On 30th November 2022, OpenAI released an early version of ChatGPT – and it went viral. Over a million users signed up in just five days.
At first, people used it for everything from writing short stories to planning holidays. But I had a strong feeling it wouldn’t be long before AI started reshaping the world of finance.
I wasn’t wrong.
Fast forward to today, and finance teams are already using ChatGPT to automate repetitive tasks, streamline reporting and improve forecasting. And with the latest advancements in GPT-4o and beyond, the possibilities have expanded even further.
Here’s how you can make ChatGPT an essential part of your team’s toolkit.
Personalise AI with CustomGPTs
One of the biggest AI breakthroughs is CustomGPTs – something I covered in a recent newsletter (which you can sign up to now if you want monthly tips about AI and all things finance!).
By default, ChatGPT is a generalist. It can answer broad finance questions, it doesn’t understand your team’s specific needs. That’s where CustomGPTs come in.
A CustomGPT is a version of ChatGPT you can train and personalise. Instead of using the same chatbot as everyone else, you can create an AI tool that understands your workflows, challenges and priorities.
For finance teams, this means no more generic answers or repeating yourself in every new chat. Your CustomGPT can remember preferred reporting formats, align with compliance requirements and even match your tone of voice – no more correcting American spellings!
How can finance teams use CustomGPTs?
Here are just a few ways finance teams are already using CustomGPTs:
- Training and onboarding: Upload essential training resources so new hires and junior team members can access key information instantly.
- Financial reporting: Train a CustomGPT on historical reports and compliance standards to generate accurate financial summaries.
- Audits and compliance: A GPT trained on regulatory requirements can flag compliance risks and offer guidance.
- Regulatory changes: Tax teams can stay ahead by training a GPT on HMRC guidelines to get instant summaries of the latest updates.
A quick heads-up: confidential data shouldn’t be uploaded unless you’re on the Enterprise plan. But even without that, CustomGPTs can significantly speed up finance processes.
And the best part? The more you use it, the better it gets at understanding your needs.
A quick-start guide to CustomGPTs
Setting up a CustomGPT is easy – you don’t need any coding skills, just a clear idea of what you want it to do.
A ChatGPT Plus subscription is required but for $20/month, it’s well worth it.
Here’s how to get started:
Step 1: Create your CustomGPT
Go to ChatGPT, click “Explore GPTs” and select “Create a GPT”.
Step 2: Define Its Role
You’ll now be in the ‘create’ section. Here, you’ll be able to give your GPT a name and description.
Step 3: Train it with Instructions
Add detailed guidelines for how it should behave. The more context you provide, the better the output. Here’s an example:
“Explain financial concepts step-by-step, offer examples and suggest exercises for learning. Provide concise definitions for accounting terms and deliver actionable insights for market analysis scenarios.”
If you’re unsure how to phrase instructions, ask ChatGPT for help – describe what you want to the general assistant, and it can generate instructions for you.
Step 4: Upload Example Data
If you have publicly shareable templates, reports or training materials, upload them. This is what your GPT will use to refine its responses.
Step 5: Add Prompt Suggestions
In the preview section, add prompts to guide the GPTs training. For example:
- Tell me how to build a discounted cash flow model.
- Explain IFRS 15 with examples.
- What key factors are affecting market volatility this quarter?
Step 6: Test and Refine
Interact with your GPT to test its responses. If something isn’t quite right, tweak the instructions until you get the output you need.
Once you’re happy with it, save and share it with your team!
Supercharge Research with ChatGPT Deep Research
If you’re tired of spending hours searching through financial reports or regulatory updates, this next feature might be worth exploring.
Deep Research takes ChatGPT beyond simple search results. Instead of just giving snippets of information, it does structured, multi-step research, evaluates sources, cross-checks facts and generates insights that would normally take you hours or even days.
What sets Deep Research apart?
While AI-powered searches find relevant information more quickly than going through Google, they often lack structure and credibility. Deep Research solves that.
- Autonomous multi-step research: Deep Research thinks like an analyst – it plans search queries, evaluates sources and produces structured reports.
- Longer reasoning time: Queries take between 5 to 30 minutes to complete, ensuring well-researched, fact-checked results.
- Transparent citations: Do you remember being told to ‘show your thinking’ at school? Deep Research does just that. Every response includes a breakdown of sources and reasoning, so you can easily verify information.
How can finance teams use Deep Research?
The best way to demonstrate this is with an example. So, let’s say you’re analysing a potential acquisition target.
Instead of manually reviewing company reports, investor statements and analyst opinions, you could ask Deep Research:
“Provide a structured report on [Company X], including its latest earnings performance, recent leadership changes, and any major regulatory risks affecting its industry.”
Within minutes, you’ll have a complete summary of financial performance, key risks and strategic insights – all with cited sources so you can verify the data.
How reliable is it?
If you’re anything like me, you probably want some data to back this up. OpenAI recently tested Deep Research on Humanity’s Last Exam, a benchmark designed to test AI on 3000 expert-level, multidisciplinary questions.
For context, GPT-4o scored 3.3%, meaning it could only answer a fraction of the specialist questions. Deep Research, on the other hand, scored 26.6% – a huge leap in AI-driven research capabilities.
That said, AI is still AI. It’s an awesome tool for speeding up research, but you’ll still want to double-check sources before making critical decisions using the information it gives you.
Automate Routines with ChatGPT Tasks
Do you ever wish that your to-do list could manage itself? ChatGPT Tasks makes that possible.
ChatGPT Tasks allows you to schedule AI-driven actions in advance, to be completed on a routine basis or as a one-off request. Although it’s still being tested, I think the feature will be a game changer for finance teams.
All you need to do is enter your prompt, set your schedule and let ChatGPT do the rest. You’ll get a notification each time it completes your scheduled task.
With this, finance teams will be able to fully automate routine processes like:
- Daily market update summaries delivered every morning.
- Scheduled reminders for reporting deadlines.
- Monthly performance summaries based on financial data.
AI-Generated Visuals for Finance Presentations
Designing engaging visuals for finance presentations can take hours, and let’s be honest, not every finance pro wants to spend their time in design tools. The good news? AI image generation can do the heavy lifting for you.
With tools like DALL·E (built into ChatGPT), you can now create professional, on-brand visuals and infographics in seconds. All you need is a clear description of what you’re looking for. Let’s say you’re presenting a report on rising operational costs.
Instead of using a generic stock image or overloaded Excel chart, you could ask ChatGPT:
“Create a clean, modern infographic showing the top 5 drivers of rising operational costs in UK retail over the last 12 months. Use icon-style visuals and bold, easy-to-read labels.”
You’ll instantly get a custom graphic you can drop straight into your slides. No fiddling with alignment, no clunky clipart, just smart, effective visuals that help you tell a story.
The key is to be specific with your prompt. Think about the purpose of the visual, who it’s for, and what message you want it to support. You don’t need to be a designer, just describe it like you would to a creative team.
Ready to Make the Most of ChatGPT?
I’d love to hear how you’re using these tools – let me know if you’ve had any wins with them!
And if you’re looking for more advice on introducing AI into your finance workflows, I’m happy to help. Keep an eye out for my upcoming AI in finance courses too – details coming soon.
