By Michael Woodall, Executive Director - Software Engineering-Candescent, Canada
Parachuting across five companies in banking, healthcare, software, legal services, and retail to lead (save) their struggling AI transformation efforts, these events have built on recognizable, real-world patterns and experiences. The failures were consistent, predictable, and avoidable. This article reflects on the lessons I learned while working closely with talented leaders who were enthusiastic about AI but unprepared for the structural, cultural, and strategic requirements needed to make it succeed. The challenges listed are far from an exhaustive list, but I included the most significant ones that are difficult to rectify if not caught and corrected early on.
Banking: Data Fragmentation and Regulatory Paralysis
I was working with Daniel, the VP of Customer Insights at the mid-sized bank. Daniel was sharp, energetic, and certain that AI would help the bank leapfrog competitors by predicting customer churn and tailoring product recommendations.
However, as I started to meet the teams, the problems appeared almost immediately. The bank’s customer data resided in seven different legacy systems. Some dated back over twenty years. Names didn’t match. Addresses were outdated or incomplete. Records were duplicated. Before we could even begin analysis and modeling, work began on data pipelines, unification, and data hygiene. We even needed to invest in synthetic data augmentation where data was missing. These efforts alone were a year of work.
The regulatory team, led by Priya, was highly competent but deeply risk averse. Instead of exploring how AI could be legally and ethically deployed, the bank’s instinct was to reject anything that wasn’t already documented in policy. Any attempt to use customer- level data triggered long review cycles. Due to regulatory constraints, trying to leverage any ML models with “black box” generated outputs from AI was a no-go and as a result, restricted our machine language model options to Supervised only.
The leadership team also lacked a solid and understood business case. They wanted AI because every bank wanted AI. But they couldn’t articulate what success looked like. That uncertainty stalled funding, delayed approvals, and created hesitation.
Healthcare: Change Management Failure and Lack of Business Input
My partner in this project was Megan, a Director of Innovation tasked with making the organization more tech-forward. She was excited and believed AI predictive analytics could identify at-risk patients before emergencies occurred. The rollout was placed under the CIO; not the ideal owner imho. I quickly realized that AI was viewed by the executives almost exclusively as a “tech/system” project rather than a collaborative business and tech endeavor. Without significant business buy-in and accountability, this was doomed to fail.
Clinical environments are complex. Any new tool must integrate seamlessly into their workflows. Our predictive model achieved strong accuracy in testing, yet clinicians were reluctant to use it. Bottom line: they did not trust AI telling them how to prioritize patient care. This was an often-seen lack of change management; you can’t roll out AI if employees don’t understand the process and are not motivated; even worse they are resistant.
Executive buy-in was weak as well. Without a strong sponsor, the project was constantly sidelined by operational emergencies. Healthcare has huge AI potential, but without leadership pushing cultural and process changes, technology never reaches the bedside.
Software Company: Wrong Staffing Approach and Misaligned Expectations
My next assignment held a lot of promise because I was in my wheelhouse. At the software company, I worked closely with Jason, the CTO. He was confident that his engineering team could handle any AI task. However, AI requires specialized skills— data scientists, ML engineers, MLOps practitioners, etc. He had none of these.
The plan was to hire a senior data scientist, but the executive team debated endlessly over the budget. Jason’s attitude was “We will be fine; my people are all technical and can do any role needed”. I pushed to bring it contractors or consultants, but he wasn’t having it. As a result, with the lack of experience and limited understanding of data science fundamentals, many mistakes were made with Data and poor ML models were chosen.
The build vs. buy argument created chaos. Product teams wanted to purchase a ready- made AI platform to accelerate delivery. Engineering insisted they could build everything internally. Leadership wanted a “unified AI strategy,” but every department had its own agenda. Misalignment became the greatest barrier.
The result was a clunky, late, and over budget AI implementation that wasn’t producing the ROI it should have.
Legal Services: Cultural Resistance and Fear of Replacing Expertise
In the legal services firm, I collaborated with Amelia, a Senior Partner overseeing strategic operations. She wanted AI-driven document review to reduce turnaround time for clients. The firm’s lawyers, however, were far from enthusiastic.
