During a recent conversation on The Executive Outlook, Buck Woody shared an idea that should reshape how every data leader approaches strategy.
The biggest mistake organisations make is not necessarily choosing the wrong technology. It is beginning the technology conversation before deciding what the business is trying to change.
He has lived through almost every major chapter of modern computing. In the 1970s, he built his first computer by following instructions in a magazine. The machine had one kilobyte of RAM, and he used a cassette tape deck for storage.
He later served in the United States Air Force, worked on NASA’s Shuttle Processing and Data Management System, joined Microsoft’s SQL Server team, and contributed to the data technologies that eventually became part of Azure.
Today, he advises organizations on data strategy, helping leaders determine not only what they should do with their data, but why they should do it in the first place.
That distinction between what and why sits at the centre of everything Buck has learned across decades in technology.
It is also the distinction many data strategies overlook.
Organizations invest in platforms, build infrastructure, introduce AI tools and create reporting systems before agreeing on what those investments are expected to change about the way the business operates.
The result is often technically impressive infrastructure that produces little meaningful action.
The most expensive mistake in data strategy is starting with how before establishing why.
Before a platform is selected, an architecture is designed, or an AI project is approved, he believes leaders must answer one clear question:
What are you going to do differently when you receive the answer?
This is not the same as asking what the organization wants to know.
A company may want better visibility into its inventory, website performance, customer behaviour, or operational costs. But visibility alone does not create business value.
The important question is what decision or behavior will change once visibility becomes available.
Will the organisation adjust inventory levels?
Will it change marketing investment?
Will it contact a customer before they leave?
Will it redirect resources toward a more profitable product line?
Will it stop a process that is creating unnecessary cost?
When leaders cannot explain what action will follow from an insight, the question is not yet ready to become a data project.
A strategy that only promises better visibility is often closer to a reporting initiative than a true business strategy. The purpose of data is not simply to help an organization see more. It is to help the organization decide and act more effectively.
He explains this through the analogy of tools.
If someone hands you a shovel and tells you to use it, the natural response is to ask what for. The shovel might be the correct tool for digging a small hole, but it would be useless if the actual objective were to paint a wall.
In the same way, companies frequently begin with a preferred cloud platform, analytics product, AI model, or licensing structure and then search for a problem that justifies using it.
That reverses the proper sequence.
First define the outcome. Then determine what information is required. Only after that should the organization decide which tools, platforms and technical approaches are appropriate.
When the why is clear, the how becomes much easier to determine.
Technology has evolved dramatically, but he argues that the fundamental requirements of reliable data have remained remarkably consistent.
Cleanliness, accuracy, availability and discoverability were essential when organizations worked with early computer systems and flat files. They remain just as important in the age of cloud infrastructure, machine learning and generative AI.
What has changed is the scale of the consequence when those foundations are ignored.
In the early years of computing, organizations were both computation poor and storage poor. Processing power was limited and storing large amounts of information was expensive.
Over time, processing became faster, and storage became cheaper. Organizations moved from being data poor to data rich and from computation poor to computation rich.
This created extraordinary opportunities, but it also amplified the cost of weak data practices.
When inaccurate or incomplete data was processed through a limited system, the potential damage was relatively contained. Today, the same poor-quality data can move across cloud platforms, dashboards, automated workflows, predictive models and AI systems within seconds.
The error does not remain in one report. It can spread across an entire organisation.
He teaches at the University of Washington, where he uses a simple principle to explain the role of data in technology:
All computing is just rearranging data.
Every major technological advancement, from flat files and relational databases to cloud platforms and AI models, represents a more powerful way of storing, processing, connecting, or rearranging information.
But greater processing power does not improve the underlying information automatically.
If the data being rearranged is inaccurate, incomplete, inaccessible, or poorly governed, the technology will simply rearrange those weaknesses faster and at a greater scale.
For that reason, data quality should not be viewed as a cleanup activity that happens before the more important work begins.
