AI Meets CFO

AI Meets CFO

Each episode Amendra Pratap sits down with a CFO, finance director, or finance-technology practitioner to talk through how they actually run their function: the tools they rely on, the tech stack decisions behind the scenes, the moments that tested them, and how they're thinking about AI's role in finance's future. No scripts, no sales pitches; just practitioner-to-practitioner conversation for people who make real budget and technology decisions every day. 🔔 Subscribe for new episodes

Episodes

  1. 1d ago

    Episode 5: When Not to Use AI: Lessons from Global Cost Forecasting | Yenny Dewi

    Sanofi has been forecasting sales with machine learning for years. Yenny Dewi explains why cost forecasting is harder, and how operational data, not AI, delivered 98% accuracy on headcount. Yenny Dewi leads AI transformation in finance at Sanofi, with a focus on planning and forecasting. She explains why a single global planning model became the foundation for AI, how a culture of experimentation reshaped their approach to cost forecasting, and why every AI project should start with a problem statement. What does AI in finance look like at a global pharmaceutical company? Yenny Dewi has spent more than 20 years working with planning systems, and now heads a global AI transformation role in finance at Sanofi. In this episode, she talks with host Amendra Pratap about what has worked, what hasn't, and why the purpose comes before the technology. Yenny explains how Sanofi brought many local planning systems into one global core model, with a standard chart of accounts and master data, and why that foundation came before any serious AI work. She shares how machine learning sales forecasts now sit inside integrated business planning, and why cost forecasting is a much harder problem. On headcount, the team learned that operational signals from the people system gave 98% accuracy without AI, so AI is now aimed at narrower questions like attrition and time to hire. In this episode: Why Sanofi uses AI: shortening the path from scientific discovery to patientsMachine learning sales forecasts as a baseline, with people making the final callWhy cost forecasting is harder than sales forecastingBuilding trust and adoption by starting with the pain pointsBalancing a global core model with local business needsA feedback loop after every budget cycleClean, standardised data as the foundation for AIWhy leadership support makes or breaks a transformationTechnology is a tool. The people decide whether it worksTM1 for planning, Power BI for storytelling, Python for custom machine learning98% headcount accuracy from operational data, with no AI neededCuriosity and critical thinking as the skills that matterHow the finance role will change, and AI use cases beyond finance About the guest: Yenny Dewi leads AI transformation in finance at Sanofi, one of the world's top 10 pharmaceutical companies, with a focus on planning, forecasting and decision support. She has more than 20 years' experience with planning systems across FMCG, retail and pharmaceuticals, and more than 12 years at Sanofi across five or six roles. Host: Amendra Pratap 00:00 - Welcome and introductions 02:46 - Yenny's AI transformation role 03:31 - Why Sanofi uses AI 06:16 - Building trust in AI 07:52 - Global core model vs local needs 10:41 - Feedback loops after every budget cycle 13:46 - What has improved most 15:46 - Data quality as the AI foundation 16:35 - Challenges along the way 17:25 - The role of leadership 18:59 - People and purpose over technology 21:56 - The finance tool landscape 23:07 - Custom vs embedded AI 23:52 - Cost forecasting use cases 26:55 - A culture of experimentation 27:42 - The skills finance teams need 29:25 - The future size of finance teams 31:54 - What comes next: from prediction to agents 33:36 - AI beyond finance 35:10 - How the business will see finance 36:33 - Final advice: start with the problem

    Episode 5: When Not to Use AI: Lessons from Global Cost Forecasting | Yenny Dewi
  2. Nov 10

