And then everything changed...
Hi there,
It's been a while since I wrote to you. Life, work, the usual. But I'm writing now because something has shifted in how I think about financial modelling, and it felt important to share it with you directly.
When I started writing the Financial Modelling Handbook, my goal was clear: teach people the craft. How to structure a model. How to get good at the mechanical "how" of modelling.
I still believe that knowledge matters.
However, over the past year, I've watched AI tools go from interesting novelties to genuinely capable modelling assistants. Not in a vague, theoretical "maybe one day this will be useful" way, but in a practical, I-use-this-every-day way. And the pace of change is accelerating. The AI labs made a deliberate choice to focus at the start on making AI great at writing code. Financial models are, at their core, structured logic, and that means they sit squarely in the crosshairs of what AI is getting good at, fast.
So here's where I've landed: the "how" of financial modelling is changing. The new "how" of financial modelling is about deploying AI effectively in your modelling work; how to direct it, govern it, and combine it with your own professional judgement to produce better work, faster.
That's the direction I'm taking the Handbook from here, with a focus specifically on Project Finance modelling since that's the space I know and love.
What I'm seeing in practice
I want to be specific about what's actually possible right now, because the hype around AI can make it hard to separate signal from noise.
We're now at the point where AI tools can:
- Add complex project finance calculations to an existing model
- Shadow-model a section to independently check for errors
- Set up an existing model template for a brand new project
You can see demos of these workflows in a webinar I'm running on Tuesday (see below for more).
What's particularly interesting is the difference between generic and specific AI tools. Generic tools like Claude for Excel work from inside your spreadsheet. They read cells, write formulas, and handle formatting. They're incredibly flexible and work for virtually any Excel task. But then there are purpose-built tools like PF Nexus, which abstract the business logic away from the spreadsheet entirely. Instead of the AI wrestling with cell references, it operates directly on an abstracted, codified representation of the model's financial logic, drawing on libraries of project finance knowledge and automatically enforcing standards like FAST.
The combination is powerful. Generic tools are brilliant for data cleaning, ad-hoc formatting, and building one-off calculations. Specific tools excel at standards-enforced modelling, controlled versioning, complex model reasoning, and multi-step workflows.
What we can learn from software engineering
One thing that gives me confidence this isn't just hype: we can look at what's already happened in software engineering.
AI hasn't reduced the number of software jobs, but it has dramatically shifted what those jobs look like. Junior developers are now expected to work with AI pair-programming tools as standard. The demand for testing, auditing, and governance skills has grown steadily. Enterprise adoption has exploded - a recent Andreessen Horowitz survey found that 80% of Global 2000 companies are now comfortable using non-Microsoft AI tools.
I think we are maybe 18 months behind the software world, but I expect to see the same pattern in financial modelling. The modellers who thrive will be those who understand both finance and the AI toolkit and can direct these tools with the confidence that comes from deep domain knowledge.
Free webinar: AI in Project Finance Modelling
I'm running a free webinar to show you exactly what this looks like in practice. We'll cover:
- What financial modellers can learn from the software engineering world's experience with AI
- An introduction to PF Nexus and how purpose-built PF tools differ from generic AI assistants
- Live demos — adding calculations, shadow modelling for error checking, and setting up models for new projects
- Q&A and discussion on where this is all heading
The webinars are on this Tuesday, March 3rd, with two time options:
Session 1: 10 am UK / 2 pm Dubai | 6 pm Singapore | Register here
Session 2: 1 pm NYC / 6 pm London | Register here
Both sessions cover the same content. Pick whichever works for you.
New course: Project Finance Modelling with AI
For those who want to go deeper, I'm also launching a new 6-week online programme with the Project Finance Institute: Project Finance Modelling with AI.
This is the first training programme designed specifically to apply AI to the complexities of project finance and PPP modelling. It's built for experienced project finance modellers who want to incorporate AI into their toolkit. As ChatGPT would put it: this is not a general overview – it's hands-on, professional-level training.
Here's what we'll cover across the six weekly modules:
Module 1: Agent workflow foundations: how to work with AI agents effectively. Clear prompting, structured planning, context management. You'll model revenue and opex from a real case study using both Claude for Excel and PF Nexus.
Module 2: Skills and modularisation: extracting reusable modelling components, building organisation-specific skill libraries, and deploying them into new work. You'll build a class-wide skills library with your cohort.
Module 3: Building with project documents: taking a contract or report, planning its implementation in a model, deploying tailor-fit calculations, and using AI to verify logic and produce documentation.
Module 4: Automated model standardisation: taking a non-FAST model and rebuilding it to FAST standard using AI, with programmatic parity verification using Python. The advanced material extends into Monte Carlo analysis.
Module 5: AI for model audit and risk management: using shadow modelling and contract-to-model skills to independently audit complex mechanisms, find errors, and validate correctness.
Module 6: Governing AI-driven change: Git and source control for AI-driven changes, review workflows, and automating model update memos.
The six week, online course starts on March 27th and costs USD 850.
Apply to join here. (We'll ask you about your PF modelling skills as this course is designed for experienced PF modellers).
This is where I'm putting my energy. I genuinely believe the modellers who learn to work with these tools now, thoughtfully, with proper governance and deep domain knowledge, will have an enormous advantage. And I want to help as many of you get there as possible.
Thanks for reading. And thanks for sticking with the Handbook through the quiet patch. There's a lot more coming.
Kenny
P.S. - I'm going to migrate this website to Substack in the next week or so - so the next time you'll hear from me will be from there.
Comments
Sign in or become a Financial Modelling Handbook member to join the conversation.
Just enter your email below to get a log in link.