Happy 2025! It’s been a minute since I last sent out a project. I've been working on some projects; before I knew it, it was the end of February. That’s my bad. The good news is that I implemented various automated processes in all my projects, which I’ll share here in the coming weeks!

I have an issue (if you couldn’t tell). I love building things. That does mean I can get distracted in the short term and lose focus on various projects. So, to make myself feel better, I’m going to pass on this issue to you as well. Today, we’re building a Streamlit app to analyze a competitor’s website, produce a SWOT analysis, and then give you some startup ideas to build!

Remember: You can get a step-by-step tutorial on this project using our AI tutor, PrompterGPT. Simply choose which project you’d like to work on, follow the instructions, and tell it what you want changed!

Build a Project

Application Steps:

1. Drop in the URL of a website you’d like to analyze

   

2. Press “Analyze”. The app will use Firecrawl to crawl the website, and use OpenAI’s API to generate the SWOT and startup ideas.

   

3. Conquer the world

App Demo:

Tech Stack:

1. Firecrawl

2. OpenAI’s GPT-4o-mini

3. Streamlit

Sounds good? Let’s get to work. Click the button below to start building. PrompterGPT will walk you through the project step-by-step and help you troubleshoot, customize, or edit its scope. Just ask away!

Build SWOT Analysis

Happy building, and see you next time.

— Nishad from Prompter

P.P.S for more daily thoughts on AI, and sneak peak of projects before they’re sent, make sure to follow us on X!

Happy Holidays! 🎄

YouTube hosts long-form podcasts, announcements, and other primarily audio-driven announcements. However, unless you pay for premium, you can’t close the app. This isn't very pleasant. I’d love to listen to some podcasts on the go, but I don’t have enough time to watch a three-hour video. I really only want to get the summary and ask questions.

Today, we’re building a Chat with Youtube app using Streamlit, pytubefix, and OpenAI.

It’s pretty simple, and the app has less than 100 lines of code! All we need to do is:

1. Drop in the YouTube URL you’d like to process

2. Click “Process”

3. Ask questions!

Remember: You can get a step-by-step tutorial on this project using our AI tutor, PrompterGPT. Simply choose which project you’d like to work on, follow the instructions, and tell it what you want changed!

Build a Project

Application Steps:

1. Drop in a URL of a YouTube video that you would like to process.

   

2. Click “Process Video”. We will extract the audio from the YouTube video and create a transcription using OpenAI’s Whisper API.

3. Ask Questions! Ask any question about the video and receive and answer directly in Streamlit.

App Demo:

Tech Stack:

1. pytubefix

2. OpenAI’s Whisper

3. OpenAI’s GPT-4o-mini

4. Streamlit

Technical Workflow:

Sounds good? Let’s get to work. Click the button below to start building. PrompterGPT will walk you through the project step-by-step and help you troubleshoot, customize, or edit its scope. Just ask away!

Build Video Chat Tool

Happy building, and see you next time.

— Nishad from Prompter

P.S. This project doesn’t interest you? Check out our previous project on how to build your own Scrum Master Check-in Tool

P.P.S for more daily thoughts on AI, and sneak peak of projects before they’re sent, make sure to follow us on X!

The daily standup. We’ve all been a part of one. A quick morning meeting to let your peers know what you’re building, what you are blocked by, and when you will get that 3-week-old Jira ticket done.

In my opinion, it serves two purposes:

1. To actually inform and give a status update (best-case scenario)

2. An accountability forum to publicly ensure that you’re carrying your weight.

I think we can accomplish these 2 purposes without meeting for 15-30 minutes daily (not including the 15 minutes before and after as people need to mentally transition from meeting move to build mode or vice versa).

Today, we will build an automation workflow to help team leads synthesize information and tasks, saving time and money spent on daily standups.

Remember: You can get a step-by-step tutorial on this project using our AI tutor, PrompterGPT. Simply choose which project you’d like to work on, follow the instructions, and tell it what you want changed!

Build Project

Application Steps:

1. Set up a database of your team members' email addresses. This will track who needs to participate in the daily standup.

2. A morning email gets sent to your team, reminding them to complete their Daily Check-In Form.

3. A bit later, a follow-up email is sent to people who haven’t completed the form.

4. Your team members will individually fill out a form - highlighting the essential questions of the day about the status of their work. (And they answer with natural language!)

5. Once everyone has filled in their responses, OpenAI summarizes them and sends them in an email to your team lead.

You’ll receive a concise email download of your team’s goals, accomplishments, and issues daily.

Tech Stack:

1. Gmail

2. Google Sheets

3. Google Forms

4. Google App Script

5. OpenAI’s GPT-4o-mini

Technical Workflow:

Sounds good? Let’s get to work. Click the button below to start building. PrompterGPT will walk you through the project step-by-step and help you troubleshoot, customize, or edit its scope. Just ask away!

Build AI Daily Stand Up

Happy building, and see you next time.

— Nishad from Prompter

P.S. This project doesn’t interest you? Check out last week’s project on how to build your own AI Executive Assistant!

P.P.S for more daily thoughts on AI, and sneak peak of projects before they’re sent, make sure to follow us on X!

Hey everyone!

Back with another project!

Today, we’ll focus on taking a step out of your daily processes. Sometimes, without you noticing, some of the most minor, mundane tasks you do are actually the best candidates for automation. It just takes a moment for you to step back and really think about what you do.

Let’s focus on one specific aspect of your day-to-day life: scheduling meetings. Most of your work is done in your email inbox, and you usually click over to Google Calendar to schedule a meeting.

Let’s try to build an automation workflow that allows you to stay within your email and schedule all your meetings from gmail.com.

Remember: You can get a step-by-step tutorial on this project using our AI tutor, PrompterGPT. Simply choose which project you’d like to work on, follow the instructions, and tell it what you want changed!

Build Project

App Demo

Application Steps:

1. Forward an email to yourself with the event details to schedule

2. App schedules your email

(yeah it’s that simple!)

Tech Stack:

1. Gmail

2. Google Calendar

3. OpenAI’s GPT-4o mini

4. Google App Script

Technical Workflow:

Sounds good? Let’s get to work. Click the button below to start building. PrompterGPT will walk you through the project step-by-step and help you troubleshoot, customize, or edit its scope. Just ask away!

Build AI Event Creator

Happy building, and see you next time.

— Nishad from Prompter

P.S. This project doesn’t interest you? Check out last week’s project on how to build your own Automated Lead Scorer!

P.P.S for more daily thoughts on AI, and sneak peak of projects before they’re sent, make sure to follow us on X!

Hey everyone!

Back with another automation project! This time, we’re ditching Python and embedding OpenAI directly into the Google Ecosystem using App Script (sorry Gemini, we will try you next time!)