    Episode 4: Visibility, Control, Accountability: Taming AI Sprawl | Ian Godsell

    Data has gravity. Ian Godsell explains why so many AI pilots never reach production, and why governance speeds AI adoption up rather than slowing it down. Ian Godsell, a partner technical specialist in data and AI at IBM, unpacks what enterprises need before AI can deliver: trusted data, clear governance and a business case that stacks up. He covers data lineage and quality, data sovereignty, token costs, ROI and a simple way to start on AI risk. Why do so many AI proofs of concept stall before production? According to Ian Godsell, it usually comes back to data. In this episode, the IBM partner technical specialist joins host Amendra Pratap to explain what a strong data foundation looks like and why finance should care about AI governance. Ian sets out the building blocks of data intelligence: lineage, quality, a semantic layer and the ability to share data products safely. He explains context engineering, the growing role of unstructured data and real-time data, and why data sovereignty is becoming a board-level question. He also warns against using AI where a simple business rule will do, and explains how visibility, control and accountability help teams scale AI with confidence. In this episode: AI has been around for decades. What changed is how visible it became to executivesEmbedding AI in software versus helping enterprises build their own agentsTrust, transparency and keeping humans in the decisionWhy a weak data foundation is the top barrier to AI adoptionData lineage, data quality and the semantic layerContext engineering: the right data, at the right time, to the right person or agentData residency, AI sovereignty and keeping customer data out of model trainingEmbedded agents, MCP servers and open standardsYou don't need AI for a fixed business ruleSaaS, on-premises and running workloads where they make senseToken costs, the business case and checking that the ROI stacks upAI governance as an accelerator: visibility, control and accountabilityAgent sprawl and the case for a single point of governanceWhere to start: building a risk register for your AI use cases About the guest: Ian Godsell is a partner technical specialist at IBM, focused on data and AI. He has more than 20 years at IBM and works with business partners to help them understand and apply IBM's data, AI and automation software. Host: Amendra Pratap 00:00 - Welcome and introductions 01:56 - How the AI landscape has changed 03:44 - Trust and transparency in AI 05:17 - Why the right data matters 07:27 - Data governance in finance 08:13 - Lineage, quality and the semantic layer 10:01 - Lakehouses and data integration 11:13 - Data residency and AI sovereignty 12:47 - AI agents embedded in software 14:31 - More agents than humans? 16:12 - When you don't need AI 17:11 - The future of SaaS and on-premises 19:46 - Open platforms and choice of model 21:18 - AI across the software development lifecycle 22:13 - Token costs and ROI 24:41 - AI governance as an accelerator 26:58 - Agent sprawl and fragmentation 28:04 - Visibility, control and accountability 29:11 - Where to start with AI governance

    Episode 4: Visibility, Control, Accountability: Taming AI Sprawl | Ian Godsell
  3. Nov 3

    Episode 3: Clean Data Before AI: Rebuilding Planning for the AI Era | Matt Clifford

    News Corp's first try at putting AI on its finance data came back untrustworthy. Matt Clifford explains what they rebuilt, and what finance should focus on once AI clears the build bottleneck. Matt Clifford leads finance systems and data at News Corp. He shares why an early AI trial failed on unlabelled data, how rebuilding a planning model that had been running since 1998 sped up scenario planning, and why the real challenge is now choosing which AI ideas to fund. What happens when you switch on AI over finance data that was never designed for it? At News Corp, the answers came back patchy and hard to trust. Matt Clifford, who leads finance systems and data there, tells host Amendra Pratap what the team learned and how they fixed the foundations. Matt came into finance from technology, starting in data and reporting at Australia's financial regulator. He explains how News Corp rebuilt a planning model that had been in place since 1998, ran old and new side by side, and brought users along by asking why they kept jumping to Excel. He then looks ahead: the "business partner in your pocket" idea, why data governance has become the foundation for AI, why locking into one model for more than 12 months is a risk, and how AI changes the question from "what can we build?" to "what should we build first?" In this episode: From technology to finance: starting in data and regulatory reportingWhy the first attempt at AI on finance data failedRebuilding a decades-old planning model without disrupting the businessGetting users to trust the platform before they jump into ExcelExecutive sponsorship and making the platform the users' ownKeeping a lean core stack but staying flexible on reporting toolsThe "business partner in your pocket": instant answers from trusted dataWhy data governance now underpins every AI moveToken costs and the real problem: prioritising ideas when build is no longer the constraintAvoiding lock-in to a single model or vendorOnboarding AI agents like team membersAI forecasting, from complex models to quick baselines for everyoneHas finance reached its peak size?Advice for graduates and for leaders frozen by the pace of change About the guest: Matt Clifford leads finance systems and data at News Corp Australia. He has worked in data and technology since the 2000s, including leading a business analysis practice at Australia's financial regulator. Host: Amendra Pratap Timestamp and Chapters: 00:00 - Welcome and introductions 01:23 - Data and regulatory reporting beginnings 02:08 - From tech into finance 02:38 - Matt's role at News Corp 04:11 - The first attempt at AI, and why it failed 05:35 - Rebuilding a planning model in place since 1998 07:48 - Excel's place in the new platform 08:43 - Bringing people along on the change 10:13 - Support and learning from others 12:10 - The tool stack 13:10 - Being willing to change tools 14:32 - Finance as a strategic partner 15:58 - Moving past the AI hype 17:50 - Use cases: a business partner in your pocket 19:01 - Data governance comes first 20:53 - Security, cost and prioritisation 23:04 - How finance teams feel about AI 25:27 - Avoiding lock-in to one model 27:11 - More agents than people? 28:53 - AI forecasting 30:39 - AI in legacy and digital businesses 31:42 - Has finance reached its peak size? 32:43 - Advice for graduates 33:35 - Advice for leaders who feel stuck 34:47 - Will AI replace finance tools? 36:18 - The cost of AI as hyperscalers seek returns 37:44 - What to focus on next

    Episode 3: Clean Data Before AI: Rebuilding Planning for the AI Era | Matt Clifford
  4. Oct 27

    Episode 2: Good Actuals First: A CFO on Forecasting Through a Crisis | Isabella Susilowati

