The Data-Driven Leader – Leveraging Data and AI to Create Business Impact with Sandro Saitta

Executive Summary

This webinar will address AI’s limitations, emphasise a problem-first implementation approach, and stress the importance of multi-dimensional maturity in AI adoption and employee upskilling. Dr Sandro Saitta will share his insights into the critical role of data governance in successful AI initiatives. Attendees will explore the broader AI landscape beyond tools like ChatGPT and learn to build an effective data roadmap. The session will conclude with essential project frameworks and phases, an analysis of AI’s limitations, and an emphasis on a problem-first approach to implementation.

Webinar Details

Title: The Data-Driven Leader – Leveraging Data and AI to Create Business Impact with Sandro Saitta
Date: 2026-02-09
Presenter: Sandro Saitta
Meetup Group: DAMA SA User Group Meeting
Write-up Author: Howard Diesel

Introduction & Session Overview

The webinar opens with Howard Diesel welcoming Dr Sandro Saitta, an accomplished data science leader with over 20 years of experience working with global brands like Nespresso and Expedia. Sandro brings a unique perspective that bridges both industry and academia, currently serving as a lecturer at HEC Lausanne and HEG Geneva, while advising companies through viadata on their data transformation journeys.

Sandro introduces his book, “The Data-Driven Leader,” which distils decades of lessons learned from both successful transformations and costly failures across industries. The book is thoughtfully structured into three essential sections, providing leaders with practical, actionable frameworks rather than theoretical concepts.

This approach reflects Sandro’s core philosophy: data leadership requires understanding not just the technology, but the business context, organisational dynamics, and human factors that ultimately determine success or failure. His cross-sector experience provides invaluable insights applicable to organisations of all sizes and maturity levels, setting the stage for a session focused on creating real business impact rather than chasing technological trends.

Figure 1 The Data-Driven Leader

Figure 2 Helping Leaders turn Data & AI into Business Impact

Figure 3 Enabling Organisations to Unlock Value through Data & AI

AI Beyond ChatGPT & Building Your Data Roadmap

Sandro emphasises the need for leaders to move beyond the AI hype cycle and focus on what truly delivers business value. He presents a sobering statistic from the 2025 MIT report: 95% of companies implementing generative AI are seeing zero return on investment. This reality check frames a crucial discussion about strategic AI adoption.

Three complementary approaches to data are introduced: Descriptive analytics helps understand what happened through visualisation and clustering. Predictive analytics forecasts future outcomes, enabling proactive decisions. Generative AI creates new content from text to images. The critical insight? Organisations fixating solely on generative AI miss valuable opportunities in descriptive and predictive methods that often deliver faster ROI.

To build an effective data strategy, Sandro presents his proven 3-step process: Generate Ideas through structured brainstorming that brings together stakeholders and data teams. Define Use Cases using a canvas that clarifies the business context, value proposition, KPIs, and success metrics before coding begins. Prioritise Projects with an impact-versus-feasibility matrix, allocating resources to initiatives with the highest potential return. Successful strategies also balance top-down business needs with bottom-up data capabilities, finding the “sweet spot” where strategic objectives align with available data assets. This intersection produces projects with both clear business value and technical feasibility.

Figure 4 What’s in It for You?

Figure 5 AI – the Solution to All Your Problems

Figure 6 MIT Report – AI Failure?

Figure 7 Generative AI – Where is the Impact?

Figure 8 Three Approaches to Leveraging Data

Figure 9 How to Build the Roadmap?

Figure 10 Generating Ideas – Top-down and Bottom-up

Figure 11 Defining Use Cases – the Data Initiative Canvas

Project Frameworks, Phases & Domain Expertise

Providing some deep insights into practical frameworks, Sandro separates successful initiatives from expensive experiments. The Data Initiative Canvas includes critical elements: business context, value proposition, key performance indicators, required data sources, technical approach, stakeholders, risks, and resource estimates. This structured tool prevents common pitfalls like building solutions, searching for problems, or starting projects without clear success criteria.

