Train your own finance AI - ChatGPT vs Poe, How AI can save you 18 hour days, AI Slack sidekick

Aug 31, 2023

1-2-3 Newsletter - Read Time - 5.5 Mins

Hello, and welcome to a another edition of the Tech for Finance 1-2-3 Newsletter (30 secs longer, sorry) where we have:

1. Tutorial - How to train your own finance AI - ChatGPT vs Poe

2. Expert resources - How AI can save finance teams from 18 hour work days + DISC framework for building a great finance team

3. Tech tools - Deploy your own chat bot, AI powered document writing + Slack Team Q & A, AI insights

1/ How to train your own finance AI - ChatGPT vs Poe

There's been some buzz recently now that you can give custom instructions to ChatGPT.

In my Ultimate GPT Framework for Finance, the first 2 elements of the SPICES prompt framework are (S) Scenario and (P) Profile.

These to elements allow you to provide context to the prompt, and a manes of the AI adopting a role / skill.

In theory, you can now cut out these first 2 steps using 'Custom Instructions'.

In this tutorial, we'll be looking at 2 tools where you can add custom instructions. ChatGPT, and Poe.

1/ ChatGPT

First, lets give it a poor prompt, without setting up a custom instruction.

Prompt - "Generate me a budget table for my departments"


Now let's set up a custom instruction and see how the same prompt works.

Step 1 - Custom Instruction

Go into your ChatGPT settings, and select 'Custom Instructions.'


Step 2 - Provide context

You'll then see 2 fields where you can provide more information about yourself and your responsibilities, as well as the way you'd like ChatGPT to respond.


Step 3 - Start a new prompt

Here's the response we now get after inputting our custom instructions using the same prompt as before - "Generate me a budget table for my departments"


This took me all of 2 minutes, and is infinitely better than the previous response that was given.

Give it a try, and shoot me a reply, I'd love to see how you're using this!

2/ Poe

You can use Poe for free here.

This can be used with multiple large language models (ChatGPT, Llama 2, and Claude)

The advantage of Poe, is that you can create multiple chat bots, with different custom instructions for different purposes.

It also has the added benefit of a community that openly shares their custom bots too. So there's a tonne you can start using straight away without having to build your own.

Step 1 - Create a new bot


Then you'll need to complete the necessary fields.

You won't see the whole prompt so here it is:


"You are an Excel genius.

You have infinite knowledge of every Excel formula and each use case.

You area of expertise is corporate finance, and you've experience with using Excel to create 3 way models, the creation of budget matrices, forecast.

When prompted, you are to use your knowledge to provide the best possible formulas, or potentially even VBA code to achieve the ends requested.

When responding, you must break down the rationale behind the formula / code and provide instructions on how to implement it.

Please ensure that before responding you clarify any missing information that's likely to affect the quality of your output."


I've set my model to Llama 2 as it's the next best thing compared to GPT-4.

And I've set my temperature to the middle to get the right balance between creativity and consistency.


Here's the response to simply typing:

"Budget vs Actual Variance Analysis"


Mind blown 🤯

My thoughts

IMO ChatGPT is still the best in terms of output, but you're limited to one 'Custom Instructon'

With Poe, the models might not be as good, but there's infinitely more flexibility in creating AIs for different purposes.

Hit reply and let me know which one you prefer.

2/ Expert Resources

1. This week we released my podcast with Ashok Manthena, Founder of ChatFin and AI + Finance thought leader.

In our discussion we cover:

- Data Quality for Predictive Finance Analytics
- Challenges in AI Contextual Understanding
- Improving Language Models With Context
- Generative AI in Finance User Interfaces
- Evolution of AI in Finance Analysis
- The future of AI in Finance

And much much more.

Check out the full podcast "Closing Time: How AI Can Save Finance Teams from 18-Hour Days" on audio here. And on video here.

2. Oana Labes recently posted her guide on how to build a great finance team using the DISC assessment to improve teamwork, communication, and productivity:

You can find it here.

Make sure to follow her, and sign up for her 'Finance Gem' newsletter too!

3/ Tech Tools

1. Mightychat - Train and deploy your own chat bot in less than 5 minutes here.

2. Calepin - AI powered assistant for professional document writing here.

3. AI Sidekick - Turn your team’s daily Q&A into actionable insights in Slack here.

IMPORTANT - Always read the data policy for any tools you try. If you’re ever unsure. Make sure your data, and any customer or supplier data is anonymised.

NOTE - I am not affiliated with any of the tools in this e-mail.

So that’s it for this week.

Until Next Time


👨‍💻Connect with me on LinkedIn

🎧Subscribe to the Podcast

📺Watch on YouTube

Subscribe to Framework Friday

Join 1,100+ finance pros, and receive 1 actionable tech framework you can use to win back your time, develop your skills, and stay ahead of the game.

Subscribe for FREE with your name and e-mail:

We hate SPAM. We will never sell your information, for any reason.