    Isabella Susilowati, CFO at O'Brien Glass, has spent more than 20 years in finance, including 14 at O'Brien. She explains why good actuals come before any forecast, how her team used TM1 and Power BI through a demand shock, and the simplest way for finance people to start with AI. Where should a finance team actually start with AI? For Isabella Susilowati, the answer is simple: download a tool and use it. In this episode, the O'Brien Glass CFO talks with host Amendra Pratap about the technology that excites her most, TM1, Power BI and AI, and how she brings her team along. Isabella explains why clean historical data is the base for every forecast, and why she believes no business can fully get rid of Excel. She shares how TM1 gave her team one source of truth when demand dropped sharply and decisions had to be made daily. She also covers AI policy for finance, keeping a human in the loop, and the leadership lessons that shaped her, from giving people more than they think they can handle to always explaining the why. In this episode: Isabella's path from FP&A manager to CFO over 14 years at O'Brien GlassWhy AI is like the internet 20 years ago, and the risk of being left behindThe first step into AI: download a tool and use AI to teach you about AIThe seven levels of AI mastery, from asking questions to agents that run in the backgroundSetting an AI policy: what data can go in, and which platforms are safeWhy human in the loop still mattersGetting actuals right in TM1 before you forecastWhy some forecasting still lives in Excel, and how to keep it tidyUsing TM1 for daily reforecasting through a sudden drop in demandBuilding Power BI on a clean data warehouse layerLeadership: stretch assignments, honest feedback and explaining the whyWhy teaching others makes you better at what you do About the guest: Isabella Susilowati is CFO of O'Brien Glass, part of the global Belron group. She has worked in finance for more than 20 years, including roles at BOC Gas and Origin Energy, and shares Power BI, Excel and AI tips on her YouTube channel, Lighthouse Analytics. Host: Amendra Pratap Timestamp & Chapter: 00:00 - Welcome and introductions 01:48 - Isabella's journey to CFO 02:51 - What O'Brien Glass does 03:19 - The tech stack and the ERP journey 04:28 - TM1, Power BI and the seven levels of AI mastery 06:25 - A message for AI sceptics 08:04 - How finance teams should get started 10:03 - AI governance and company policy 11:25 - More agents than people? 12:21 - Why human in the loop matters 12:48 - AI for forecasting 14:05 - Bringing the team along: sharing AI wins 15:36 - How the team uses TM1 18:03 - Reforecasting through a demand shock 19:27 - Power BI and the data warehouse 20:52 - Motivating a finance team 27:17 - Why she teaches: the teachers who shaped her 31:44 - The YouTube story 34:50 - Finance, fractional CFO work and what's next

    Episode 2: Good Actuals First: A CFO on Forecasting Through a Crisis | Isabella Susilowati
  5. Oct 20

    Episode 1: It's a Maths Model, Not a Monster: AI for Finance Leaders | Nick Renotte

    Nick Renotte from IBM joins Amendra to talk about broken spreadsheets, late-night month-end reports, and what AI can actually do for finance teams. Welcome to the first episode of AI Meets CFO. Host Amendra sits down with Nick Renotte, who leads AI developer work at IBM, runs a YouTube channel with 320,000 subscribers, and is also a Chartered Accountant who learned Excel macros at eight years old. They skip the sales pitch and talk about what actually works: • Why ChatGPT is just a chatbot until you connect it to your ERP and spreadsheets • How to get AI to draft month-end variance notes so you do not work late on Friday • Why asking giant AI models simple questions gives you a massive surprise bill • Why staff secretly use outside tools when IT takes three months to change a report • Why being scared of AI is like being scared of high school math formulas • How Nick connects AI directly to TM1 and IBM Planning Analytics Chapters: 00:00 - Meet Nick Renotte: 320k YouTube subscribers and IBM AI lead 02:24 - How a Chartered Accountant ended up building AI at IBM 04:11 - What an AI agent actually is (and why ChatGPT is just a chatbot) 07:20 - Why month-end reports are painful and how AI helps 09:34 - How simple AI questions turn into massive monthly bills 14:45 - The difference between personal AI tools and company tools 17:20 - Why Microsoft Copilot still has not fixed your job 21:00 - Why being afraid of AI is like being afraid of algebra 25:20 - Connecting AI directly to TM1 and Planning Analytics 31:44 - Where does your company financial data actually go? 37:51 - Why staff secretly use ChatGPT when IT is too slow 42:50 - Do you really need the most expensive AI model? 45:20 - The single easiest place for finance teams to start

    Episode 1: It's a Maths Model, Not a Monster: AI for Finance Leaders | Nick Renotte

About

Each episode Amendra Pratap sits down with a CFO, finance director, or finance-technology practitioner to talk through how they actually run their function: the tools they rely on, the tech stack decisions behind the scenes, the moments that tested them, and how they're thinking about AI's role in finance's future. No scripts, no sales pitches; just practitioner-to-practitioner conversation for people who make real budget and technology decisions every day. 🔔 Subscribe for new episodes