Understanding the AI project lifecycle helps leaders set realistic expectations. Sandro introduces CRISP-DM’s six phases: Business Understanding, Data Understanding, Data Preparation, Modelling, Evaluation, and Deployment. Critically, this isn’t linear but iterative—insights from one phase often require revisiting earlier ones. Leaders should budget for this iteration rather than viewing it as a failure.

A crucial revelation: modelling should represent only 10% of total effort. The remaining 90% involves understanding, preparation, and deployment, all of which require domain experts. Domain experts know which data is reliable, which patterns represent true insights versus quirks, and which outputs make business sense. Smart organisations embed domain experts throughout the entire lifecycle, not just requirements gathering.

An attendee then raises an important question about consistently measuring ROI across projects. Sandro responds by emphasising the need to define value metrics upfront in the canvas, using a consistent methodology across projects, and distinguishing expected from realised value through regular retrospectives. This discipline builds credibility and enables continuous improvement.

Figure 12 Prioritising Projects – Impact Vs. Feasibility Matrix

Figure 13 The Key Steps of Any AI Project

Figure 14 Success Depends on Your Ability to Ask the Right Questions

Is AI Really Intelligent? Understanding Limitations

While examining AI’s actual capabilities versus marketing hype, Sandro believes leaders should be grounded in reality to ensure responsible deployment. He presents compelling examples revealing fundamental limitations. The Wolf vs. Husky case study shows how an image classifier “cheated” by detecting snow in backgrounds rather than animal features—achieving high accuracy for wrong reasons. This illustrates that AI learns patterns without understanding causation or context.

The resume screening exploit demonstrates how applicants game AI systems using white-on-white text invisible to humans but parsed by algorithms. AI lacks common sense regarding manipulation or appropriateness and processes inputs mechanically. Sandro also explores ChatGPT’s random number bias, showing how it favours certain numbers, such as 42, when asked for random output—revealing that probabilistic models don’t truly generate randomness.

The calculation risk highlights why using probabilistic tools like generative AI for deterministic tasks (mathematics, Excel formulas) is dangerous. The Müller-Lyer illusion illustrates how AI replicates human cognitive biases in training data. Finally, discussions of ethical dilemmas, such as self-driving car scenarios, reveal that AI struggles with complex moral reasoning that requires human judgment. These examples aren’t meant to discourage AI adoption but to foster realistic expectations. Leaders must understand both the powers and limitations, implement appropriate governance, test for spurious correlations, and maintain human oversight for high-stakes decisions.

Figure 15 Is AI Really Intelligent?

Figure 16 Wolves and Huskies

Figure 17 Tricking the System

Figure 18 Generating Random Numbers

Figure 19 Calculating in Excel

Figure 20 Analysing Images – Which Line is Longer?

Figure 21 The Challenge of Bias

Figure 22 ChatGPT Cannot Solve All Problems

Practical Tips: Start with Problems, Not Technology

The first actionable guidance tip for leaders seeking genuine business impact is: Start with business problems, not technology. Flip the question from “What can we do with AI?” to “What problem are we trying to solve?” This problem-first approach ensures initiatives deliver measurable value rather than becoming technology experiments searching for application.

Data quality matters more than perfect data. Sandro introduces the concept of “fit for purpose” data—data of sufficient quality for your intended use, rather than unattainable perfection. Many companies sit on gold mines of data they can’t use because they’re waiting for perfect cleanliness. Start with good-enough data and improve iteratively.

Follow the pain, not the trend. Sandro reviews hype cycles from 2000 to today—Machine Learning, Big Data, GenAI. The core challenge across all these trends? Good data quality. Organisations that address fundamental data issues position themselves to leverage any emerging technology, while those chasing trends without foundations repeatedly fail.

Leverage data visualisation first. Before jumping to complex AI models, focus on exploratory and explanatory visualisation. Simple, well-designed charts provide profound insights quickly and transparently. Know when to use AI: when manual coding is impossible, data volume overwhelms humans, or abundant training data exists. Avoid AI when you need predictability, explainability, or perfect reproducibility.