It is the important work.
A strong data strategy treats trustworthy data as an ongoing organizational capability. It establishes ownership, definitions, access, controls, and accountability so that business leaders can rely on the information they use to make decisions.
He uses another analogy to explain why AI creates both enormous opportunity and significant risk.
Give a child a small shovel, and the child can move only a small amount of dirt. Give an adult a larger shovel and the adult can dig a meaningful hole over time. But give a child a mechanical digger, and the situation changes completely.
The machine gives the child extraordinary power, but the child may not know to check for water pipes, electrical lines, internet cables, or structural risks before beginning to dig.
AI is the mechanical digger.
Its capability is extraordinary. But the risk increases when the person using it does not understand what the system is doing beneath a simple and accessible interface.
Traditional computing has trained organizations to expect exactness.
A database query asking for records that match a defined condition is expected to return those records consistently. The same input should produce the same output.
Generative AI does not operate in the same way.
Large language models are built on statistical and probabilistic systems. They generate responses based on patterns and likelihoods rather than retrieving one guaranteed answer from a fixed table.
That difference is critical.
A confident but incorrect response does not always mean that the model has stopped functioning. It may be the result of a probabilistic system generating the most likely answer from incomplete, inaccurate, outdated, or poorly matched information.
Leaders therefore cannot apply the same assumptions of certainty to AI that they apply to deterministic software.
AI outputs must be evaluated according to the importance of the decision, the quality of the supporting data and the potential impact of being wrong.
The higher the business risk, the stronger the validation, governance and human oversight must be.
A general-purpose language model may understand language, recognize broad patterns, and reason across common knowledge. But it does not automatically understand the internal realities of a specific organization.
It does not know the company’s customers, operating procedures, product definitions, inventory constraints, internal policies, or strategic priorities unless that context is provided.
Organisations are attempting to close this gap through approaches such as retrieval-augmented generation, fine-tuning and purpose-built model training. Each approach connects a model more closely with company-specific information.
But every one of them depends on the quality of the data being introduced.
If the company’s documents are outdated, contradictory, poorly structured, or difficult to locate, the model will inherit those weaknesses.
Connecting an AI system to organizational data does not automatically create reliable intelligence. It can also create a faster and more persuasive way of distributing existing confusion.
Before introducing AI into important business processes, leaders must understand what information the model is using, who owns that information, how frequently it is updated, and how the resulting answers will be checked.
The AI strategy cannot be separated from the data strategy supporting it.
He offers a practical definition of what a successful data strategy should ultimately achieve:
The right data should reach the right person at the right time.
When those conditions are met, the organization has a much better chance of making the right decision.
Timing matters because even accurate information can lose value when it arrives too late.
A report that reveals a customer was likely to leave after the customer has already cancelled provides historical understanding, but it does not create an opportunity to intervene.
An inventory warning received after stock has already run out does not help the organization prevent lost sales.
A website insight delivered after a major campaign has ended cannot improve the campaign that generated it.
Effective data strategy is therefore not only about collecting information or creating dashboards. It is about connecting information to a person who has the authority and opportunity to act while the decision still matters.
This is where technology should support the business.
The objective is not to build the largest data platform or the greatest number of reports. It is to reduce the distance between a meaningful signal and an informed action.
He traces the development of analytics through a progression that helps leaders understand where their organizations currently operate.
The first stage is descriptive analytics.
Descriptive analytics explains what happened. It includes historical reports, dashboards, operational summaries and performance measurements.
This stage is necessary because an organization cannot understand what it does not understand. But descriptive analytics primarily look backwards.
The second stage is predictive analytics.
Predictive analytics uses historical patterns, statistical methods, and machine-learning models to estimate what is likely to happen next.
It may predict demand, identify customers at risk of leaving, estimate equipment failure, or forecast financial performance.
Prediction gives the organization time to prepare, but it still leaves an important question unanswered:
What should be done in response?
That is the role of prescriptive analytics.
