During a recent conversation on The Executive Outlook, Julia Bardmesser shared a message that challenges how many organizations approach artificial intelligence.
There is no such thing as a successful standalone AI strategy.
It is not because AI and data do not require planning, investment, or leadership. It is because a strategy disconnected from what the business is trying to achieve is not really a strategy. It is a technology initiative with no clear definition of success and no reliable way to prove that it created value.
Julia has spent more than 25 years helping organizations use data, technology, and AI to support meaningful business outcomes. Her career has taken her through some of the world’s largest financial institutions, where she has seen technologies evolve, platforms change, and new waves of innovation reshape enterprise priorities.
Yet one lesson has remained consistent throughout that evolution.
Technology creates value only when it helps the business achieve something that matters.
Julia did not begin her career with a long-term plan to become a data and AI leader.
She started as a programmer at Bloomberg Financial. When two people joined her team, one was assigned user interface work and the other was assigned data work. She received the data assignment.
Years later, while working with technology supporting trading operations, she decided she wanted to move into management. Once again, two opportunities appeared. She wanted the role of leading a team that built applications for traders. Instead, she was offered a position managing developers who created databases and data assets used by multiple teams.
It was not the role she originally wanted, but it became a defining moment in her career.
She discovered that she was good at data work and, more importantly, that she enjoyed it.
What kept her interested was never the idea of building the most sophisticated database or the most impressive data warehouse. The real attraction was understanding how data and technology could enable the business.
The field also continued to change.
When she began, relational databases were still creating new possibilities for organizations. Data volumes, system performance, and technical limitations shaped what teams could realistically achieve. Business intelligence tools were only beginning to emerge, and in one case, Julia’s team had to build its own BI system because a suitable product was not available.
Today, organizations can process far more data, apply machine learning to complex decisions, and use generative AI to interact with information in natural language.
The capabilities have changed enormously. The business question has not.
What are we trying to achieve?
She frequently encounters organizations that begin their strategy of conversations with technology.
They say they need an AI strategy. They need a data lake. They need a new platform. They need to identify use cases for generative AI.
She begins somewhere else.
What is the business strategy? What is the organization trying to achieve?
The technology of conversation comes after those questions, not before them.
An organization may want to increase customer retention, acquire more customers, improve margins, introduce new products, strengthen risk management, or create more efficient operations. These are business outcomes.
Data and AI are capabilities that may help deliver those outcomes.
When organizations reverse that sequence, the technology itself becomes the objective. Teams invest in infrastructure, tools, models, and reports without first agreeing on the result those investments are expected to produce.
This problem is not unique to the current AI era.
The language has changed repeatedly throughout Julia’s career. At one point, organizations believed they needed a data warehouse. Later, they believed they needed a data lake. Today, many believe they need an AI strategy.
The underlying mistake is the same.
A data warehouse is not a business strategy. A data lake is not a business strategy. Chatbots are not a business strategy. An AI model is not a business strategy.
An organization may need any of those capabilities, but the decision should begin with the outcome they are expected to support.
When the connection is clear, the return on investment becomes easier to explain. The organization is not investing in infrastructure simply to modernize its technology. It is investing in the ability to improve sales, reduce customer churn, manage risk, or increase operational efficiency.
The business outcome gives the technology its purpose.
Her approach begins by temporarily removing data and AI from the conversation.
She first works with business leaders to understand what the company wants to achieve this year, next year, and over the longer term.
Who are the company’s customers? What challenges does it face in acquiring and retaining them? What are its plans for products and services? Where are the margins under pressure? What risks need to be managed? What operational or supply chain challenges are limiting growth?
These questions help the organization articulate its business strategy clearly.
Once the priorities are understood, the next step is to identify the data and AI capabilities required to support them.
The process involves assessing what the organization already has, where the important gaps exist, and which investments should be made first.
Sequencing matters because organizations can pursue countless data and AI initiatives. They cannot pursue all of them at once.
The most valuable starting point is often a smaller gap that is closely connected to an important business outcome. Delivering an early, measurable result helps build confidence and creates a stronger case for the larger foundational investments that may follow.
This is what separates a business-aligned data strategy from a technology program supported by an aspirational document.
The organization is not investing in AI simply because AI is important. It is investing in a specific capability because that capability helps achieve an agreed business result.
For years, data science and machine learning teams have produced valuable predictions, recommendations, and analytical insights.
The problem was often not the quality of the insight.
The problem was that the person who needed it never used it.
She describes this as the last-mile problem: the gap between generating an insight and delivering it to the right person in a format that fits naturally into the way they work.
She shared an example involving a machine learning model designed to support sales teams.
The model used internal and external information to create highly targeted lists of prospects. It could help salespeople identify who they should contact and where new commercial opportunities might exist.
Technically, the model worked.
Commercially, it did not create the expected impact because the recommendations were placed in a separate area of the customer relationship management system. Looking at that information was not part of the sales team’s established daily process.
The insight existed, but it did not influence behavior.
Generative AI creates a new way to address this problem because it can present information in natural language and through interfaces that are more convenient for the person consuming it.
A salesperson preparing for a meeting may not want to open and interpret another report. They may benefit more from a short voice briefing that explains the client’s product footprint, recent interactions, potential cross-sell opportunities, and relevant notes from previous conversations.
The underlying intelligence may still come from traditional data systems and machine learning models. Generative AI makes it easier to deliver that intelligence in a format that fits the user’s immediate context.
