00:00 Intro02:12 Why stakeholder requests are often wrong05:09 Business discovery before solutions10:37 Treating dashboards as data products17:00 When you disagree with the stakeholder21:50 The Mom Test and better discovery meetings28:02 Teaching data professionals to really listen40:09 Can junior analysts think without AI?44:44 Management starts with listening48:06 Culture, communication and consultingGreat data professionals do not just take requirements. They listen closely enough to understand the problem behind them.In this episode of The Data Hustle, we speak with Gabriela Costa, an analytics engineer and team lead at Indicium AI, about why communication and discovery are as important to good data work as SQL, modeling or engineering.Gabriela explains why immediately accepting “I need a dashboard” can send a data team in completely the wrong direction. Instead, she separates business discovery from data discovery: understanding how somebody works today, what decisions they are trying to make, what constraints they face and what problem they are actually trying to solve before deciding what should be built.We discuss treating dashboards and other analytical outputs as living data products rather than one-off projects, measuring adoption, uncovering requirements stakeholders themselves may not know they have, and what to do when your professional judgment tells you that the solution a stakeholder is asking for is the wrong one.We also get practical about stakeholder interviews. Gabriela shares how ideas from The Mom Test influence her discovery process, why scripted question lists can make junior consultants worse listeners, how she prepares teams for client conversations, and why understanding who actually owns a decision can prevent endless cycles of conflicting requirements.Later, we get into one of the more uncomfortable questions facing junior data professionals today: are people becoming too dependent on AI? Gabriela talks about analysts presenting code they cannot explain beyond “ChatGPT told me to do it,” and why using AI effectively still requires understanding the logic well enough to question, validate and defend what it produces.Finally, we talk about management, genuinely listening to the people you lead, and how Gabriela’s background in international relations shaped the way she approaches negotiation, different cultures, competing perspectives and consulting.Find Gabriela:LinkedIn: Gabriela Costa — @gabrieladadosIndicium AI: indicium.ai

00:00 Intro01:09 From test automation to web data03:27 Trust, guardrails and AI seatbelts13:27 When agile automation becomes fragile21:39 Restoring trust in messy systems28:14 Who owns the open web?39:21 Deterministic AI for regulated workflows48:12 What’s next?AI is moving from generating answers to making decisions. That changes the stakes.In this episode of The Data Hustle, we speak with Sarah McKenna, CEO of Sequentum, about why AI needs the equivalent of seatbelts: clear guardrails, acceptance criteria, audit trails, deterministic workflows and accountable human review.Drawing on two decades across software test automation, DevOps, data quality and enterprise web extraction, Sarah explains why automation becomes fragile when teams optimise for speed without building in trust. We discuss reusable components, versioning, validation, low-code visibility, security reviews and the role of AI coding tools in maintaining large-scale data systems.We also zoom out to the growing conflict around the open web: publishers, AI crawlers, web-scraping companies, bot identification, paid data access and agentic commerce. Sarah explains why regulated organisations often reject AI at runtime, even while using it aggressively to accelerate development and strengthen governance.Find SarahLinkedIn: https://www.linkedin.com/in/sarahransommckenna/Sequentum: https://www.sequentum.com/

00:00 Intro00:49 From consulting to building a product05:00 Finding the right co-founder07:07 What data teams should delegate to AI11:21 Context engineering and the future of dashboards18:36 Avoiding the agent cemetery23:59 Do you need a semantic layer first?29:47 Inside Nao’s open-source analytics agent38:22 How data professionals create visible value44:48 What’s next for Nao?AI agents promise to let anyone analyse company data through a conversation—but reliable agentic analytics requires far more than connecting an LLM to a database.In this episode of The Data Hustle, we speak with Claire Gouze and Christophe Blefari, co-founders of Nao, about building an open-source framework for agentic analytics. They explain how their experience in consulting and data leadership exposed a recurring problem: business teams want answers faster, while data teams remain trapped as overloaded bottlenecks.We discuss which parts of data work should be automated, why analytics engineers may evolve into context engineers, whether traditional dashboards are becoming obsolete, and why companies risk replacing their dashboard graveyards with agent cemeteries.Claire and Christophe also break down the role of semantic layers, metadata and reusable context; why teams should not wait for perfect foundations before experimenting; how Nao structures its analytics agent; and how data professionals can use AI to become more visible and valuable rather than automate themselves out of the process.

Find Nao at:- https://getnao.io- https://github.com/getnao/nao- https://www.linkedin.com/in/claire-gouze/- https://www.linkedin.com/in/christopheblefari/