Prescriptive analytics connects a likely outcome with a recommended or initiated response. It moves the organization from knowing what may happen to determining what action should follow.
Agentic AI is beginning to extend this capability by allowing systems to interpret information, propose actions, and, in controlled circumstances, initiate parts of a workflow.
However, the goal should not be to remove human judgement wherever possible. The goal is to determine where automation is appropriate, where human review is essential, and where the combination of both produces a better result.
Moving from descriptive reporting to predictive and prescriptive capabilities requires more than purchasing an AI product.
It requires reliable data, clear business objectives, appropriate governance and confidence in the actions produced from the analysis.
One of Buck’s most practical warnings is that organisations should use the entire analytical toolbox rather than treating AI as the answer to every problem.
Some business questions may benefit from a large language model. Others may be answered more accurately, economically and transparently through SQL, linear regression, logistic regression, statistical testing, or conventional machine-learning techniques.
He discusses Benford’s Law as one example. Benford’s Law examines the frequency with which certain leading digits appear across naturally occurring numerical datasets. When a dataset significantly departs from the expected pattern, the result may justify further investigation. It has been used as one of many methods for identifying unusual patterns in financial information.
A company does not necessarily need a large language model to perform that type of analysis. A well-designed query or statistical calculation may provide the required answer more efficiently.
The correct tool is determined by the business question, the available data, the required level of accuracy, and the consequences of an incorrect result.
Choosing AI simply because it is the most visible technology is another way of starting with how instead of why.
One of the most useful ideas Buck shares comes from an anecdote he associates with former US President Barack Obama.
When faced with large and complex problems that could not be solved completely, the question became whether something could still be done to improve the situation.
The principle was simple:
Better is good.
He has adopted this as an operating philosophy for data strategy.
Many organizations begin transformation programs with messy information, disconnected systems, unclear ownership and inconsistent processes.
The instinct is often to design a complete solution that addresses everything at once.
That leads to long transformation programs with distant outcomes. By the time the solution is delivered, business requirements may have changed, executive sponsors may have moved on, and users may no longer trust that the project will solve their problems.
He prefers visible and continuous improvement.
First, identify what the organization currently has.
Then identify what it needs.
Find one meaningful problem that can be improved and begin there.
The first step may not solve the entire data challenge, but it moves the organization closer to the desired outcome. It also demonstrates that the data function can deliver value.
Each visible improvement builds credibility. That credibility makes the next investment easier to support and the next organizational change easier to introduce.
He does not reject maturity models. He understands their purpose and recognizes that their stages can be valuable.
His concern is that organizations often struggle to maintain the discipline required for a rigid, multi-year transformation while continuing to manage the daily demands of the business.
A strategy must therefore be structured enough to create direction but flexible enough to continue producing value when circumstances change.
Progress should not be postponed until perfection becomes possible.
Across Buck Woody’s journey, one principle connects every lesson.
Data strategy begins with a business decision, not a technology purchase.
Leaders must first define what the organization will do differently when the required information becomes available.
From there, they must identify what data is needed, determine whether that data can be trusted, and select the most appropriate analytical or technological approach.
They must recognize that AI is probabilistic rather than deterministic and govern its outputs according to the consequences of being wrong.
They must build toward an environment where the right information reaches the right decision-maker while action is still possible.
And they must continue delivering visible improvement rather than waiting for a perfect transformation to arrive.
Data strategy does not begin with a platform, architecture, or AI model.
It begins with a decision.
What will the organization do differently when the answer becomes available?
Until that question is clear, every technology conversation is premature.
Once it is clear, as he argues, how it becomes much easier to determine.
Buck Woody, joined The Executive Outlook, shares what five decades in technology have taught him about data strategy, AI, and decision-making. Are you a data, technology, or business leader with a perspective that deserves a wider audience? Get featured on The Executive Outlook and share your story with global executives.
Listen in for business strategies and interviews with business experts to help you build a business and a life that you love.
About Us