For this project, we will be focusing on lead scoring.

If you’re in an established service business, you probably have a lead scoring system going already. If you don’t, a lead scoring system is set up due to a pretty simple reason: it’s pretty ineffective for you to take calls with every single inbound prospect, so you need a way to score them quickly and accurately.

Having a form for prospects to fill out is great, but manually reviewing them can be painful, and creating a fixed-reponse selection form doesn’t really provide enough insight on what the potential client actually wants.

So, let’s build an AI business analyst to score them autonomously.

Remember: You can get a step-by-step tutorial on this project using our AI tutor, PrompterGPT. Simply choose which project you’d like to work on, follow the instructions, and tell it what you want changed!

Build Project

Application Steps:

1. Link a Google Form to your website, a prospect will fill out, answering some basic questions.

2. In the code (will be shared shortly) define your ideal customer - along with timelines, product type, marketing goals, etc. This is what the prospect’s form responses will be scored against.

3. When responses are submitted, in an new column, a score with reasoning will be populated.

   

App Demo:

Tech Stack:

1. Google Sheets

2. Google Forms

3. OpenAI’s GPT-4o mini

4. Google App Script

Technical Workflow:

Sounds good? Let’s get to work. Click the button below to start building. PrompterGPT will walk you through the project step-by-step and help you troubleshoot, customize, or edit its scope. Just ask away!

Build Leader Scorer

Happy building, and see you next time.

— Nishad from Prompter

P.S. This project doesn’t interest you? Check out last week’s project on how to build your own Competitor Website Monitor!

P.P.S for more daily thoughts on AI, and sneak peak of projects before they’re sent, make sure to follow us on X!

Hey Prompter Community!

It’s a dog-eat-dog world out there. With this new technological revolution with LLMs, these dogs are building complex features on tighter budgets with smaller teams in a shorter time. You best stay on your A-game. It’s vital to keep an eye on your competitor, but ensure not to get tormented by their product announcements and their seeming progress.

As Sam Altman says:

A quick word about competitors: competitors are a startup ghost story. First-time founders think they are what kill 99% of startups. But 99% of startups die from suicide, not murder. Worry instead about all of your internal problems. If you fail, it will very likely be because you failed to make a great product and/or failed to make a great company.

- Y Combinator Startup Playbook

So, how do we keep tabs on our competitors without wasting precious time focusing on our internal issues?

Let’s automate it.

Remember: You can get a step-by-step tutorial on this project using our AI tutor, PrompterGPT. Simply choose which project you’d like to work on, follow the instructions, and tell it what you want changed!

Build a Project

We will build a relatively simple Flask app that allows you to upload various “competitor URLs,” whether it’s their newsroom or a site like macrumors.com. When new content is published on these sites, we will scrape, summarize, and email them to you once a week.

Application Steps:

1. Create a “Competitor Profile”

   

2. Add URLs of websites

3. Schedule Run on Google Scheduler

3. Receive email with a summary of changes 1 time a week (or your choice)

App Demo:

Tech Stack:

1. OpenAI’s GPT-4o Mini

2. Google Cloud Run

3. Flask

4. Google Scheduler

5. Beautiful Soup

6. SMTP Email Sending

Technical Workflow:

Sounds good? Let’s get to work. Click the button below to start building. PrompterGPT will walk you through the project step-by-step and help you troubleshoot, customize, or edit its scope. Just ask away!

Build Competitor Monitor

Happy building, and see you next time.

— Nishad from Prompter

P.S. This project doesn’t interest you? Check out last week’s project on how to build your own Applicant Tracking System!

P.P.S for more daily thoughts on AI, and sneak peak of projects before they’re sent, make sure to follow us on X!

Hey Prompter Community!

Let’s face it: the state of the job market is a weird AI trench warfare battle. AI recruiters are scanning millions of resumes, looking for buzzwords from resumes blasted out by people using AI to apply for hundreds of jobs.

Chances are, if you’re using a social media channel to hire, you’re getting hundreds of applications you can’t parse through. It’s just the state of the market.

So, instead of paying hundreds of dollars a month to some ATS software company, let’s build one on our own in half an hour - for free!

Let’s look at this week’s project: a DIY Applicant Tracking and Rating system.

Want to build today’s project but don’t know where to start? Get a step-by-step walkthrough by PrompterGPT - our AI tutor tuned on our projects and ready to help - no matter your skill level.

Build this Project

Application Steps:

1. Admin uploads job and description

2. Admin sends Streamlit link out to applicants.

   ---

3. Applicants insert their name and email and attach their resume.

   

   ---

4. OpenAI compares a resume to the job description, gives a summary and score, and inserts data into a sheet.

   

   ---

5. The resumes are saved to a folder.

App Demo

Tech Stack

1. Streamlit

2. Google Cloud Storage

3. Google Sheets API

4. Google Drive API

5. OpenAI’s GPT 4o-mini

Sounds good? Let’s get to work. Click the button below to start building. PrompterGPT will walk you through the project step-by-step and help you troubleshoot, customize, or edit its scope. Just ask away!

Build ATS

P.S. This project doesn’t interest you? Check out last week’s project for a free time tracking app template!

P.P.S for more daily thoughts on AI, and sneak peak of projects before they’re sent, make sure to follow us on X!

What happens to product roadmaps when the marginal cost of writing a line of code goes to zero?

It’s an exciting thought. We’ve stumbled upon an age where a simple addition to an app that could have taken weeks takes mere seconds. What took hundreds of expensive developer hours can now take mere minutes of a non-technical conversation back and forth. So what if we shifted those minutes to the customer?

Should we give customers control over their experience? No-code tools such as Monday.com, Bubble.io, and others attempt to have this sort of end-user flexibility with apparent success, but another obvious factor is the limitations of no-code.

To be clear, I was pro no-code, until I realized you can code with no code like a pro using code (gen).

So, let’s take an example of what the future could look like. Today’s project is less of an automation project and more about building an app in 2024. I’ll show you how I share my projects with PrompterGPT and how you can build and customize anything (and I’ll give out a free template for a time-tracking app you can build, customize, and deploy now!).

Step 1: Using PrompterGPT, get today’s time-tracking app project

I upload each of my automation projects to a custom GPT so you can easily build, deploy, customize, and troubleshoot.

Get Time Tracking App

Step 2: Follow the instructions and copy the provided code to a new working directory.

Step 3: Assess the current app’s capabilities

Follow PrompterGPT’s instructions to test the app locally and see what you like/dislike about it (my feelings won’t be hurt, don’t worry!)

Step 4: Customize and implement with PrompterGPT

I’ll be adding a scheduled email send feature to send information to my leadership group

There we go—simple as that. I’m not sure about you, but this really feels like the future of internal tooling SaaS. I get a basic template and the flexibility to change what I need. The customer has the power of creation, while the provider sets up the architectures and frameworks.

