What's Next For Accounting? - Issue #64

2023 AI Prediction Roundup for Public Accounting

Here's a link to a write-up I did. Common traps I'm seeing people fall into when considering AI 👇

It's like a confident intern that doesn't know anythingThink of it as a *free* intern that never sleeps, is free, doesn't get sick, never forgets what you tell it, and is free. Focus on a single use-case for ChatGPT, develop an amazing prompt to train it to do exactly what you want it to do, and put it work. See my intern Steve if you need ideas.

It's a long way from ever doing X as well as I doFor most tasks, AI makes for a great assistant, not a replacement. Consider how it can help you, not how it can replace you.

Sweet I can generate a ton of blog posts with thisLong-term, assume AI-generated text will be recognizable. Who knows if future models will be, but current models can be recognized, so assume search algorithms will penalize your GPT-3 content, and future browsers will have built-in detection.

It's not readyOur minds jump to the most complex work we do, when in reality we're pushing our noggins to the limit 1% of the time. Start simple, with the stuff you'd delegate to a VA or jr staff member.

It can't do meaningful work because it can't see into my systemsAs I've shared the best version of this is one that's embedded into the systems you use every day. For example, email response generation embedded within your practice management system can see projects, files, tasks, engagements, past emails, even meeting transcripts.

The discussion of what AI can't do is entirely separate from the discussion of what AI can do. Focusing on the latter is more productive.

I always come back to this Paul Graham quote: "The future is not hard to see. You've already seen it. You just didn't realize it, because you dismissed it as a toy."

🤖 Automate My Thing

Reusable Data Transformations

You know that service that gives you a CSV, and you have to do a bunch of data reformatting to get it to a format for import into another app? Of course you do we're accountants that's our job.

Here's a spin on this concept, to create a reusable data transformation in ChatGPT:

App A exports a CSV in this format:

  • Date (DD/MM/YYYY)

  • Description

  • Useless ID

  • Debit

  • Credit

  • Card Number (full number)

App B imports a CSV in this format:

  • Date (MM/DD/YY)

  • Reference

  • Amount (positive or negative in a single field)

  • Card Number (last 4)

This ended up being way more powerful than I thought it'd be. Here's the full build guide I posted to twitter, that among other things, converts data from format A to B.

    💎This Week's Sponsor

    👷‍♀️ This Week I Made

    😍From The Community

    My community Realize is a private space for people to swap ideas & collaborate. At $180/mo it’s also the way I finance everything else I do, pay a full-time creative director, and a part-time developer in an effort to make the accounting industry a slightly better place 💗 The community is currently about 270 members strong.