They feared that AI might replace meaningful portions of their billable work. Even though the AI system performed well in early testing, attorneys questioned every output. They asked for explanations that neither the model nor the vendor platform could provide in a legally defensible way.
The firm’s leadership supported the project but did not reinforce it strongly. Without active championing, lawyers simply continued working the way they always had. Adoption remained low.
This situation echoes the broader industry hesitation reported in analyses of AI in law firms—many see the value, but cultural and financial concerns overshadow operational improvements. These firms tend to be task-based; they are not used to continual operational monitoring and metrics – they should be defining KPI’s and OKRs and tracking.
Retail/ Baby Supplies: Misaligned Strategy and Unknown ROI
My contact with the baby supplies company was Josh, the Director of eCommerce. He was enthusiastic about AI but had no clear goals. He wanted AI because competitors were using it—not because the business had a defined problem to solve.
The company proposed generative AI for content creation, product recommendations, and supply chain optimization. All great ideas, but none had clear ROI expectations. Without a solid business case, every project began unfocused. As part of my investigation, I talked to random employees across the company, and most did not know why they were going through an AI transformation, and how it would impact their jobs.
Funding dried up halfway through. Leaders approved a budget for experimentation but not the additional funding needed to deploy and maintain AI systems in production. This is a common pattern in retail—pilots get attention, but long-term operations get ignored. They had failed to set up continuous financial monitoring of the AI initiative. They also underestimated the significant cost required in hardware (cloud usage/storage) and tooling (AI tools and platforms).
The organizational structure made it worse. AI didn’t belong to any team. Every department participated, but none owned outcomes. There seemed to be a lack of accountability, and a “hot potato” mentality across the C-Suite when it came to AI implementation. As a result, the initiative splintered. Funding evaporated because nothing tangible was delivered.
Cross-Industry Patterns: What These Failures Have in Common
Across all five companies, similar themes appeared. Data fragmentation slowed progress everywhere. Leadership alignment mattered far more than any model or algorithm. Companies underestimated staffing needs, operational readiness, and long- term funding. Hiring for AI and building in-house models are harder than you think; consider growing from within augmented by outsourcing/consulting to start. Also, try and buy (or choose open source) AI tooling where possible unless you really have the capabilities or derive a competitive advantage from it.
Organizations also chased hype rather than solving real business problems. Without a clear value proposition, AI becomes expensive experimentation. Departments worked in silos, and governance was inconsistent. The governance needs to include establishing KPIs and monitoring them continuously. AI Education at all levels is critical for both buy- in and upskilling and was sorely lacking in most companies.
Data is more critical than anyone originally thinks it is, it must be a starting point and of the utmost quality before you can do anything. Given the privacy concerns encountered, a Data governance leader/team (ideally with external advisors) including AI Ethics usage policies need to be established. Governance is an ongoing process not a one-time audit. Data always changes.
The truth is simple: AI transformation succeeds only when the organization treats it as a long-term capability. It requires investment, ownership, and cultural adaptation—not just technology. This is a permanent mind shift and not a one-and-done project (like implementing an internal system). The organization needs to be set up for success including clear ownership at the C level.
Conclusion: Avoiding Failure Patterns, Steps to Success
These fictional stories, inspired by real-world cases, highlight the real reasons AI initiatives fail. They reveal that success requires alignment, strong leadership, and thoughtful planning. AI is not a switch to flip. It is a journey that reshapes how a business operates. This helps explain why, according to research, “only 21 percent of the organizations have embedded AI in various parts of their business; and only 3 percent of large organizations have integrated AI in their enterprise workflows”.
Companies that succeed invest early in data quality, governance, organizational alignment, and staffing. They focus on business value, not hype. They champion adoption across all levels.
Always start with a small AI pilot to uncover issues and demonstrate ROI before you run into an enterprise-wide problem. AI rewards those who build the right architecture—technical, cultural, and strategic. Those who ignore the foundations repeat the same costly mistakes.