Next week, I’ll be back with another automation project, so stay tuned!

Happy building, and see you next time.

— Nishad from Prompter

P.S. This project doesn’t interest you? Check out last week’s project on how to build a media monitor to capitalize on breaking news about your company.

P.P.S for more daily thoughts on AI, and sneak peak of projects before they’re sent, make sure to follow us on X!

Hey Prompter Community!

Today, we’re building a simple application to help you monitor earned media mentions. Whether you’re part of a large organization and want to monitor a product launch closely, or you’re a growing startup just beginning to get some press, this lightweight app will ensure you’re maximizing your coverage without spending more than 1 cent a week.

As always, if you’re interested in building this project, start with a step-by-step walkthrough from our custom GPT, PrompterGPT! PrompterGPT is trained on all of our projects, so you can tune, customize, or troubleshoot - all for free! Just mention which project you’re working on.

Build Media Monitor

Let’s get started:

Application Steps:

1. User opens app

2. User inserts company name to track

3. One time a week (which can change), the app uses News API to gather 25 new articles

4. OpenAI determines the relevance of each article.

5. The article link, summary, and publish date are appended in Google Sheets if relevant.

App Demo

Tech Stack

1. OpenAI

2. Google Cloud (Run, Scheduler, Sheets API)

3. News API

4. Flask / Python

Technical Workflow

Sounds good? Let’s get to work. Click the button below to start building. PrompterGPT will walk you through the project step-by-step and help you troubleshoot, customize, or edit its scope. Just ask away!

Build Media Monitor

Happy building, and see you next time.

— Nishad from Prompter

P.S. This project doesn’t interest you? Check out last week’s project on how to clean up your Google Drive with just a click!

P.P.S for more daily thoughts on AI, and sneak peak of projects before they’re sent, make sure to follow us on X!

If you’re like me, your Google Drive is a mess. I try my best to keep it clean, but among the important files and documents, there is a hodgepodge of “Untitled Documents”, “project_v34_final_final”, and random PDFs with god-awful names.

Honestly, some days I can’t find the right file for the life of me, and I just know in the back of my mind that my earlier procrastination on giving it a proper name and organization schema is biting me in the ass.

So, let’s build something to reward procrastinators worldwide and automate cleaning up your file’s names.

As always, if this project interests you and you’d like to build it yourself, you can have a step-by-step walkthrough with Prompter’s custom GPT (and even ask it to tune the project to your needs).

Build Drive Organizer

Here’s how it’ll work:

Application Steps:

1. User opens Streamlit.

2. User authorizes access to their Google Drive account.

3. The user copies the file link into the app.

4. App processes files (not other folders within the folder) and extracts the first 1000 characters.

5. Based on the extraction, the app determines whether it is appropriately named and either suggests a new name or confirms that it is appropriately named.

6. The user selects which file names they would like changed.

7. App changes the file’s name.

App Demo

Tech Stack

1. OpenAI’s GPT-4o-mini.

2. Google Cloud Vision API (for reading images, non-selectable PDF’s).

3. Google Drive API.

4. Google Cloud.

5. Streamlit.

Technical Workflow

Sounds good? Let’s get to work. Click the button below to start building. PrompterGPT will walk you through the project step-by-step and help you troubleshoot, customize, or edit its scope. Just ask away!

Build Drive Organizer

Happy building, and see you next time.

— Nishad from Prompter

P.S. This project doesn’t interest you? Check out last week’s project on how to automate the extraction of new SEC filings to keep you informed about your portfolio!

P.P.S for more daily thoughts on AI, and sneak peak of projects before they’re sent, make sure to follow us on X!

Hey Prompter Community!

Frankly, I’m tired of circumnavigating around paywalls to get the latest news about my stock holdings.

So, let’s use Beautiful Soup to extract text from SEC filings and OpenAI to summarize them so we can receive breaking news about our portfolios straight to our email inboxes.

Here’s how the app will work:

Application Steps

1. User will select companies to track.

2. App performs a lookup to identify CIK codes to track SEC filings.

3. When the process is triggered, the CIK code is searched, and new filings are gathered. (Process can be triggered manually or with a cron job via Google Scheduler, ex: 1 time a week)

4. App processes and cleans up extracted text.

5. Text is sent to OpenAI.

6. OpenAI summarizes text.

7. SMTP email is sent with summaries of company’s most recent filings.

8. Most recent filing is added to a JSON to ensure no duplicate filing fetches.

App Demo

(note: if Google Cloud Scheduler is set up, this will automatically send new filings to my inbox each week.)

Tech Stack

1. OpenAI’s GPT-4o mini

2. Google Cloud Run

3. Google Cloud Scheduler

4. Flask / Python

5. SEC.gov API

6. Beautiful Soup

Technical Workflow

Sounds good? Let’s get to work. Click the button below to start building. PrompterGPT will walk you through the project step-by-step and help you troubleshoot, customize, or edit its scope. Just ask away!

Build SEC Tracker

Happy building, and see you next time.

— Nishad from Prompter

P.S. This project doesn’t interest you? Check out last week’s project on building a personalized finance coach!

P.P.S for more daily thoughts on AI, and sneak peak of projects before they’re sent, make sure to follow us on X!

Hey, Prompter community!

Happy to be back with a new project. This one is less of a workflow automation and more of a cool useful app you can set up in minutes!

Chances are you have some financial goals. Tediously parsing through your credit card statement or using the relatively unhelpful built-in visualizations in your bank’s app leaves you wanting a more accessible and personalized way to set financial goals month-to-month.

Today, we’re going to build a small Streamlit app (and by the end of this, you should know why I’m obsessed with this framework for internal tools) that allows you to set financial goals using natural language and upload your credit card statement to see if those goals have been met.

Let’s dive in:

Application Steps:

1. The user creates a goal using natural language (e.g., “I want to keep my going out-to-eat expenses under $200 this month”).

2. OpenAI’s API extracts key details and creates a success metric.

3. The user uploads a credit card statement.

4. Goals are measured as complete or not complete.

5. Results saved to JSON.

App Demo:

Tech Stack:

1. Streamlit

2. Python

3. OpenAI’s GPT-4o Mini

Technical Workflow

Sounds good? Let’s get to work. Click the button below to start building. PrompterGPT will walk you through the project step-by-step and help you troubleshoot, customize, or edit its scope. Just ask away!

Build AI Finance Coach

Happy building, and see you next time.

— Nishad from Prompter

P.S. This project doesn’t interest you? Check out last week’s project on building an AI email assistant that auto-sorts your emails!