This is where she sees one of Gen AI’s most valuable enterprise roles.
It can help bridge the gap between analytical capability and practical adoption.
It does not eliminate the need for change management, but it can make adoption easier by reducing the amount of change required from the user.
The accessibility of generative AI also introduces an important risk.
Unlike traditional machine learning outputs, generative AI can communicate in a highly confident and human-like way. That makes it easier to use, but it can also make incorrect answers more difficult to recognize.
She experienced this while developing material for a course about retrieval-augmented generation, commonly known as RAG.
She used a generative AI tool while creating the presentation. The system produced material that appeared plausible but did not accurately reflect what she was trying to explain about RAG.
She recognized the problem because she understood the subject well enough to challenge the output.
A user without the same knowledge might have accepted it.
This is one of the central risks of generative AI. It can produce answers that sound reasonable and authoritative while still being incorrect.
That is why AI does not remove the need for data management, validation, governance, or human judgment. In many cases, it makes those disciplines even more important.
Organizations looking for shortcuts around difficult data management work may discover that AI amplifies the weaknesses already present in their information environment.
Reliable AI still depends on reliable data, clear controls, appropriate oversight, and people who know when to question the answer.
She is often asked to help organizations fix what leaders believe is a technology problem.
In many cases, she finds that technology is not the real issue.
Projects fail because the organization focused on building the capability without thinking deeply enough about the behavior the capability was supposed to change.
They fail because users are introduced to the solution after it has already been designed. They fail because a team measures deployment rather than adoption. They fail because the new tool sits outside the user’s established workflow. They fail because employees are told to work differently without being involved in the process or shown how the change will help them.
The sales prospecting example illustrates this clearly.
The model produced the intended recommendations. The failure occurred because the recommendations were not integrated into the way salespeople worked.
The project was technically successful but commercially ineffective.
This is why change management cannot be treated as a communication exercise that begins shortly before launch. It must be treated as a design requirement from the beginning.
Teams need to speak with users before building the solution. They need to understand how decisions are currently made, where friction exists, and how the new capability will fit into established processes.
Technology should support the user’s work rather than require the user to reorganize everything around the technology.
The Use Cases That Deliver Business Value
She identifies several areas where well-managed data and AI capabilities can produce measurable business impact.
One is sales intelligence.
Machine learning can help organizations identify promising prospects, recommend cross-sell opportunities, and provide sales teams with useful client context. Generative AI can then make that intelligence easier to consume by presenting it through natural language, email, voice, or other familiar formats.
Another major area is customer retention.
Organizations can use data to understand customer lifetime value, identify customers at risk of leaving, predict churn, and design more targeted retention efforts.
These capabilities allow the business to focus its attention where action is most likely to protect or increase long-term value.
Data monetization can also create opportunities, particularly for established organizations whose traditional business lines are experiencing margin pressure.
The data generated through existing operations may have value beyond its original purpose. However, she stresses that organizations must approach data monetization carefully.
Using data for a purpose other than the one for which it was originally collected raises important legal, ethical, and governance questions. Organizations must understand their data rights and ensure that any new use is appropriate, permitted, and responsibly managed.
Risk management is another significant area of impact.
In financial services, this can involve financial, credit, market, regulatory, or operational risk. Across other industries, similar principles apply to supply chains, operations, compliance, and customer concentration.
The technology differs by use case, but the strategic principle remains the same.
Start with the business decision or outcome. Then determine which data and AI capabilities can improve it.
She also makes an important distinction about the role of external advisers.
A consultant can provide a framework, facilitate a structured conversation, and help leaders clarify what they are trying to achieve.
However, the strategy must belong to the organization.
An external partner cannot simply enter a company and declare what its business strategy should be. The leadership team understands the company’s customers, products, constraints, ambitions, and operating environment in a way that an outsider cannot fully replicate.
The adviser’s role is to help the organization surface what it already knows, challenge assumptions, identify gaps, and translate strategic priorities into the capabilities required to support them.
This creates ownership inside the business.
It also produces a strategy that senior leaders can explain in clear language because it is connected to goals they already understand and are responsible for achieving.
Across Julia Bardmesser’s experience, one message remains consistent.
Data and AI are not the destinations.
They are capabilities that help a business reach its destination.
Organizations create problems when they begin with a platform, a model, or a trend and then search for a reason to justify it. They create value when they begin with the result they need and work backward to the data, technology, processes, and behavioral changes required to achieve it.
The strongest AI strategy is therefore not a document built separately from the business strategy.
It is a clear connection between what the company wants to accomplish and the capabilities it needs to accomplish.
The organization is not building a data warehouse. It is improving its ability to make decisions.
It is not deploying generative AI. It is helping employees act on information more effectively.
It is not investing in analytics for the sake of modernization. It is increasing retention, supporting growth, managing risk, improving margins, or creating better customer outcomes.
When leaders make that connection clear, the purpose of the investment becomes easier to understand. Success becomes easier to measure. The conversation moves away from technology for its own sake and toward the results the organization was created to deliver.
The AI and data strategy provides capability.
The business outcome was always the point.
Has a leadership perspective shaping the future of data, AI, or enterprise transformation?
Get featured on The Executive Outlook and share your experience with a global audience of executives, technology leaders, and decision-makers.
Listen in for business strategies and interviews with business experts to help you build a business and a life that you love.
About Us
