Hey there!

Welcome to Briefed, your go-to source for the latest Accounting x AI news. In today’s post, we cover Chaser’s new letters, the Big 4 implementing AI, IRS headlines, GPT-4’s performance as a financial analyst, and the potential for AI Agents to transform the industry.

Let’s dive right in…

[Read time: 5 minutes]

📰 Accountech news

1. Chaser is sending letters in the post to your clients

But wait, isn’t this newsletter meant to be all about tech and AI?! Well, yes, but I guess it doesn’t hurt to be reminded that some things are better done in person. The accounts receivables automation tool Chaser has just launched a feature that enables SMEs to send letters in the post to debtors automatically with a QR code that can be used to make the payment online. Apparently, 90% of postal letters get opened while only 20% of emails do, so they might be onto something.

Back in the world of AI, Chaser did announce some features at the start of the year around late payment predictions; recommended chasing times; and payer ratings, which are “using artificial intelligence”, but to be honest, I can’t imagine these being much more than simple regression models.

2. PwC and KPMG are implementing AI… but there’s a difference

PwC is entering the AI world with a clear focus on business growth. “Our UK and US firms have signed an agreement with OpenAI, making PwC OpenAI’s first reseller for ChatGPT Enterprise and the largest user of the product.” They say the move is set to “streamline operations” internally, though I’m sure they’re equally excited by the opportunity to drive profits by being an implementation partner for ChatGPT Enterprise.

On the flip side, KPMG is taking a more altruistic approach with their AI Impact Initiative 2024. They’re focusing on how AI can benefit non-profits and drive positive social change. This initiative plans to implement AI in various ways to support non-profits, such as optimising resource allocation, enhancing data analysis, and improving service delivery to underserved communities. So, while PwC is using AI to supercharge its business, KPMG is harnessing AI to support non-profit organisations and make a meaningful impact in society.

3. The IRS is making headlines: AI, direct filing, and budget battles

Across the pond, AI is stepping in to help Uncle Sam close the tax gap. The Government Accountability Office reports that AI models are already being used to improve audits and estimate the tax gap. The IRS is leveraging these models to catch errors and reduce the difference between taxes owed and taxes paid. You can read more on the US Government Accountability Office website.

In other news, the IRS announced that it’s making its Direct File platform a permanent fixture. This move is expected to make tax filing simpler and cheaper for millions, perhaps at the expense of existing platforms like TurboTax. But not everyone is thrilled about these changes. House Republicans are proposing to defund the IRS’s Direct File initiative, which could impact the IRS’s ability to implement these new tools and services. So, while the IRS is pushing forward with AI and making tax filing more accessible, the Republicans are trying to hold them back.

4. GPT-4 outperforms professional human analysts at predicting earnings changes

Here’s one for the financial analysts. A recent study by researchers at the University of Chicago Booth School of Business explored whether large language models (LLMs) like GPT-4 can effectively perform financial statement analysis, a task traditionally handled by human analysts. The results are impressive: even without narrative or industry-specific information, GPT-4 outperformed human analysts in predicting the direction of future earnings. The AI model showed a relative advantage, particularly in complex situations where analysts typically struggle.

Interestingly, the study also found that GPT-4’s performance was on par with specialised machine learning models, highlighting its capability to generate valuable insights about a company’s future performance from purely numerical data because the LLM is able to generate “useful narrative insights”. Moreover, trading strategies based on GPT-4’s predictions yielded higher returns compared to those based on other models.

🤖 AI debrief

What is an AI Agent?

So you’ve probably heard the term ‘AI Agent’ thrown around loosely over the past few months, and you’d be forgiven for assuming it’s not all that different from ChatGPT. However, while the two are built on the same technology, the automation potential differs significantly.

AI agents are advanced AI systems designed to interact with their environment autonomously. Unlike traditional AI platforms, such as ChatGPT, which rely heavily on user prompting, AI agents utilise a comprehensive ecosystem of infrastructure and tools. This ecosystem includes access to external knowledge bases (acting as long-term memory), evaluation capabilities (self-reflection on task performance), and various tools like internet browsing and email inbox extraction.

For example, an AI agent can remember user corrections over time, ensuring continuous improvement and accuracy in repetitive tasks. This long-term memory function is crucial for personalised and context-specific workflows. AI agents also have the capability to evaluate their performance. They can assess how well they performed a task, such as evaluating the accuracy of data extraction. If an AI agent detects low confidence in its output, it can notify a human for review, ensuring high standards and reducing errors.

Additionally, AI agents can leverage the internet to retrieve up-to-date information or obtain additional context from other data sources, such as an email inbox. This ensures that no critical information is overlooked, enhancing the completeness and accuracy of processes.

What does this mean for accountants?

Imagine this: an AI that learns from your corrections during bookkeeping tasks. Over time, this AI becomes increasingly adept at handling client-specific details, reducing the need for manual interventions. It can evaluate its own performance and alert you when something needs a closer look, ensuring that nothing slips through the cracks.

For example, suppose an accountant is managing invoices and receives an email with an attachment. An AI agent can analyse the attachment, read the email content, and cross-reference this information with existing records. If there are discrepancies or additional details needed, the AI can flag these issues for human review.

The implementation of AI agents in accounting promises a significant shift in efficiency and accuracy. By leveraging their ability to learn, self-evaluate, and access comprehensive information, AI agents can take on more complex and nuanced tasks, freeing up accountants to focus on strategic decision-making and client advisory roles.

That’s all for today! Subscribers will receive new posts directly in their inbox every other Wednesday. In the meantime, feedback is always welcome—hit reply and let me know what you think.

Until next time 🫡

Reuben