P.P.S for more daily thoughts on AI, and sneak peak of projects before they’re sent, make sure to follow us on X!

Hey Prompter Community!

Happy post-Labor Day! I hope you got some well-deserved rest. I have good news and bad news. The good news is that it’s a short week. The bad news is that you probably have a ton of emails.

So that’s why I’m excited to bring you our latest project: Gmail Auto-Sorting.

This week, we will deviate from workflow process automation and focus on something else: leveraging LLMs as a point solution for our emails.

Generally, for automation solutions, there are a few categories:

1. Workflow Automation

   1. Workflows are multi-step guardrail automations. Think of these as a factory line. You’ve developed a process, solidified it, optimized it, and made it more efficient (without applying automation). Now, you’ve implemented some robotic process automation or other methods to have a smooth-running machine.

2. Point Solutions

   1. These are single-point automation solutions. Point solutions are slightly more complex as individuals and focus on optimizing a single step within a process.

3. Agents / General Automation

   1. This is a growing category that will explode over the next few years (contingent on LLMs improving in a few weak areas). These solutions are multi-faced and responsible for automating an area rather than a workflow. They are responsible for making their own decisions on how to complete a process.

So, today, we will look at how to build a point solution that can make your life a little more organized.

If you’re like me, you’ve had the same email address for a while - and it’s full of important emails, payment info, and subjects I’m interested in inside a sea of spam, marketing emails, and more crap. My inbox is a mess.

Let’s fix this.

We’re going to create several labels with descriptions in a web app. Our app will look at our new emails each hour and apply a label based on the contents via an assessment from OpenAI’s GPT-4o mini to organize our inbox.

Application Steps

1. Create labels and descriptions for what kind of emails should be categorized in each label.

2. Google cron job triggers each hour

3. Application cycles through new emails and extracts body text.

4. The body text is sent to OpenAI along with a list of labels.

5. OpenAI returns the best-matched label

6. Label is applied to email.

7. If there is no match, the label is not applied.

App Demo:

Tech Stack:

1. Google Cloud Run

2. Gmail API

3. Google Cloud Scheduler

4. Flask / Python

5. OpenAI’s GPT-4o mini

Technical Workflow:

Sounds good? Let’s get to work. Click the button below to start building. PrompterGPT will walk you through the project and help you troubleshoot, customize, or edit its scope. Just ask away!

Important note: To build this app we use Gmail’s historyID - meaning that the first time you run this application, no new emails will be processed, as it grabs the first ID as an anchor.

Build Project

Happy building, and see you next time.

— Nishad from Prompter

P.S. This project doesn’t interest you? Check out last week’s project on building a Reddit bot that finds your product’s target audience!

P.P.S for more daily thoughts on AI, and sneak peak of projects before they’re sent, make sure to follow us on X!

Hey Prompter Community!

I’m excited to bring you our latest project: Automated Reddit Discovery. This project is focused on leveraging Reddit as an outlet for discovering and engaging with your target audience, capitalizing on customer discovery, participating in product promotion, finding beta testers, and more.

I believe Reddit is a unique platform that will grow massively over the next few years. While it pales compared to some of the more prominent social platforms, it has a few highly unique characteristics that make it an incredible platform for startups and businesses to find dedicated users and customers.

1. Self-governance—Reddit’s subreddits have moderators, meaning that the culture and vibe of each subreddit vary drastically. This is important because most of these moderators will not allow excessive self-promotion if it provides no value. So, you must get creative and focus on delivering value to the conversation first and foremost. Uncreative, spammy Reddit users fail here.

2. High-quality content - Reddit’s posts and conversations are higher quality and usually less spammy. These conversations provide such value that Google ranks them pretty high in search results.

If you can thoughtfully respond to posts and creatively promote your product before anyone else (see this article by Matt McGary on how to do this right), you can dominate a social platform many fail to conquer.

You can find a treasure trove of potential users by adding value to conversations, answering questions, and understanding when, where and most importantly HOW to plug your solution. However, constantly combing these subreddits for potential opportunities is not in your best interest and takes a lot of time.

A note: You get the highest engagement within the first 18 hours, the standard lifecycle of a Reddit post.

But here’s the thing—how do we find a Reddit post similar to what we are selling where we can add value and see our users, all within 16 hours?

Let’s automate it.

Today, we will build a Reddit post similarity search bot that will deliver the most relevant posts to our product to our inbox, allowing us to leave highly visible comments first.

Application Steps:

1. Add subreddit(s) to track.

2. Create a detailed product description.

3. Add logical reasoning rules.

4. Then, the app runs at preset times.

5. Recent posts are pulled from tracked subreddits.

6. A similarity search is done on each post.

7. Logical reasoning rules are further applied to filter down posts.

8. Final posts are emailed to your inbox with a link for you to comment.

App Demo:

Tech Stack:

1. Google Cloud Run

2. Google Cloud Scheduler

3. Flask / Python

4. OpenAI’s Embeddings and GPT-4o mini models

5. Reddit API

Technical Workflow:

Sounds good? Let’s get to work. Click the button below to start building. PrompterGPT will walk you through the project (number 8) and help you troubleshoot, customize, or edit its scope. Just ask away!

Build Project

Happy building, and see you next time.

— Nishad from Prompter

P.S. This project doesn’t interest you? Check out last week’s project on building an AI accountant who lives in your inbox and manages your receipts for you!

P.P.S for more daily thoughts on AI, and sneak peak of projects before they’re sent, make sure to follow us on X!

Hello Prompter Community!

Seeing this community grow has been thrilling, and feedback from each project week after week has been extremely valuable and welcomed! I’m excited to bring you the final installment on how to automate your accounting operations.

As a look-back, the last two accounting series sends were focused on leveraging OpenAI’s new affordable model, GPT-4o mini, and testing its ability to parse through unstructured data like invoices and credit card statements. We’re going to continue that, but focus on how to integrate this workflow into your day-to-day operations.

But first, let’s take a look back on what we covered in the previous installments:

Part 1 - Bulk invoice processing

First, we examined automating the extraction of text from a bulk group of invoices and slotting them into a Google Sheet.

Part 2 - Reciept Reconciliation and Expense Verification

Then, we built a workflow to ensure that there were matching receipts for recorded expenses in our budget tracker.

Now, let’s take this one step further to wrap up our accounting deep dive:

Part 3 - The Virtual Accountant

Today, we are going to build a workflow that lets users email a receipt or invoice to a predetermined email address, then the program will parse and slot the data right into your expense sheet. Never type an expense line by line ever again!

Let’s see it step by step.

1. Email - send an email with an attached invoice to a predetermined email address.

2. Parse - At the end of the day (or whatever time you set), the application will comb through your inbox and look for unread emails with the subject “EXPENSE”. Then it will extract and parse the text using GPT4o-mini.

