Data-Driven CFO Hosted discussion on AI

The Data-Driven CFO panel discussion looks at AI from a practical finance perspective: where generative AI sits on the hype cycle, which types of AI accountants need to understand and why many AI projects fail. It closes with advice for the data-driven CFO on starting small and building a sound business case.

Open Data-Driven CFO discussion with Public Sector CFOs and Accountants on AI, Hype and Business Outcomes. The discussion also included an expert in AI from Monash University.

Detailed Summary of the Sequel CFO Podcast Episode:

“AI, Hype & Business Outcomes with Chartered Accountants Australia & New Zealand”
(Hosted by David Boyar, featuring Professor Geoff Webb from Monash University)
Listen here

This episode, presents a thoughtful and professionally grounded panel discussion hosted by David Boyar, CEO of Sequel CFO, in partnership with Chartered Accountants Australia and New Zealand (CA ANZ). The session was designed specifically for accountants, financial leaders, and business advisors seeking clarity amid the overwhelming noise surrounding artificial intelligence (AI). Rather than contributing to the hype cycle, the panel deliberately dissects the realistic business outcomes of AI adoption, offering a roadmap for professionals to evaluate, implement, and measure AI-driven value without falling into common traps.


Panel Composition and Format

David Boyar, a seasoned CFO advisor and host of the SequelCFO podcast, serves as the moderator. Known for his pragmatic approach to finance transformation, Boyar frames the discussion around the daily realities of accounting practices—time constraints, regulatory compliance, client expectations, and ROI accountability.

The primary expert on the panel is Professor Geoff Webb, a globally recognized authority in machine learning and data science from Monash University. Webb is the Research Director of the Monash Data Futures Institute and has published extensively on scalable data mining, predictive analytics, and the limitations of AI systems. His inclusion ensures that technical explanations remain accessible yet rigorous, avoiding both oversimplification and jargon overload.

The format is a live panel recording, originally delivered as a webinar or in-person event for CA ANZ members, later adapted into a podcast episode. While the SoundCloud version is audio-only, the conversational flow suggests visual aids (e.g., slides on AI workflows) may have accompanied the live session.


Core Themes and Discussion Breakdown

1. The AI Hype Cycle: Where Are We Now?

The panel opens with a candid assessment of the current AI landscape. Boyar references the Gartner Hype Cycle, placing generative AI (post-ChatGPT) at the “Peak of Inflated Expectations” in 2023–2024, now sliding toward the “Trough of Disillusionment” by 2025. Real-world anecdotes from accounting firms illustrate this:

  • Firms investing six figures in AI consulting, only to achieve marginal automation in expense coding.
  • Marketing claims of “80% time savings” that evaporate under audit-quality requirements.

Professor Webb reinforces this by distinguishing between narrow AI (task-specific, reliable) and general AI (still largely science fiction). He warns that much of the “AI revolution” in business tools is rebranded automation—rules-based software with basic machine learning layered on top.

“Most of what’s sold as AI today is sophisticated pattern matching—powerful, but not intelligent in the human sense.” – Prof. Geoff Webb


2. How AI Actually Works (Demystified for Accountants)

Webb dedicates a significant segment to explaining machine learning fundamentals in accounting-relevant terms:

ConceptExplanationAccounting Example
Supervised LearningAlgorithm learns from labeled historical data to predict outcomesPredicting client payment delays using past invoice aging data
Unsupervised LearningFinds hidden patterns without predefined labelsSegmenting clients by risk profile from unstructured financial notes
Natural Language Processing (NLP)Extracts meaning from textAuto-categorizing bank statement descriptions
Anomaly DetectionFlags outliers in data streamsIdentifying fraudulent journal entries

He emphasizes that data quality is the bottleneck, not the algorithm. Garbage in, garbage out (GIGO) remains the #1 reason AI projects fail in professional services.


3. Practical Use Cases in Accounting & Advisory

The panel transitions into tangible, low-risk, high-value applications:

  1. Automated Data Extraction & Reconciliation
  • Tools like DocuClipper, Rossum, or custom ML models extract line items from PDFs/invoices with 95%+ accuracy.
  • Reduces manual entry from hours to minutes; audit trail preserved via confidence scoring.
  1. Predictive Cash Flow Forecasting
  • Using time-series ML (e.g., Prophet, LSTM models), firms forecast client cash flow with 20–30% improved accuracy over Excel trends.
  • Enables proactive advisory: “Your cash runway drops below 60 days in Q3—here are three mitigation strategies.”
  1. Risk-Based Audit Sampling
  • Replace random sampling with ML-driven risk scoring.
  • Example: Flag 100% of high-risk transactions (e.g., round-dollar amounts, weekend postings) while sampling only 10% of low-risk items.
  1. Client Query Automation (via Internal Chatbots)
  • Train a secure, firm-specific LLM on engagement letters, tax rulings, and prior advice.
  • Junior staff get instant, compliant draft responses; partners review exceptions.