Figure 23 How Far Should AI Decide for Us?

Figure 24 Practical Tips for Leaders

Figure 25 Start with the Business Problem

Figure 26 Work on Data Quality

Figure 27 Follow the Pain, Not the Trend

Figure 28 Leverage Data Visualisations

Figure 29 Know When (Not) to Use AI

Multi-Dimensional Maturity & Upskilling Employees

Successful transformation extends beyond technology. Sandro presents his multi-dimensional maturity model encompassing five critical dimensions: Data (quality, governance, accessibility), Skills (data literacy across the organisation), Organisation (culture, roles, operating models), Tools (technology stack enabling data work), and Processes (how data flows through and informs decisions).

The critical insight: overall maturity is constrained by your weakest dimension. Organisations might invest millions in cutting-edge tools, but if employees lack literacy or silos prevent collaboration, initiatives fail. Leaders must assess all dimensions honestly, identify bottlenecks, and invest proportionally for balanced progress.

Regarding the Organisation dimension, Sandro emphasises culture, clear data ownership, decision rights, and cross-functional collaboration. Upskilling employees follows a “no pain, no gain” reality—becoming data-driven requires genuine effort. The key is training non-data people to become data-aware rather than turning everyone into data scientists.

Coaching versus classroom training becomes a crucial distinction. Hands-on coaching on real projects proves far more effective than one-off classroom workshops for long-term retention and behaviour change. When employees work with coaches on actual business problems using real company data, learning sticks and capabilities develop organically. This experiential approach transforms organisational culture more effectively than any amount of theoretical training.

Figure 30 Keep in Mind the Multiple Dimensions

Figure 31 Upskill Your Employees

Figure 32 My Key Message to Keep in Mind

Key Takeaways & Book Giveaway

AI is a tool, not magic. While AI possesses remarkable capabilities for pattern recognition and prediction, it also carries significant risks and limitations. Leaders must maintain realistic expectations and understand both power and boundaries. Over-promising undermines organisational trust in data initiatives.

Data and people are the real challenges. Technology is increasingly commoditised—powerful AI tools are widely available. Genuine competitive advantage lies in quality data and people who effectively leverage it. Organisations investing in data infrastructure and human capability consistently outperform those chasing the latest trends.

Success requires three foundations: data quality appropriate to your use case, widespread data literacy that enables employees to interpret and act on insights, and the organisational ability to ask the right questions that connect capabilities to business needs.

The Q&A continues with fascinating discussions on prompt engineering across languages, where Sandro explores how AI handles different languages and whether prompt precision matters across cultures. He explains that while models perform differently with varying training data volumes, the principles of clear communication remain universal for effective prompting.

Figure 33 Key Takeaways

Figure 34 Closing Slide

Deep Dive: Data Governance as Foundation

The session concludes with an extensive, critical discussion of data governance, which Sandro considers foundational to successful AI implementation. Participants raise thoughtful questions about governance challenges, and Sandro provides comprehensive insights drawn from decades of experience.

Data governance encompasses policies, procedures, roles, and responsibilities that ensure data quality, security, privacy, and appropriate use. Without strong governance, even the most sophisticated AI initiatives crumble. Governance provides the framework for data democratisation—making data accessible to those who need it while maintaining appropriate controls.

Sandro distinguishes between immature and mature governance programs. Immature programs focus on restrictive policies that frustrate users, creating shadow IT and workarounds. Mature programs balance accessibility with responsibility, empowering employees to use data confidently while protecting organisational interests. This requires clear data ownership, documented lineage showing data sources and transformations, quality metrics and monitoring, and role-based access controls.

The final discussion emphasises that governance isn’t a one-time implementation but an evolving practice. As organisations mature, governance becomes embedded in culture rather than enforced through compliance. Leaders should invest early in governance foundations, even if starting small. Sandro’s final recommendations are to define data owners, establish quality standards for critical datasets, and create simple processes for data requests. Organisations that treat governance as an afterthought inevitably face costly remediation, compliance issues, or catastrophic data breaches.