3. Record - Finally, it will slot the expense into your budget sheet (feel free to adapt to your configuration!) and mark the email as read.

Let’s take a look at what tech stack we are using and the technical flow:

Tech Stack

1. Google Cloud (Specifically, Cloud Functions, and Cloud Scheduler)

2. Google Sheets

3. OpenAI GPT-4o mini

4. Gmail

Technical Workflow:

Here’s how the technical workflow pans out:

Sounds good? Let’s get to work. Click the button below to start building. PrompterGPT will walk you through the project (number 7) and help you troubleshoot, customize, or edit its scope. Just ask away!

Build Project

Happy building, and see you next time.

— Nishad from Prompter

P.S. for more daily thoughts on AI, and sneak peak of projects before they’re sent, make sure to follow us on X!

Let’s be honest, as a small business owner, you kick the can down the road on over a hundred tasks, until it bites you in the ass. This is fine! It’s just prioritization, but sometimes that tin can you kick hits a curb and pops you right back in the face. Let’s prevent that hypothetical bruise with automation.

Automation can significantly help you and ensure that critical, non-prioritized tasks get done without wasting your time. One perfect example of this is receipt reconciliation. If you’re moving fast to grow or build your business, chances are you are purchasing more and giving more autonomy to others to buy goods and services. However, cash is tight; it’s vital to ensure you’re paying for the right goods and that you can illustrate your expenses correctly, especially when tax season comes. Invoice and receipt reconciliation is essential, and according to Stripe, having a “smoothly functioning invoice reconciliation process reduces the risk of fraud and contributes to an organized, transparent financial system.”

In my last article, I detailed an automation flow using receipts as the source of truth to build an expense report. Today, we will use an expense report to understand what receipts we have and which we are missing based on our expense records.

Here’s how our application will work:

1. The user drops an Excel file full of expense records into the application.

2. The user points the application to a folder full of receipts.

3. The application parses each receipt and attempts to match the expense to the receipt.

4. Then, a downloadable report showing whether each expense has a matching receipt and the receipt's file name is produced.

Here’s a video of the automation in action, using an example of 5 expense records with only four matching receipts in a folder.

App: Automated Receipt Reconciliation

Expenses Input:

Automated Process:

Matched Expenses Output:

It takes just a few seconds to reconcile those receipts! Now, you can move on to better things, knowing your finances have been audited.

Tech Stack

Here’s the tech stack we’re using:

1. OpenAI’s GPT-4o Mini (its low cost makes this type of parsing affordable!).

2. Google Drive API

3. Google Cloud

4. Google Sheets/Excel (we are dropping an Excel sheet into the app, but you can customize it to add dynamic integration with GSheets!)

5. Streamlit (how your application will be hosted on the web (for free!)

Technical Workflow

Here’s how the technical workflow pans out:

Sounds good? Let’s get to work. Click the button below to start building. PrompterGPT will walk you through the project and help you troubleshoot, customize, or edit its scope. Just ask away!

Start Project

Happy building, and see you next time for part 3!

— Nishad from Prompter

P.S. for more daily thoughts on AI, and sneak peak of projects before they’re sent, make sure to follow us on X!

I’m thrilled to bring you the first part of a three-part series on financial automation for small business owners. We will examine a few labor-intensive tasks that will provide an immediate ROI on your time and allow you to get back to growing your business.

The first flow we will examine is Spend & Budget Management. This tedious and manual category has spawned some huge companies, including Expensify (peak $4B val), Concur (acq, $8.3B), and Rydoo ($30m ARR).

Chances are, you’re either holding out before you need to buy expense management software, or you have, and you’re just about to realize you spent too much on it ;).

Managing your expenses can be difficult, especially when you’ve just set up shop or recently hired a few employees and need to track reimbursements. Manual receipt reconciliation or invoice spend data entry is tedious work, yet millions still do it! Let’s change that with this project.

Let’s look at the manual flow

1. Gather invoices, statements, and bills

2. Open each PDF

3. Insert Vendor, Amount, Date, and Source into an expense tracker

4. Move the recorded file into a new folder to mark as complete

5. Repeat 2-5 until completed

Remember, Prompter caters to any business owner; you DON'T need to be technical to build this project; just take it slow and step by step, but pull the trigger and take a shot! Here’s what the program will do:

Now, let automate it

We will be automating more than 70% of the process, just leaving you to gather the invoices into one accessible area!

The tech stack we are using today is as follows:

1. OpenAI’s GPT-4o

2. Google Sheets API

3. Google Drive API

4. Google Cloud API

5. Streamlit (optional)

What does it look like?

And here’s a quick video of the automation in action:

Pretty cool, huh? I’ve adopted this into my weekly workflow and easily have saved hours!

If you’re ready to build this, hit the button below, and PrompterGPT will walk you through it at your pace, your skill level, and your own time. Just ask PrompterGPT to customize or troubleshoot whenever you need!

Start Project

Happy building, and see you next time for part 2!

— Nishad from Prompter

P.S. for more daily thoughts on AI, and sneak peak of projects, make sure to follow us on X!

Subscribe now

Share

Hello!

It’s been a bit! It's a little longer than I’d prefer to do in the future, but regardless - I'm excited to share the next project. This one typically falls outside of most common workflows shared by most knowledge workers and bleeds into helping you in your personal life - but regardless - I think this has a lot of value.

Today, we will explore content automation - and how to do it correctly. We’ve all seen a boatload of AI-generated images and videos and probably even have our suspicions about text from various sources. I think they lack 2 important things:

Soul & Effort.

This means we need to automate the content generate process elegantly. Let’s consider the task of selling some clothes or items on Facebook marketplace. Maybe you’re doing some spring cleaning, maybe you’re trying to sell some things out of your office. Regardless, using AI to generate captions of your items can be robotic and without soul, especially if they are not in your voice or missing context to your brand.

Platforms like Depop (which we will focus on today) are as much about the people selling the items as the items themselves. That’s why a consistent, friendly, self-tuned voice can help move inventory and save you time.

So, since we can’t automate taking pictures of the items we are selling, I think we can generate friendly captions without needing to upload pictures and images to ChatGPT Plus and constantly feed it context.

Let’s see how an application like this could work.

Subscribe now

Scenario:

We are selling clothes. A LOT of clothes. Filling in the description of what you’re selling can take a lot of time & effort, but we still want to sound like ourselves. How can we do this?

Here’s an example function that leverages Gmail, OpenAI’s GPT-4o, and Google Cloud to email a picture of the item you want to sell to a pre-determined email address and receive a custom caption in your voice in seconds.

Let’s take a look at real-world examples!

