About This Episode
In episode 7 of Data & AI Heads, host Ian Allison sits down with Dawar Dedmari, a data engineering and analytics leader with over 15 years of experience spanning Meta's Reality Labs, Amazon, Oracle, and Accenture. Dawar unpacks his "compass and engine" model for positioning data teams inside an organization, explaining how teams must shift from fixating on tech stacks to deeply understanding executive goals, business inputs and outputs, and the levers that connect them. He shares concrete frameworks for earning executive trust, measuring data team impact through organizational outcomes rather than internal metrics, and building goal maps that make the right decisions automatable and the complex ones faster.
The conversation moves into the realities of AI adoption, where Dawar draws on firsthand experience to separate genuine use cases, such as synthetic data generation and anomaly detection, from overhyped vendor promises like enterprise chatbots built on top of poorly modeled data. He argues that a well-constructed data model and proper data labeling remain 90 percent of the real work, and that no LLM layer rescues a weak data foundation. The episode closes with career advice grounded in Dawar's own journey from chip design through the full arc of data warehousing, data lakes, and cloud infrastructure: invest in foundational skills over specific tools, adopt an ownership mindset, and above all, build the capacity to keep learning.
This episode is essential listening for data leaders who want to align their teams with business value, evaluate AI investments honestly, and develop talent that stays relevant through every technology cycle.
Key Takeaways
- Data teams should operate as both a compass, providing directional intelligence across all business functions, and an engine for growth, identifying inefficiencies and amplifying what is working.
- New data leaders should prioritize executive engagement in their first weeks, mapping business inputs, levers, and outputs before touching the tech stack, so the team directly serves organizational goals.
- The most credible way to measure data team success is to connect the team's output to top-line organizational goals, such as incremental revenue driven by data-informed programs, rather than relying solely on internal metrics like dashboard uptime or SLA compliance.
- AI chatbot tools built on top of poorly labeled enterprise data create real organizational risk, including executives receiving and distributing hallucinated financial figures, so strong data modeling and governance must precede any generative AI layer.
- Legitimate near-term AI wins for data teams include synthetic data generation, assistive code generation, boilerplate dashboard creation, and anomaly detection and monitoring.
- Goal maps that chart every input, lever, and output metric for a business can reveal which decisions are fully automatable and which require a human, making it possible to put the right information in front of decision-makers at exactly the right moment.
- Foundational skills, such as telling a clear story with data, transfer across any tool or platform, while proficiency in a single technology is easily made obsolete as the industry evolves every two to three years.
- The fastest path to career growth in data is combining theoretical grounding with hands-on project work and adopting an ownership mindset, volunteering to solve problems outside your defined scope rather than waiting to be assigned.
What We Cover
Frequently Asked Questions
How should data teams position themselves within an organization to earn executive respect?
According to Dawar Dedmari, data teams should operate under a dual mandate he calls the compass and engine model. As a compass, they provide directional intelligence to every business function. As an engine, they identify inefficiencies and model processes to supercharge growth. Earning executive trust requires early engagement with senior leaders to understand their goals and pain points, then demonstrating how the data team helps them achieve those goals rather than showcasing internal technical achievements.
What is the input-lever-output model for data teams?
The input-lever-output model, described by Dawar Dedmari, is a framework for understanding how a business actually works. Data leaders identify the inputs that drive a business, the levers the company can adjust, and the output metrics that define success or failure. By mapping these relationships and building measurements across each layer, data teams can provide intelligence on exactly which dials to turn to move the business in the direction leadership wants. This model also reveals which decisions can be automated and which require human judgment.
How can data teams prove their value to executives beyond internal data quality metrics?
Dawar Dedmari argues that executives rarely care about internal data metrics like SLA compliance or dashboard uptime. He recommends two approaches that resonate more strongly: tracking adoption and usage of data products by the business teams they serve, and directly connecting data team work to organizational outcomes. In one example he shared, data-driven programs including better targeting and A-B tested interventions drove hundreds of thousands of incremental sales conversions and millions in additional revenue, creating a clear ROI case for the data team's contribution.
What are the real risks of deploying AI chatbots on enterprise data?
Dawar Dedmari warns that AI chatbot tools are frequently marketed beyond their actual capabilities, and the underlying problem is rarely the AI layer itself. In enterprise environments, data is almost never clean, consistently labeled, or structured the way vendor demos suggest. He cited a real incident where an executive queried a chatbot on a financial question, received an incorrect answer, and circulated it in a document, creating significant organizational risk. His position is that strong data modeling and data labeling must be in place before any LLM-based tool can function reliably.
Where is AI actually delivering value for data teams today?
Dawar Dedmari identifies several practical areas where AI is already helping data teams. Synthetic data generation has been particularly valuable where training data at scale is difficult to obtain. AI also works well as an assistive technology for generating boilerplate code and dashboards, and it has shown real utility in monitoring and anomaly detection. He contrasts these concrete use cases with more speculative applications, noting that some capabilities are still more fiction than fact at this stage.
What career advice does Dawar Dedmari give to data professionals navigating rapid technology change?
Dawar Dedmari advises data professionals to invest in foundational skills rather than specific tools, because tech stacks change every two to three years. He uses the example of chart-building: learning how to tell a compelling story with data transfers across Tableau, Power BI, Python, or any future tool, whereas mastery of a single platform has limited shelf life. He also stresses the importance of an ownership mindset, proactively solving problems outside your defined scope, which he credits directly with expanding his own career opportunities including taking on data governance work that was not originally part of his role.
What is the most effective way to develop skills in data and analytics?
Dawar Dedmari emphasizes that nothing replaces actually doing the work. Regardless of whether someone learns best through reading, video, or courses, the critical step is moving from theory to hands-on practice by building real dashboards, writing actual code, and constructing real models. He compares it to swimming: watching videos can provide a starting point, but skill only develops in the water. Ian Allison reinforced this from his own teaching experience, finding that learners given an open-ended project and pointed to resources consistently outpaced those given step-by-step instruction.
About the Guest
Dawar Dedmari
Dawar Dedmari is a data engineering and analytics leader with over 15 years of experience building and scaling data organizations across major technology and enterprise companies. He currently leads AR, VR, and AI data initiatives at Meta's Reality Labs, where his work spans product launches including the Quest headset and Ray-Ban smart glasses lines. Earlier in his career he held data leadership roles at Amazon, where he was part of the generational shift from on-premises databases to cloud-based data warehousing on Redshift, as well as at Oracle and Accenture, where he built data engineering teams, implemented Kimball-principle data warehouses, and led large-scale data governance programs. His career spans the full arc of the modern data stack, from traditional ETL and dimensional modeling through data lakes, big data, Spark, and Delta lakes. He is also a mentor focused on developing the next generation of data professionals and is based in the Pacific Northwest.
Connect on LinkedIn →More Episodes