4. The Business Case: ROI, Risks, and Governance

Boyar presses Webb on how to build a defensible business case. Key frameworks introduced:

  • AI Value Matrix (Impact vs. Feasibility)
  • High Impact + High Feasibility = Quick Wins (e.g., bank feed categorization)
  • High Impact + Low Feasibility = Strategic Bets (e.g., predictive insolvency models)
  • Total Cost of AI Ownership (TCAIO)
  • Includes: software licenses, data cleaning, change management, compliance reviews, and rework when AI confidence <90%.
  • Ethical & Regulatory Guardrails
  • APES 205 (Code of Ethics) and Privacy Act compliance when using client data for training.
  • Recommendation: Use synthetic data or federated learning to avoid privacy breaches.

5. Avoiding Common Pitfalls

The panel concludes with a “Top 5 AI Failure Modes” in accounting firms:

  1. Starting with Generative AI for Regulated Outputs (e.g., letting ChatGPT draft tax opinions).
  2. Underestimating Data Preparation (70% of project time).
  3. Treating AI as a Black Box (no explainability = audit failure).
  4. Vendor Lock-In with proprietary models.
  5. Ignoring Change Management—staff resistance kills adoption.

Key Takeaways for Listeners

  1. Start Small, Measure Ruthlessly: Pilot one process (e.g., creditor reconciliations), track time saved and error rates.
  2. Build Internal AI Literacy: Every senior accountant should understand bias, confidence intervals, and data lineage.
  3. Partner with IT & Data Teams: AI is a cross-functional sport.
  4. Think Advisory, Not Just Compliance: Use AI to deliver forward-looking insights—that’s where fees grow.
  5. Stay Skeptical, Stay Curious: Question vendor claims; demand proof-of-value in your data.

Final Thoughts from the Panel

David Boyar closes with a call to action:

“AI won’t replace accountants—but accountants who use AI strategically will replace those who don’t.”

Professor Webb adds a scholarly nuance:

“The future isn’t AI vs. humans. It’s humans with poor tools vs. humans with AI-augmented judgment.”


Who Should Listen?

  • Accounting firm partners evaluating AI investments
  • CFOs in mid-sized businesses
  • Advisors wanting to offer data-driven insights
  • Students & early-career accountants entering an AI-transformed profession

We put this into practice in EXL Cloud, our own SaaS modelling platform. It connects Xero to Excel and adds a library of Claude skills that build, check and report on driver-based financial models, so the AI works on a structured model instead of guessing.

Frequently asked questions

Where is generative AI on the hype cycle for a data-driven CFO?

The panel placed generative AI at the peak of inflated expectations in 2023 and 2024, sliding towards the trough of disillusionment by 2025. For a data-driven CFO, that means treating claims such as 80% time savings with caution: many fail under audit-quality requirements, and much business AI is rebranded rules-based automation.

What types of AI should accountants and CFOs understand?

The discussion covered four. Supervised learning learns from labelled history, such as predicting client payment delays. Unsupervised learning finds hidden patterns, such as client risk segments. Natural language processing reads text, such as bank statement descriptions. Anomaly detection flags outliers, such as potentially fraudulent journal entries. For a data-driven CFO, the aim is practical value rather than hype.

Why do AI projects in professional services fail?

The panel named data quality as the bottleneck, not the algorithm: garbage in, garbage out. Other failure modes were using generative AI for regulated outputs such as tax opinions, underestimating data preparation (around 70% of project time), treating AI as a black box, vendor lock-in with proprietary models and ignoring change management. A data-driven CFO starts with the business problem, not the tool.

How should a data-driven CFO start with AI?

Start small and measure rigorously: pilot one process, such as creditor reconciliations, and track time saved and error rates. Build internal AI literacy around bias, confidence and data lineage, partner with IT and data teams, focus AI on forward-looking advisory work, and ask vendors for proof of value on your own data. That is where a data-driven CFO adds the most value.

Will AI replace accountants?

The panel’s view was that AI will not replace accountants, but accountants who use it strategically will replace those who do not. The future it described is people with AI-augmented judgement, supported by clean data, clear governance and compliance with ethics and privacy obligations whenever client data is involved.

Need clearer reporting, dashboards or Power BI insights for your business? See our Finance Intelligence as a Service.


About the Author

Lance Rubin is CEO and Founder of Model Citizn, and founder and Chief Excel Officer of EXL Cloud, the SaaS platform that connects Xero to Excel with driver-based financial models, dashboards and Claude AI skills. He is a Chartered Accountant, FMI Certified Trainer, co-author of the CA ANZ Financial Modelling Study Guide, and creator of the AI-Powered Accountant programme on Maven, currently live at Level 1 with Level 2 in design, and author of the companion book “AI-Powered Accounting with Excel and Power BI” (Packt). He has worked at PwC, KPMG, Investec, and NAB, and is based in Melbourne, Australia.

Related reading: Finance Tech Stack Part 1 – Your Tech Stack Roadmap and Finance Tech Stack Part 2: Nine Criteria Every Finance Professional Should Know about Tech.

Share This

Related Articles

Craig is the founder of Beyond Excel (BXL) and 5g. BXL was founded in 2006 to teach readers and viewers how to pull data from enterprise systems into Excel using VBA.
Finance and Accounting teams are expected to take a more strategic seat given the high rates of uncertainty in 2021 in relation to profits, cash flow and general business stability.
If you want to learn more in 2019 and also understand how you can upskill in Financial Modeling, be sure to reach out to us or stay tuned.