We want to caption a green shirt for sale. Using the image of the shirt, I attach it to an email and send it off to a predetermined email address that processes my emails to look for a specific phrase in the subject. At this point, it’s “CAPTION”.

A few moments later, I received an email back in the voice that I had pre-set up the application to talk in, including specific text that I wanted to include. This is the result!

Perfect! Now I copy and paste this into my listing, and voila! The caption is generated.

This project can be challenging for novices, so I hired some help. PrompterGPT is a great way to get a guided walkthrough through building all of our projects. Fine-tuned with knowledge from our projects, PrompterGPT can help you customize, troubleshoot, or edit the project however you like. To build this project, just hit the button below and let PrompterGPT know which project you’d like to work on (this one is called “Merchandise Your Inventory Faster”.

Start Project

Also, big shoutout to the Depop store this was built for, Maggie’s Closet! You can check out all of their styles at the link.

What’s next?

Over the next few weeks and months, I’m excited to highlight a three-part series on automating accounting for small businesses and projects and how to make a stack that keeps you from doing repetitive, mundane tasks and refocuses your valuable attention towards more important tasks.

Stay tuned, and as always, if you liked this article, consider subscribing and sharing!

I’ve also launched a referral program, trading my time to automate tasks at your business for subscribers. Check it out!

Cheers,

The Prompter Team

Subscribe now

The future is here, the future is now, and the future is weird. We are entering an ugly era, transitioning from a capped era of content production to an uncapped era. Just a few years ago, the fastest we could generate text-based content was capped at how fast a person could type. Now, the lid is off the pot. Essays, tweets, emails, articles (even newsletters!) are generated within seconds, seriously devaluing the utility of a piece of content.

This new reality has divided the camp into two distinct groups—one all-in on automated content and the other turned off by any automation. This false dilemma is not the way to proceed. Identifying where automation (and especially LLMs) can be the most beneficial is critical.

Today, we will examine cold emails and automate the outreach workflow. This kind of marketing tactic is a numbers game. With a naturally low conversion rate, you want to send as many emails as possible without draining resources. If you have a simple CRM or target list in Google Sheets, let’s see how we can automate a tedious process and help you become more efficient.

For this example, we’ll pretend we’re a marketing automation agency that finds clients based on their previous campaigns, and pitches to automate their next one.

Here’s what our automation flow will look like:

At the end of this project, we should see personalized emails in our drafts in Gmail—ready to send as marketing outreach.

The tech stack we will use is as follows:

1. Google Cloud

2. Google Sheets

3. VSCode (or the IDE of your choice)

4. OpenAI

5. Gmail

Make sure these are downloaded and/or set up before attempting this script! Before we start, here’s some exciting news to help you with this project:

Prompter has its own GPT!

I’m thinking about what’s next for the future of long-form interactive content. A core weakness of tutoring or educational content is the lack of back-and-forth between “teacher” and “student.” PrompterGPT aims to fix that. All articles and projects have been pre-uploaded to PrompterGPT’s memory, allowing you to troubleshoot errors in the code, customize the project to fit your needs, or even learn more about a particular module or application.

I’m incredibly excited to see where this goes! You can check it out here:

Check out Prompter GPT

If you’re struggling with this article or any project, check out PrompterGPT for help. As always, feel free to reach out if you still need help!

Let’s dive in:

Section 1: CRM Set Up

First, let’s set up our CRM. We will use a simple one with just four fields: our prospect’s name, business, email, and a short note about their last marketing campaign. This last field is how we will create a more personalized email. Your CRM should look like this:

Make sure you create this sheet using the same Google account from which you will send emails.

Section 2: Environment Set Up

First, make sure you have Python and VSCode set up. If you have never done this, here’s a short and easy tutorial link. Set up a folder for this project and open it in your VSCode environment.

The next thing we will do is set up Google Cloud. This allows us to interact with our applications, such as Google Drive, Google Sheets, and Gmail.

Here’s how to set up the project in Google Cloud:

1. Visit Google Cloud Console:

   • Open your web browser and go to the Google Cloud Console.

   • Sign in with your Google account if you're not already signed in.

2. Create a New Project:

   • In the Google Cloud Console, on the top toolbar, click on the project dropdown (next to "Google Cloud Platform").

   • Click on “New Project” in the top right corner of the dropdown.

   • Give your project a name, for example, AutoEmail.

   • Click "Create".

Step 2: Enable APIs

1. Enable Gmail, Sheets, and Drive APIs:

   • Once your project is created, ensure it is selected in the top toolbar.

   • Navigate to “APIs & Services” using the left sidebar, then click on “Dashboard”.

   • Click on "+ ENABLE APIS AND SERVICES" at the top of the page.

   • Search for “Gmail API”, click on it, and then click "Enable". Repeat this step for “Google Sheets API” and “Google Drive API”.

Step 3: Configure OAuth Consent Screen

1. Configure the OAuth Consent Screen:

   • Go back to the “APIs & Services” dashboard.

   • On the left menu, click on "OAuth consent screen".

   • Select "External" for the user type if you plan to use applications outside of your organization. Click "Create".

   • Enter the app name, user support email, and developer contact information.

   • IMPORTANT: Add your email address under "Test users" to be able to use this application.

   • Click "Save and Continue" until you complete the setup.

Step 4: Create Credentials

1. Create OAuth 2.0 Credentials:

   • In the “APIs & Services > Credentials” section, click on “Create Credentials” at the top.

   • Select “OAuth client ID”.

   • Application type should be “Desktop app” for this script.

   • Give your client ID a name, for example, Desktop Client for Email Automation.

   • Click "Create".

   • After the credentials are created, click "Download JSON" on the OAuth 2.0 Client IDs section. This downloads your credentials file.

   • Rename the downloaded file to client_secret.json and save it to your project folder.

Step 5: Prepare for Python Code

1. Move the client_secret.json to Your Project Folder:

   • Ensure the client_secret.json file is in the same folder as your Python script. This file contains sensitive information and should be kept secure.

Section 3: Adding Python Code

Now it’s time to add your Python code to a new file in the same directory as your .json file.

First, let’s start with the required packages.

