Hello Reader,

Every day people are asking how AI can help them.

This inspired me to launch a series titled "How AI can help Finance".

I have had the chance to spend 2,000 hours working on AI for Finance over the last 18 months.

Every day, I meet with experts in the field and founders of new AI startups specialized in Finance which gives me a glimpse into the future of Finance.

Let me share with you where you can start and how AI can help you.

Today we start with Finance's most critical function: Accounting.

Without accounting the FP&A team has no financial data to analyze.
Without accounting, the Tax department cannot finalize the tax filings.
Without accounting, the CFO cannot report the figures to the shareholders.

This is why I wanted to start this series with accounting!

Please make sure to reply to this email by telling me which function you want to see me covering next in this series!

​
Nicolas

Free Masterclass

On March 19th I will discuss about the future of finance with my friend Christian Wattig, expert on FP&A (sponsored by DocuWare).

Here is what to expect:

• Learn which finance processes will be eliminated in the future of finance
• Understand the skills and tools finance teams need to succeed in the future of finance
• How to use the freed-up time for higher-value finance tasks
• How effective finance business partners drive growth and profitability
• More tips and tricks on how to use ChatGPT

​Join us for free here!​

How AI can help Accountants?

If AI can help accounting, the whole finance function as well as the whole company will benefit from it.

Main Challenges Faced by Accounting

Let's start by understand the main challenges accountants face.

Having managed a team of accountants in my previous job, here are, in my opinion, the top 3 challenges to focus on:

• Manual processing of invoices and receipts. Problem: prone to errors and inefficiencies.
• Time-consuming bank reconciliations due to high volume of transactions and different types of bank statements formats and internal reports.
• High amount of manual entry of journal entries, increasing the risk of inaccuracies but also time spent on low value activities.

1) AI Solutions for Invoice processing

Let's start by the invoicing. Where can AI help?

The solution: digitize and automate your Invoice Processing.

AI, particularly OCR technology, can automatically extract data from invoices and receipts.

Then it will use NLP (Natural Language Processing) to understand the information read from the invoice and contextualize it.

This step is new as before, accountants had to use rules-based software where they had to manually teach the software how to map each supplier's invoices the ERP fields.

Now the NLP technology allows an automatic mapping as it understands the context. For example, seeing "Invoice number" on the invoice will make the model map the charactes with the field "Invoice number" in the ERP.

Moreover, if a software provider utilizes collective intelligence, you can benefit from the patterns learned by other users. Meaning, if a number of clients have already validated the way the software encodes a FedEx invoice, you can be sure that the model has an accuracy close to 100% on FedEx invoices.

Which tool to use ?
Azure Cognitive Services for Invoice Processing

If you are already in the Microsoft Azure environment, you can use the Azure Form Recognizer module (see here for more info from Microsoft).

• How does it help: Automates the extraction of data from invoices and receipts, significantly reducing manual processing time and increasing accuracy.
• Practical Implementation: Accounts payable teams can integrate Azure Form Recognizer into their workflow to automatically process incoming invoices, extract relevant data (such as vendor name, invoice number, and amount), and populate this data into their financial systems without manual input using an API

​Here is a video showing how Form Recognizer works and here is a specific one on receipts.

I am not affiliated with Microsoft Azure but I like how you can use it without installing any other software if you already have Microsoft Azure.

Why? Because you can create an automation that gets invoices from Outlook or a folder, then process it through the OCR and NLP of the Form Recognizer. After that it will feed a flat table using the JSON language which is ideal.

JSON is one of the common language used for APIs (Application Programming Interface)

This API will be what enables you to book the invoice in your ERP based on the output of the Form Recognizer process.

Other alternatives?

If you don't have Microsoft Azure, you can try solutions that are off the shelf like: Nanonets, Parsio, Yokoy or Vic.ai.

In the coming months, I plan to meet with the founders of these companies and test their solutions. I will share the results of this work in the AI Finance Club. If you want to be at the cutting edge of AI in Finance and leverage my expertise, I recommend you to join the club by clicking here.

2) Automate your reconciliations

The second area where you can use AI to improve your work is reconciliations.

Here are two ways you can do it:

First option, build your own mini algorithm.

You might think that this is not for you. But without knowing it, you have already built hundreds of algorithms. Here are two real life examples:

• Cooking: When following a recipe, you're essentially following a set of instructions (an algorithm) that outlines the steps to prepare a dish. Each step, from chopping vegetables to adjusting cooking times, follows a specific sequence to achieve the desired outcome.
  ​
• Excel: When working with Excel, you often create algorithms through formulas and functions to perform calculations, analyze data, and automate tasks. For instance, creating a formula to calculate sales tax or using conditional formatting to highlight specific data points are examples of algorithmic processes in Excel.

