Catching Your AI Breath

AI adoption is accelerating, but are we giving it the time and thought it deserves? Too many business leaders seem to be sprinting to implement AI without fully understanding its long-term implications. While AI holds immense promise, rushing into it without a thoughtful strategy, risks not just setbacks but potentially damaging consequences for the business.

Let’s pause for a moment and consider how to best shape our AI strategies for sustainable success. AI is not a sprint—it’s a marathon, and the right approach starts with catching your breath and planning with care.

AI: A Business Capability First

Gen AI has been with us for two years already, and while AI as a technology has been with us a lot longer than that many organizations are still figuring out how to harness its potential at scale. The sense of urgency I see in many boardrooms often leads to a hasty focus on “doing AI” rather than building it as a business capability. Yes, AI is a technology, but it’s unlike any other. It will evolve, make mistakes, and change how we work. Unlike traditional software, AI doesn’t need a structured flow and won’t stay static—it adapts. This demands a shift in how we think about AI in the context of its interaction with other business capabilities.

A recent conversation made me realize that many believe AI integration should be simple, citing the frequent release of new software. However, unlike conventional software, AI introduces entirely new challenges. For instance, we’ve never before needed Ethicists to monitor how the behavior of AI agents evolves in the quality and type of response they provide to employees or customers over time. When we recognize that AI needs to be treated as a business capability, we start to recognize the need to understand its complexity being interwoven with other capabilities, highlighting the need for thoughtful planning. Here are some themes that I believe are critical for business leaders as they endeavor to scale AI:

Data Quality: The Foundation of AI Success

If your business is still manually cleaning data and teams don’t trust the data, AI will always struggle. In one example, a business was excited to use generative AI to provide frontline teams with real-time market insights. However, upon closer inspection, these insights were often two weeks old due to data cleansing processes. As a result, the business had to embark on a data maturity program to build trust in the data, so that insights could be generated in near real time and allow front line teams to drive market impacts either in how they sell and serve customers. Without high-quality data, even the most sophisticated AI models can’t perform to their true potential.

Compliance: Navigating Regulatory Waters

The introduction of the EU AI Act adds a new layer of complexity to existing compliance frameworks like GDPR. Many leaders assume that their software providers will take responsibility for compliance given that it is their software or service that they are providing, but it’s not that simple. Vendors will often argue that customizations make it impossible to fully oversee compliance for each unique deployment. This means you need to scrutinize your contracts and service agreements to understand where accountability truly lies and where your software providers or SI partners are claiming compliance to these new policies and regulatory needs, just be curious and ask the question of how they do this, to give your team more confidence.

Process Standardization: Avoiding Assumptions

AI doesn’t need structured data, but it does need clarity in the definition of processes and the role it is being asked to play within those processes to function effectively. Yet, many business processes are ad hoc and live only in people’s heads. When AI is introduced without standardized inputs and outputs, it leads to misinterpretations and false outcomes. Standardizing key processes that support AI use cases will minimize these issues and ensure smoother scaling.

Ethical AI: Beyond the Ethics Board

Many businesses have ethics boards, but Ethical AI needs to move beyond that. The dynamic nature of AI means policies alone aren’t enough. That’s why we’re seeing the rise of the "Ethicist" role, responsible for pressure-testing AI agents to ensure they can’t be manipulated into providing inappropriate responses as well as regularly assessing the health of AI agents to ensure they are not deviating from their intended output with regards to the type and quality of responses they are sharing. A recent experience by a business eager to expose customers to an AI enabled chat resulted in an impact to their brand and a mistrust of their Chat features, all because no one internally was able to track how the bot was evolving its responses (whether influenced by a tech savvy customer or not). As AI matures, the risk of ethical breaches grows, and businesses must be vigilant.

Redesigning the Operating Model

Treating AI as a business capability means rethinking how work gets done. Trying to add AI to existing operating models often results in confusion and workflow disruptions. It is important to rethink how human and machine work together, incorporating AI into the operating model itself rather than treating it as an add-on. Imagine merging two organizations, one powered by AI and the other by humans. Simply combining them won’t work. A thoughtful redesign is required for them to function as one and ensure the value of AI to activities is realized. It is interesting to note that in a recent survey of early AI adopters, people were complaining that it was taking more time to work with AI to complete tasks or run through a workflow than what it did without it, this highlights we are not spending enough time to think through the implications of .

IT Platform: Bottlenecks in Legacy Systems

Legacy IT systems, especially highly customized ones, can become bottlenecks when AI is introduced. While people can work around slow systems, AI can’t. For businesses trying to scale AI, legacy system limitations will likely be one of the biggest challenges. This doesn’t mean that every AI use case will be hit by these same challenges, however assessing the potential IT systems impact upfront for a use case will give your team much higher confidence in the path to persistent value.

Change Management: It’s Not Just Another Technology

Given the perception that AI is just another technology there has been a view shared with me by a few business leaders that existing IT Change management frameworks will suffice for rolling out AI use cases. I would challenge this thinking, as IT Change management tends to think only about the initial deployment and ensuring teams know how to use the new capability. The introduction of AI will have a profound impact on an organization as one business learned in getting employees excited about the AI use cases without explaining the broader implications to people’s jobs. This resulted in an internal negative perception of leaderships value of employees and teams started blaming the AI use cases for all issues that were being experienced to anyone what would listen. Change management must focus on rethinking people’s roles and responsibilities, not just introducing a new tool.

Tracking Value and Managing Costs

Early AI adopters often underestimate the costs associated with scaling AI, particularly when it comes to compute and storage. Having a clear baseline and tracking value throughout the AI lifecycle will help businesses to make decisions as to whether the growth value is justifying the increasing costs and ensure the benefits are realized vs an expected plan. Remember, as AI evolves, so do the cost models tied to it whether that’s the load on compute, the volume of data created (it is expected that AI will be one of the largest data ‘generators’ within an organization as use cases scale, more so than your campaign or billing systems for example).

Governance: AI’s Unique Challenges

Governance for AI requires a multi-disciplinary approach, involving procurement, HR, legal, finance, and more. AI governance will differ from traditional governance forums, and ongoing oversight is crucial to ensuring AI remains aligned with business objectives. Whether it is endorsing use cases to move from pilot to production, assessing the health of AI use cases in production or challenging the ongoing value of use cases that may no longer be adding the value or more likely have deviated too far from the original expectations and thus requiring a reset to avoid impacting the business negatively. Many businesses are thinking that in empowering teams, after the initial sign off of an AI use case each team will manage its use cases on its own, over a period of time this may be possible especially when considering to leverage AI to monitor AI ( to be clear different solutions not a poacher turned game keeper scenario) or more specifically AI Agents in what they are delivering as an output and how they are doing it.

?

Final Thoughts

I know this list can seem overwhelming. But awareness is the first step. The key is to prioritize based on the AI use cases that will bring the most value to your business. I encourage you to think of AI not as a sprint to adopt but as a marathon where thoughtful planning leads to long-term success. Consider establishing a cross functional team that ensures AI as a business capability can be integrated across all roles and processes. So, take a moment to catch your AI breath—shaping a steady and progressive path wins this race.

?

As always, this is my perspective, and I’d love to hear yours. What areas do you think are most critical? Are there others I’ve missed? Does this even really matter? Let’s continue the conversation.