the tech—it’s about the people. We’re training our teams, educating leadership, and ensuring the whole organization understands that AI isn’t magic. It’s a journey that starts with improving what we already have.
Focus on Real Impact
We’re staying grounded in what really matters. Rather than chasing every shiny AI trend, we’re asking: What will make a real difference to our business? That focus keeps us from wasting resources on initiatives we’re not ready to support.
Communicate and Collaborate
Finally, it’s about transparency. We’re honest with our teams and stakeholders about where we are today and what it will take to get where we want to be. And we’re not afraid to bring in partners to help bridge gaps in expertise or infrastructure.
The key is pacing ourselves. AI adoption is a marathon, not a sprint, and building the right foundation ensures we’re set up for success when we’re ready to go all-in.
I see 3 major challenges to this, most have been discussed in this thread. First and foremost is the data foundation and the need for certified data products that meet the standard for 'usability' for AI to be applied. The second is not as much about computing power as the dependency on 3rd party tools with the most advanced AI capabilities operating in environments that contain highly proprietary content. The hardware is available and can be quickly added, but there are concerns regarding how to protect against the 'machine' creating content and knowledge that can be contained within the four walls of the business. The last one is probably the hardest, which is agreeing to the hardened business processes that you are willing to trust the machine to perform. To some degree this is about change management and conditioning the workforce to 'trust' the machine, to a larger degree it is about the non-digital path many processes follow that the machine is unable to learn. A buyer is frequently contemplating past experiences, informal knowledge channels and thought patterns that exist outside of what is currently in digital form to make buying decisions.