Hi Reader,

Last month, OpenAI dropped GPT-5.
And as usual, the internet was flooded with hype.

But if you’re a Finance leader, you don’t need hype.

You need clarity on:

• What’s actually new?
• Where does it fit into Finance workflows?
• And how do you train your team without wasting months experimenting?

That’s why I went through the full GPT-5 release notes and tested it myself.
Here’s the CFO-ready breakdown.

But before we dive deep into GPT-5, here's something important...

Free Masterclass: Use cases and AI roadmap (in German)

Learn how to become an AI-CFO in just one hour.
In this live masterclass, you’ll discover the AI roadmap and practical use cases that are already transforming Finance teams today.

You’ll see:

• Which Finance AI use cases bring the highest ROI
• Best practices that work in real Finance teams
• A concrete roadmap you can apply right away to future-proof your team

Note: (This session is in German — but English and French editions are coming soon.)

Who’s this masterclass for?

1. CFOs and Finance leaders ready to scale AI
2. Decision-makers looking for scalable AI tools in Finance
3. Finance executives who are still unsure of AI’s full potential, but are open to learn what works

👉 This free webinar is offered to you by Agicap.

GPT-5 has landed (here's what's new)

GPT-5 isn’t just “faster.” It’s a complete redesign focused on reasoning, reliability, and personalization.
​
For Finance leaders, that means AI you can finally trust for forecasts, reconciliations, and scenario planning, not just email drafts.

Here’s the CFO-ready breakdown of what’s new, how to access it, and why it matters.

Enjoy!

Before we start, you need to understand what a reasoning model is compared to the traditional LLM models that you have been using until now.

Traditional LLMs are trained mainly to predict the next word in a sequence, so they excel at generating fluent, human-like text but can sometimes produce answers that sound right without being logically consistent.

Reasoning models, on the other hand, are designed to explicitly follow logical steps, check consistency, and arrive at a conclusion in a structured way—more like solving a problem step by step than just predicting text.

In short: LLMs generate language, while reasoning models aim to generate logic-driven answers.

1. What’s Actually New in GPT-5

GPT-5 is the new default model in ChatGPT (replacing GPT-4o and others) that is designed to combine speed and deep reasoning.

• Lower hallucinations, higher reliability → fewer factual errors vs GPT-4o on web-enabled prompts.
• More efficient thinking → it matches or beats o3 while using fewer tokens.
• Enterprise variant → “GPT-5 Pro / Thinking Pro” exclusive for Team/Enterprise/Education extends reasoning cycles for highest reliability.
• Ecosystem integration → now integrated in Microsoft 365 Copilot and Copilot Studio (relevant if you use Excel/Outlook/Teams).

2. GPT-5 Mode Switching

GPT-5 isn’t just one model. It has a family of modes with a real-time router determining which model to use based on complexity.

• Auto → default mode, decides how long to think (with the help of routing).

You can still manually choose instead of relying on auto:

• Instant → speed over depth; great for summaries, quick checks.
• Thinking → slower, deeper reasoning; best for forecasts, reconciliations, policies.
• Pro → most reliable, nuanced; ideal for mission-critical analysis.

Even the free tier gives you 10 messages per 5 hours + 1 “Thinking” message/day.

3. How to Access GPT-5 “Thinking”

This is actually the most important part of this model.

Before, only 5% of the users were using the reasoning model. Remember the o3 model? Maybe not, this is probably because you were also part of the 95% not using (or not knowing about it). This is why now with ChatGPT 5, you have the model choosing for you if you need reasoning or not based on the complexity of your request.

But there are multiple ways to force the model into deeper reasoning:

• Select “GPT-5 Thinking” manually in the picker
• Use the “Think longer” button
• Add the phrase: “think hard about this” in your prompt
• Pro accounts → longest, deepest reasoning

For Finance: I recommend testing Thinking mode for reconciliations, stress-testing forecasts, and drafting sensitive documents where reasoning matters more than speed.

4. How You and Your Team Can Use GPT-5 Today

The release of GPT-5 doesn’t just make ChatGPT “smarter.” It changes how finance leaders interact with it day to day.

With this update, you can now:

1. Rely on adaptive intelligence → no more guessing which model to pick. GPT-5 automatically switches between modes.

2. Trust the outputs more → GPT-5 produces significantly fewer hallucinations compared to GPT-4 and o-series models.

3. Work with bigger, messier documents → extended context window (up to ~196k tokens in Thinking mode).

4. Control depth and style of reasoning → controls things like “Think longer” and verbosity settings.

5. Seamlessly embed into Microsoft 365 → if you are using Copilot Business in Microsoft 365, you can activate GPT-5 (button at the top right like in the image below).

For Finance: I recommend using GPT-5 as your adaptive tool. Use it for quick variance summaries, deeper scenario stress tests, and drafting board-ready narratives where accuracy and reasoning matter more than speed.

5. Where GPT-5 fits into Finance workflows

Some use cases where you can use GPT-5 in finance:

Some more use cases...

CFO → draft board-ready narratives with assumptions and Q&A points; or stress-test strategic scenarios (e.g., interest rate shifts, FX exposure).

FP&A → automate variance analyses and rolling forecast updates; or run scenario modeling with assumptions clearly documented.

Treasury → generate 13-week cash forecasts from ERP + bank feeds; or flag liquidity risks tied to late AR collections.

6. How to Train Your Finance Team on GPT-5 (Without Wasting Months)

If you're lost on what to do, you can start with the following:

Start with “Quick Wins” → pick 2–3 everyday tasks where time savings are obvious (e.g., drafting variance explanations, reconciling AR exceptions, summarizing AP issues).

Let the team see the benefit in days, not months.

Give Them Playbooks, Not Prompts → instead of “experiment and see,” give ready-to-use templates (use the prompts examples in section 5 above and then make your own).

Build “Show & Tell” Rituals → weekly 30-minute team sessions: one person shows how they used GPT-5 to cut time or improve clarity.

The rest of the team can adopt proven tactics immediately.

Have you already tried GPT-5 and what are your first impressions?
​
Let me hear your feedback and learn different use cases from you 😉

​

Best,
Your AI Finance Coach, Nicolas
​
​
​P.S. If you want to see my AI roadmap in action and how to develop in AI CFO in just 1 hour, join me at the Agicap webinar (German edition first — English & French coming soon).​
​ 👉 Reserve your free seat now​