In the terminal part of your VSCode environment, type the following in and hit enter to install the needed packages:

pip install --upgrade google-api-python-client google-auth-httplib2 google-auth-oauthlib gspread openai

Next, create a new Python file called “main.py” and add the following code:

import os
import pickle
import gspread
from google.auth.transport.requests import Request
from google.oauth2.credentials import Credentials
from google_auth_oauthlib.flow import InstalledAppFlow
from openai import OpenAI

client = OpenAI(api_key='ENTER YOUR OPENAI KEY')
from googleapiclient.discovery import build
import base64
from email.mime.text import MIMEText
import logging

# Setup logging
logging.basicConfig(level=logging.DEBUG, format='%(asctime)s - %(levelname)s - %(message)s')

# Define the OAuth2 scopes
SCOPES = [
    "https://www.googleapis.com/auth/spreadsheets",
    "https://www.googleapis.com/auth/gmail.compose",
    "https://www.googleapis.com/auth/gmail.modify",
    "https://www.googleapis.com/auth/gmail.send",
    "https://www.googleapis.com/auth/drive"
]

# Function to authenticate and create the Sheets and Gmail API client
def authenticate_google_apis():
    creds = None
    if os.path.exists('token.pickle'):
        with open('token.pickle', 'rb') as token:
            creds = pickle.load(token)
            logging.debug("Loaded credentials from token.pickle.")
    if not creds or not creds.valid:
        if creds and creds.expired and creds.refresh_token:
            creds.refresh(Request())
            logging.debug("Refreshed credentials.")
        else:
            flow = InstalledAppFlow.from_client_secrets_file('client_secret.json', SCOPES)
            creds = flow.run_local_server(port=0)
            logging.debug("New credentials obtained.")
        with open('token.pickle', 'wb') as token:
            pickle.dump(creds, token)
            logging.debug("Credentials saved to token.pickle.")
    return creds

def get_google_sheets_data(creds, sheet_name):
    try:
        gc = gspread.authorize(creds)
        sheet = gc.open(sheet_name).sheet1
        return sheet.get_all_records()
    except Exception as e:
        logging.error(f"Error accessing Google Sheets: {e}")
        raise



def generate_email_content(name, business, campaign_details):
    try:
        completion = client.chat.completions.create(
        model="gpt-4o",
        messages=[
            {f"role": "system", "content": f"""
    You are writing a cold email to solicit marketing services by my company Acme Marketing. We use automations to enhance marketing campaigns.
             Write an email to {name} for their business {business} and reference their most recent marketing campaign: {campaign_details}. Share how we can improve their costs through automation.
               Sign it off with Nishad P., Acme Marketing, my email npasricha@acmemarketing, and my phone number (123) 456 7890 Write it professionally. Do not include a subject line.
    """},
        ])
        return completion.choices[0].message.content
    except Exception as e:
        logging.error(f"Failed to generate email content: {e}")
        raise

def authenticate_gmail(creds):
    try:
        service = build('gmail', 'v1', credentials=creds)
        return service
    except Exception as e:
        logging.error(f"Error accessing Gmail API: {e}")
        raise

def create_draft(service, user_id, message_body):
    try:
        message = {'message': message_body}
        draft = service.users().drafts().create(userId=user_id, body=message).execute()
        logging.info(f"Draft created: {draft['id']}")
        return draft
    except Exception as e:
        logging.error(f"Error creating draft: {e}")
        raise

def main():
    creds = authenticate_google_apis()
    sheet_name = "Prompter Example CRM"
    data = get_google_sheets_data(creds, sheet_name)

    service = authenticate_gmail(creds)

    for record in data:
        try:
            email_content = generate_email_content(record['Name'], record['Business'], record['Most Recent Campaign Details'])
            message = MIMEText(email_content)
            message['to'] = record['Email']
            message['subject'] = 'Follow-up on your recent marketing campaign'
            raw = base64.urlsafe_b64encode(message.as_bytes()).decode()
            message_body = {'raw': raw}
            create_draft(service, 'me', message_body)
        except Exception as e:
            logging.error(f"Failed to process record for {record['Email']}: {e}")

if __name__ == "__main__":
    main()

Your environment should look roughly like this:

Next, we need to customize just three aspects of this code before we are ready to run:

1. Your OpenAI Key

2. Your Prompt

3. Your Google Sheet Name

Starting with your OpenAI API key (I teach you how to get that here), find the block of code near the top that looks like this:

client = OpenAI(api_key="ENTER YOUR OPENAI KEY")

Now, insert your key between the quotes. Remember to keep your key a secret; that is how your account is billed. (Each call only costs a fraction of a cent)

Next, let’s edit your prompt that generates the email content. Find this block:

completion = client.chat.completions.create(
        model="gpt-4o",
        messages=[
            {f"role": "system", "content": f"""
    You are writing a cold email to solicit marketing services by my company Acme Marketing. We use automations to enhance marketing campaigns.
             Write an email to {name} for their business {business} and reference their most recent marketing campaign: {campaign_details}. Share how we can improve their costs through automation.
               Sign it off with Nishad P., Acme Marketing, my email npasricha@acmemarketing, and my phone number (123) 456 7890 Write it professionally. Do not include a subject line.
    """},
        ])

Between the two triple quotation marks is where you’ll want to edit the kind of email you are writing. The variables in the { } brackets pull in each record’s value from your CRM. This allows ChatGPT to reference small bits of data and not have to type each record’s name individually. (Note: We are not including a subject line, as this code just creates the body of the email.)

Finally, the last thing to do is edit the name of the Google Sheet this program is finding. Ensure that this line matches the name of your sheet exactly:

sheet_name = "Prompter Example CRM"

Section 4: Run our code!

Click the play button on the top left to run your code. Depending on how many records are in your CRM, each record will take around 3-5 seconds to generate a response for and create a draft in Gmail.

If this is your first time running the application, you will be taken to a new browser window and asked to sign into Google. Ensure that you are signing in with the same account you are permitted to as a “test user,” or else you will hit an error.

After the program finishes (new text will stop generating in the Terminal box in VSCode), check your Gmail! In the drafts inbox, you should see fresh outreach emails ready to send to your prospects.

That’s it! Your emails are ready to be sent. Make sure to customize the prompt as little or as much as you want to shorten the length, tune the content, or more.

Thanks for tuning in to this project, I believe this is a great way to automate a necessary task that requires a lot of effort for such a little success rate.

If you’re having issues, check out PrompterGPT for troubleshooting. Just make sure to identify which project you’re working on!

If you liked this project or want to learn more about how you can use AI in your career, subscribe below or send this to a friend!

Subscribe now

Cheers,

- Nishad from Prompter

The ongoing joke (but not really) in the industry is that many B2B SaaS products are just fancy versions of Excel. The future might see a shift away from these visually appealing but complex interfaces towards using AI for back-office tasks through simple APIs, potentially reducing the need for many subscriptions.

Today, we'll concentrate on a particularly generic but crucial area of automation: expense software. The uniformity in accounting standards has made expense management a prime target for early automation efforts.

However, managing expenses remains cumbersome until it's possible to upgrade to a more comprehensive system. I’ll guide you through replicating an expense management tool, focusing on how to reconcile receipts with expenses after a project or event. This is especially useful if you're organizing an event or managing a project and don't have access to sophisticated company credit card software, making receipt management cumbersome.

