Why adopting AI is not enough, it needs resilience too [Q&A]

Last October’s AWS outage underscores the fact that as companies race to deploy AI, outdated infrastructure and inadequate data protection systems create compounding risks. At the same time shadow AI applications expose sensitive data, while legacy systems can’t support the real-time insights AI promises.

We talked to Ric Opal, digital leader at Global BDO, to find out why it’s not enough just to roll out AI, it’s essential to build in resilience too.

BN: How can business leaders effectively balance the pressure for rapid AI adoption with the critical need to meet cybersecurity and data management requirements?

RO: The truth is you cannot deploy AI successfully without clean data and proper security.

Our recent survey of over one thousand global business leaders revealed that 57 percent of organizations are fast-tracking AI adoption, but only 42 percent admit they have the infrastructure to fully leverage AI. This gap represents the biggest challenge. Additionally, 76 percent of survey respondents express concerns about cybersecurity in their businesses. Leaders feel the pressure to move quickly, but they are building on shaky ground. An unstable foundation can lead to critical mistakes such as data breaches that could cause long-term reputational harm.

The balance comes from integrating operations, IT, and risk professionals together efficiently. When risk teams and business leaders understand how security affects outcomes, they can start aligning on strategy. The organizations getting this right are taking this one step further to make sure AI initiatives support overall business goals.

Business leaders need to understand that it is not a race to AI adoption, but true competitive differentiation will come from effective implementation.

BN: Why do legacy systems specifically fail to support the real-time insights AI promises, and what are the top infrastructure upgrades or architectural changes an organization should make to become ‘AI-ready’?

RO: Legacy systems typically fail to support real-time insights because they weren’t built for the speed and scale AI demands, and they typically house siloed or messy data that is unstructured, incomplete, inconsistent, or error prone. As stated above, less than half (42 percent) of business leaders say they have the data infrastructure needed to fully leverage AI technologies. Our research also shows that messy and siloed data is the number one barrier to leveraging technology benefits. Companies need to modernize their technical infrastructure and break down data silos to get the most out of their innovation investments.

Most organizations (91 percent) are now technically ‘in the cloud’, but migration is just the starting point. They need to also implement scalable cloud architecture. Too many businesses have simply lifted legacy systems into a new environment, like moving into a new house before clearing out the clutter. Without clean, structured, and secure data, even the most advanced systems are rendered ineffective or dangerous.

To become AI-ready, organizations need to make three upgrades to make the most of their data. First, clean before you scale. Before investing in AI or analytics, bring together data from across the organization, eliminate duplicates, and resolve inconsistencies. Second, treat data as a core strategic asset. Embed data thinking into every part of the business and foster a culture where data drives decisions. Third, commit to continuous data governance. This isn’t a tick-box exercise — ongoing monitoring is essential to maintain and improve data integrity over time so insights remain accurate and trusted as the business evolves.

BN: Which are the most common failures you see in AI rollouts, and what can be done to identify these before a failure occurs?

RO: The biggest concern we’re seeing in AI rollouts is the disconnect between senior leadership visions and IT teams’ realities. Some organizations jump straight to tools without ever defining what success looks like or figuring out where AI can deliver measurable value. So, AI becomes an expensive distraction instead of strategic systems that move the business forward.

The second thing we’re observing is leaders underestimating the people side of AI implementation. Our research shows that only 49 percent of leaders think change management will be central to their technology strategy. This makes it vital to train the workforce on not just how to use these AI tools, but why it is important to embrace them.

Third is lack of focus — trying to do everything at once instead of targeting high-impact use cases. Start by identifying your biggest inefficiencies and pain points, then deploy AI against those specific problems first.

To address these setbacks early, start with three fundamental questions before any AI investment: What specific business problem are we solving? How will we measure success? And how will we reinvest the productivity gains? If you cannot answer these clearly, you do not have a strategy.

Watch for warning signs of disconnection. Is AI being implemented solely because competitors are doing it? Is there an inability to articulate what success looks like? A lack of clear KPIs tied to AI investments, or projects scattered across departments with no coordination, can cripple an organization.

On the people side, survey your workforce before rollout. Do they understand why you are implementing AI? Do they feel equipped to use it? Do they see how it benefits their work? If employees view AI as a threat rather than a tool, or if adoption rates remain low, you need to invest in communication, training, and change management before deployment.

Finally, benchmark yourself against peers by communicating with them directly. Without data, you are guessing. Without an external perspective, you are missing the context that helps you identify blind spots. It is important to remember that implementing AI does not mean that AI is driving value. AI infused into an organization’s strategy and processes, not simply bolted on as a standalone tool, will separate the winners from the losers.

BN: What specific skills, beyond data science, do employees and leaders need in order to successfully build and maintain an AI-resilient organization?

RO: I think it is important to remember that a new tool might last a few years, but your people’s careers span decades. Their ability to adapt, keep learning, and think critically is what determines whether your digital investments pay off.

As technology gets more powerful, it raises the bar for the unique value people bring to the table. Organizations will need to rethink roles around higher-value work, including creativity, critical and strategic thinking. Employees should demonstrate and emphasize their business acumen, data literacy, creativity, and curiosity while on the job or when searching for a new role. The time saved through automation must be reinvested into meaningful work, not just trimming headcount.

For leaders specifically, AI-resilience requires change management capabilities to guide teams through transformation. Our research shows that only 49 percent of leaders believe effective change management will be central to their technology strategy, yet this is critical for success.

Additionally, leaders need to build a culture where people feel safe experimenting. Innovation happens when people can test ideas and fail fast without facing intense consequences. With that said, you still need governance and knowledge sharing. Focus on fostering a culture that celebrates bold thinking, rapid iteration, and lessons learned as well as outcomes achieved.

BN: How can you measure the long-term value of AI adoption when the initial investment is often tied up in infrastructure which doesn’t always provide immediate ROI?

RO: The key is alignment. When your AI initiatives support clear business objectives — for example, cutting down manual work in finance, speeding up forecasting, or getting better customer insights — then the value is measurable. Focus on real business problems where technology can make a tangible difference.

The challenge with building infrastructure is it typically does not provide immediate ROI, so you need to track progress with stage-specific KPIs. In the early stages, measure data readiness: How quickly can you clean your data sets and update your legacy systems? Are your teams ready to be integrated? Then over time, shift to operational metrics. Are we saving time, money, and/or resources?

Employee adoption is another critical metric. I speak with leaders who have specific AI adoption goals built into their performance reviews ensure accountability. When AI adoption is tied to someone’s compensation and performance metrics, they drive toward it.

The long-term value extends far beyond immediate ROI — it is about building organizational resilience and agility to withstand future disruptions from cyberattacks to major cloud service outages. In fact, resilience ranked as the top priority among survey respondents in Chapter One of BDO’s Techtonic States Report. Strategic infrastructure investment provides exactly this: benefits in continuity, recovery speed, and adaptability that position organizations to thrive amid uncertainty.

Image credit: Tongsupatman/Dreamstime.com