So now that we have debunked the notion that you are capable of building an algorithm, how can you build your own for reconciliation?

Here are the 5 steps:

1. Get the data from your bank and your system using digital files (get CSV or Text files if you can).
2. Clean this data using rules (formatting of date, removing columns, combining files, enriching data).
3. Create your matching rules using algorithm
4. Analyze exceptions
5. Act on exceptions: either book the transaction in your system or ask the bank to correct (sometimes it happens that the bank make mistake)

For step 1 to 3 you can automate the process using a combination of Python and ChatGPT. Use ChatGPT to generate Python code that will automate the cleaning of data and create rules for reconciliations.

If you have access to a Large Language Model, you can add a step in your reconciliation where it reads the text to add context in the reconciliation and determine the possibility of applying fuzzy matching.

You might notice that I refer to Python here. Again, don't be afraid to use it. This opens the door to automations and integrates AI into your processes.

If you want to use my knowledge but especially the expertise of Christian Martinez (the best person combining FP&A, Data Analytics and AI knowledge), then you can come to our course on Advanced ChatGPT for Finance. In this live course, we teach how to combine ChatGPT and Python to automate Financial Analysis.

​

Second option: off-the-shelf solutions

I found some solutions that are automating reconciliations. I haven't test them yet, so this is to take with a pinch of salt.

Here are the ones you can explore: Docyt, SmartStream and Reconzile​

3) Automation of Journal Entries

If you have already worked on the first and second areas listed above, normally most of your journal entries are already automated.

But there are still areas where accounting has to book a lot of entries like:

• Revenue recognition
• Accruals
• Tax booking

For this, you need to have a system that automate these entries based on the contracts with your clients (for revenue recognition), your accruals rules and your tax specificities.

For this, I have found three innovative accounting systems that start ups are using and which are leveraging AI:

​Puzzle: Puzzle is the solution for SaaS companies who wants to automate their accounting. It uses integration with Stripe, Ramp, Mercury Bank, Gusto and Rippling to ease the booking and accelerate the closing.

​Truewind: This week i had the chance to talk to the founder, Alex Lee who showed me how they use AI to automate the bookkeeping of invoices, receipts and soon revenue recognition. The tool is ideal for SMEs or for fractional CFOs who want to help their clients.

​Booke.ai: Booked uses AI to categorize expenses but also to digitalize the accounts payable process. I have not seen the solution yet, but the fact that it covers several accounting processes is interesting and I am looking forward to test it and present it in exclusivity to the members of the AI Finance Club.

My perspective on these tools is independent as I have no affiliation with these tools. However, I recognize the efforts they are putting in bringing AI to accounting and they are right now the most relevant accounting softwares that are AI native.

Wrap up:

I have covered the three most important areas where AI can streamline accounting tasks.

Yet, there are many more areas to think about like travel and expenses review (done by appzen) or cash recovery of VAT (done by zapliance) or creating a ChatBot to reply to the most asked questions to the other departments (look for Chatbot powered by AI, there are plenty of solutions).

The potential for innovation is there and the opportunities for you are evident.

I spend 100s of hours every month researching the tools that will help Finance, meeting their founders and testing what are the best ways to use AI in finance processes.

I do it because I strongly believe we in finance can reap the benefits of using AI in our processes.

However, I am also conscious that most of you, including CFOs, finance managers, controllers, FP&A teams and accountants don't have the time to undertake all of this work by yourselves.

This is why you can have access to all of my work, along with insights from many experts by becoming a member of the AI Finance Club.

On top, all the members who join us get access to all of my courses on AI for Finance.

Consider joining us, this is the best place to learn how to use AI for Finance.

I am curious which finance role do you want me to cover in my next email of the series "How AI can help Finance". Let me know!

I hope you enjoyed this email as I spent a significant part of my Sunday on it, but I feel it was important to share with you this practical learnings.

Looking forward to reading your responses!

Nicolas

​

P.S.: Here is how I can help you:

1. Learn ChatGPT to boost your productivity and stay ahead of the competition. Take my 2-hour course.

2. Master PowerPoint and impress your boss with my PowerPoint guide for Finance professionals.

3. Do you want to move to FP&A? Learn all the technical and soft skills you need here.

4. Do You want your team to start reaping the benefits of ChatGPT? Hire me for a 2-hour digital workshop tailored to your needs.
Reply to this email if you are interested.