The final system will look like this:

Subscribe now

I’m going to break this project into three sections:

1. Environment Set-Up

2. Development

3. Execution

1. Environment Set Up

Our stack is going to look similar to last time - using just three tools:

1. Google Colab

2. OpenAI’s API

3. Google Drive

Let’s set up a new environment with Google Colab. You can create a new notebook here: https://colab.research.google.com/.

Name your notebook, and before we write any of our main code, we will need to install some packages. In the first block, copy and paste this code and hit the play button to the left of the code:

!pip install PyPDF2 openai

Next, let’s set up and obtain an OpenAI API key. I’ve gone over this before, and if you’ve been following my previous articles, I outline how to get one. But as a refresher (you'll probably need a new key here anyways), here’s a refresher article.

Finally, let’s set up our Google Drive environment. We will read a folder of PDF receipts directly from Google Drive and return a compiled Excel file to the Drive. So, to make things easy, let’s set up a folder as close to your home directory as possible with the name “Expenses.” Inside the folder, let’s create another folder called “Receipts.”

The official path to the receipts folder would be: “/content/drive/My Drive//ExpenseApp/Receipts”

2. Development

This part is an easy copy and paste with a few edits, but essentially, the steps we are coding serve these sequential functions:

1. Extract text from PDF

2. Using OpenAI’s API - pass the text through the model and prompt it to deliver information in a specific format. This helps account for the variability in the structure/format of different invoices.

3. Write to Excel - Create a new Excel sheet with each record being a new line.

Here’s the full code to copy and paste into a new block in Google Colab:

# Import necessary libraries
import os
import re
import pandas as pd
from PyPDF2 import PdfReader
from openai import OpenAI

client = OpenAI(api_key='INSERT API KEY HERE')

# Function to extract text from PDF
def extract_text_from_pdf(pdf_path):
    text = ""
    with open(pdf_path, "rb") as file:
        pdf = PdfReader(file)
        num_pages = len(pdf.pages)
        for page in range(num_pages):
            text += pdf.pages[page].extract_text() if pdf.pages[page].extract_text() else ""
    return text

# Function to use OpenAI API for parsing text

def parse_receipt(text):
    completion = client.chat.completions.create(
        model="gpt-3.5-turbo",
        messages=[
            {"role": "system", "content": "You are a helpful assistant."},
            {"role": "user", "content": f"Extract the AMOUNT, DATE, and VENDOR from this receipt text: {text}. Please respond exactly in this format: AMOUNT: $amount DATE: MM/DD/YYYY VENDOR: vendor"}
        ]
    )

    print(completion.choices[0].message.content)
    return completion.choices[0].message.content


# Function to parse the structured string into components
def extract_info(formatted_string):
    pattern = r"AMOUNT:\s*(.*?)\s*DATE:\s*(.*?)\s*VENDOR:\s*(.*)"
    match = re.search(pattern, formatted_string, re.DOTALL)
    if match:
        return match.groups()
    return None, None, None

# Function to write data to Excel
def write_to_excel(data, filename="output.xlsx"):
    df = pd.DataFrame(data, columns=["Date", "Vendor", "Amount"])
    df.to_excel(filename, index=False)

# Main function to process all PDFs in a folder
def process_receipts(folder_path):
    data = []
    for filename in os.listdir(folder_path):
        if filename.endswith('.pdf'):
            pdf_path = os.path.join(folder_path, filename)
            text = extract_text_from_pdf(pdf_path)

            if text is None:
                print(f"No text extracted from {filename}. Skipping.")
                continue

            print(f"Extracting from {filename}:")
            print("Extracted Text:", text[:500])  # Print the first 500 characters of the extracted text

            formatted_string = parse_receipt(text)
            print("AI Response:", formatted_string)  # Debug print of what the AI parsed

            if formatted_string:
                amount, date, vendor = extract_info(formatted_string)
                print("Parsed Data:", amount, date, vendor)  # Debug print of the parsed data

                if all([amount, date, vendor]):
                    data.append((date, vendor, amount))
                else:
                    print(f"Data missing in fields for file {filename}")
            else:
                print(f"No formatted data returned for file {filename}")

    # Writing the data to Excel, adjust the filename or path as necessary
    write_to_excel(data, filename='INSERT PATH TO FOLDER HERE'/Processed Receipts.xlsx')
    print("Data written to Excel successfully.")  # Adjust path as needed

# Mount Google Drive
from google.colab import drive
drive.mount('/content/drive')

# Set folder path and process receipts
folder_path = 'INSERT PATH HERE'  # Adjust the path to your receipts folder
process_receipts(folder_path)

You’ll need to make three main edits here:

1. Insert your OpenAI API Key

   1. Insert your key at the line near the top that is “client = OpenAI(api_key='INSERT API KEY HERE'):

2. Insert the path to your receipt folder

   1. Insert the path near the bottom at folder_path = 'INSERT PATH HERE'

3. Insert the path to where you want the program to deposit your new receipts

   1. Insert the path where it says: write_to_excel(data, filename='INSERT PATH TO FOLDER HERE/Processed Receipts.xlsx').

   2. Keep the “/Processed Receipts.xlsx” part, as that is what your new Excel file will be called.

That’s it! You’re ready to run it.

3. Execution

Drop a few PDF invoices in the newly created “Receipts” folder, navigate to Google Colab, and run your code by clicking “Runtime” on the menu bar and navigating to “Run All.”

You’ll be prompted to connect to Google Drive, and your script will run! After a few seconds of running it, you should see a new Excel file, that looks like this, with your invoices nicely extracted:

Now this is just an example of a small slice of what this program could do, if you want, you can have ChatGPT edit your code to have it organize your personalized expenses off your credit card bill!

Limitations and Next Steps

Why Not Use Google Sheets?

You might wonder why we're not using Google Sheets for our project. The reason is straightforward: accessing Google Sheets via its API can be complex, and we aim to keep our tools simple and accessible for everyone. I'll gradually introduce more technically challenging projects, but for now, the focus is on keeping things simple.

What's Next?

At the stage this project marks a significant turning point as everyone has unique accounting and reconciliation needs. In this project, we explore a new approach that uses OpenAI to parse data from invoices, making the process more intuitive than using traditional methods like Regex. Using ChatGPT, take a swing at having it help you code a the program to meet your specifications. It’s very good at error handling for simple applications like this as well!

A Thank You to Our Subscribers

A special shout-out to all our new subscribers, many of whom joined after my recent LinkedIn post. Sharing content with peers I respect is always a bit daunting, but your support has been incredibly encouraging.

Help Spread the Word

If you found this project useful, please consider sharing it. Your support is crucial as we aim to equip the workforce for a new era of AI and automation.

Share

See you in two weeks!

— Nishad