A full-time employee with zero Make.com skills built a personal AI agent team

Please also check the document attached which gives a detailed overview over the project in a better formatted way. Project Background/Project Idea: Please also check the document attached which gives a overview over the project in a better formatted way. The “project” came to life through numerous individual automations I created after getting to know Make.com, Airtable and various AI tools and APIs during the AI Mastery program. In order to…. ·      get deeper into what we have learned during the AI Mastery (since I am not self-employed and do not have my own business with clients to start AI projects for specific use cases with, I have focused on implementing my own ideas and projects in order to apply what we have learned😊) and ·      boost efficiency, save time and become more organized….. …. I created many different automations (here an automation for expense tracking, another for to-do lists, another for meeting recaps, another automation for creating children's stories—including generating illustrations with Flux—and so on). This resulted in many different input forms or channels, each triggering its respective automation. I started wondering whether I could consolidate all the automations (at least those that I really use frequently in my private life—and sometimes also for work—into one single automation, if possible. In addition, I also wanted to take on the challenge of exploring the increasingly discussed and relevant topic of AI agents (something I find fascinating and I am currently also exploring with using various tools such as Flowise, Lindy, n8n, Relevance AI etc..). For this project my goal was to create my own personal “AI agent team” with several "sub-agents" within Make.com. The result: ·      A Make.com “Personal AI Agent Team” Automation featuring a "supervisor" (represented by a ChatGPT module) that assigns and delegates tasks to the various "sub-agents" (corresponding to different paths in Make.com). Input is provided exclusively through a single input channel (in this case, a specific Slack channel) via Text, Audio or by dropping a PDF file. ·      Additional few complementary Automations and Tools have been created, in order to complete the “Personal AI-Powered Agent and Automation system”: o  Meeting notes summaries and Tasks creation automation (Fireflies, Airtable and Make.com) o  Invoice Management (Airtable and Make.com) o  Virtual Phone Assistant (Retell AI, Make.com and Airtable) Remark: The initial automations and databases were created in German Language and with (my) real data. I tried to translate everything to English as good as possible. The demo and the links refer to “Fake data” or some “Test data” created specifically for showing this solution in English language.  Project Description and Outcome This project consolidates numerous personal automation workflows into a unified AI-powered system and automation à”Personal AI Agent team”. Utilizing a central AI "supervisor agent” with various "sub-agents," the system streamlines tasks such as expense tracking, research, meeting summaries, creating calendar events and more via a single input channel in Slack. The outcome is an efficient, centralized, and intelligent assistant system tailored for private (can be customized as desired) and potentially also professional use. At its core is an AI "supervisor" module, “represented” by ChatGPT module, that identifies the tasks or requests and delegates them to various "sub-agents" (corresponding to different paths with specific filter criteria within Make.com scenario). The “Supervisor Agent” accepts inputs through a dedicated private Slack channel using text, audio or PDF uploads, ensuring seamless communication and interaction. By combining automation with AI, this project helps to become more organized and increases efficiency in both personal and (potentially also) professional life. Any task (that might be relevant in the future) can be added to the “Personal AI Agent team” by adding a new path within the Make scenario.  What can this Agent Team do? ·      Expense Tracker: o  Add ongoing expenses (e.g., a coffee here, a dinner there, supermarket purchases, etc.) to my expense tracking in Airtable. ·      To-Do-Manager: o  Add to-do items to my Airtable to-do list (especially useful for tasks that come to mind spontaneously while on the go). ·      Research Assistant: o  Conduct research on a topic and convert the results into an audio format and send it to a dedicated Slack channel for research (I like to listen to these while traveling or driving). The results are also stored in Airtable and Google Drive in a specific folder. ·      Email Drafter: o  Draft emails based on input given à While the idea of creating, sending and responding to emails in automatic way could be expanded further, I am cautious here and prefer not to have emails automatically sent. ·      Meeting Recap Manager: o  Provide meeting recaps for meetings that occurred within a specific time frame (e.g. “Please give me an overview over all meetings from last week, January 6th to January 10th”). These recaps are made available as audio files. o  Provide meeting recaps for past meetings related to a specific topic or project, also delivered as audio files. (e.g. “Please provide me an overview over what has been discussed in previous meetings for “Project X”.) - Children story creator: - Create children’s stories with image generation via Flux (I absolutely love Flux!). The audio version of the story, the illustrations, and a written version for reading aloud are sent to my sister via email so she can read them to my nephew. This is not really related to organization and productivity, but an important task for me personally, as my nephew loves to get a new story which my sister can read to him in the morning or evening. By just entering one sentence about what the story should be about, the story, image and audio will be created and will also be send automatically. - Idea and Notetaker: - Adding ideas and notes to my ideas/notes table in Airtable (with the option to trigger another automation from there that takes the idea and prompts the AI to explore solutions, additional aspects, perspectives, thoughts, and a concept for it). Often, ideas come to mind spontaneously and on the go, which I would like to capture and revisit later. ·      Google Calendar Manager: o  Provide overview over calendar events scheduled in my google calendar for a mentioned time frame o  Create new calendar events for specified time ·      PDF Summarizer: o  Summarize PDFs (using Claude API in this case as PDF Analysis with Claude is amazing), creating a concise summary and generating an audio file. These are saved to Google Drive and sent to the Slack channel (I like to listen to these on the go or while driving). The summary and audio are also stored in Airtable. Additional complementary automations and tools In addition to this “Personal AI Agent team” system/automation there are few additional automations that are completing the whole “Personal AI-Powered Agent and Automation system”. ·      Meeting notes taker (Airtable and Make.com + Fireflies to record and transcribe meetings) o  automatically transcribes and summarizes meetings whenever a meeting was recorded with Fireflies. Without taking any specific action the automation, which is running daily, is creating “to do’s” and meeting summaries which will be added to the Airtable Meeting database “Meetings english”, to which the “Personal AI Agent team” above is having access to in order to create and provide meeting overviews for a requested timeframe or topic o  the meeting summaries and tasks/action items created with this automation will also be sent as email o  the summary of the meeting will also be provided as an audio file in addition to text summary ·      Invoice Manager (Airtable and Make.com) o  Invoices that are either not paid immediately (and therefore should not yet be recorded as "expenses" in my system) or are relevant for tax purposes (e.g., home office expenses) or health insurance (typically accompanied by a PDF invoice or a photo (image) of the invoice) are added separately to the database “Expense Tracking (in english) into table “Invoices” via an Airtable form. As soon as a new entry has been created with the Airtable form, a webhook in Make.com will be called and the automation below will be triggered -->Using AI vision, the information (amount, invoice number, due date, and, if available, bank details for the bank transfer, etc.) is extracted and updated into the Airtable table “Invoices” for the respective record. Additionally, the invoices are categorized using the Airtable AI. The categories align with those used in the expense tracking system managed by the “AI agent team”. o  Both expense tracking (table “Variable expenses”) and invoices (table “Invoices”) are recorded and organized within the same database but in different tables. o  An additional automation runs daily early in the morning to check which invoices are still not paid and are due the following day. These are then sent via email through an Airtable automation. o  When an invoice is paid, by marking the "Paid" checkbox in the table, adds the entry to the "Already Paid" view, which triggers another Airtable automation. o  This automation calls a webhook in Make.com via a script, triggering another automation that adds the paid invoice to the expense tracking table (as it is now finally paid, the expense should be considered within the expense tracking “Expensed variable” table). ·      Ideas and notes follow up o  For the ideas and notes added to the “To Do list (English)” database to table “Ideas and notes” via the “Personal AI-Agent team”, it is possible to trigger an automation that is taking this idea or note and to follow up on this by analyzing the idea, creating a concept draft, providing different perspectives and considerations o  This could also have been done directly when entering the idea or note via the “Personal AI-Agent team”, but as not every idea or note is worth a follow up I wanted to trigger this manually and separately àtherefore it will be triggered by using a button in Airtable, which triggers a Webhook in Make ·      Virtual phone assistant (Retell AI, Make.com and Airtable) o  With Retell AI, I created my own virtual phone assistant, which is linked to my personal number and activates whenever I don’t answer a call within 15 seconds. My virtual phone assistant, Joe, answers the call and informs the caller that I am currently unavailable. Joe collects some information, such as the caller's name, the reason for the call, and whether additional details should be forwarded to me. o  Once Joe has gathered all the information, she triggers a Make automation via a webhook. This automation sends the details to me via email and also creates an entry in my Airtable database with the information and details from the call.  o  At the end, Joe politely says goodbye and thanks the caller and assures that I will call back. Key Benefits of the “Personal AI-Powered Agent and Automation system” - Saving Time by using automations for various tasks in my daily life - Increase Productivity and Efficiency by being now able to easily add expenses, to do’s, ideas etc. automatically to the respective data bases or formats (before, my expense tracking or to do list was a chaos and basically not existent as it was not carried out consistently due to the effort involved and the use of various systems, tables, or tools); better invoice handling - Research Summaries: Automated topic research with results converted also to audio files for convenient consumption. - Meeting Notes creation, Tracking and Recap: Creation of audio and text summaries of meetings, storing them and providing recaps and overviews categorized by date and topic, enhancing information retrieval. - Creative Storytelling: Children's stories with AI-generated illustrations and audio. This is not really related to organization and productivity, but an important task for me personally, as my nephew loves to get a new story, which my sister can read to him in the morning or evening. By just entering one sentence about what the story should be about, the story, image and audio (with the story) will be created and sent to them automatically. - Document Summarization: Concise summaries of PDFs with corresponding audio outputs, stored in Google Drive for easy access. - Easy Access to Calendar events by just asking the “Personal AI-Agent team” - Easy way to create calendar events - Become more organized in general as I now have a “system” which makes it easy for me to add for examples expenses, invoices, ideas and notes, meeting notes etc. - Enhanced Accessibility: Key information is available in multiple formats (text, audio, visual). - Simplified Communication and limited input channels or forms: A single Slack channel serves as the communication channel with the “Personal AI-Agent team”, minimizing distractions and streamlining operations. In addition to this, there is one additional Airtable form specifically for uploading invoice documents.  Tools used Make.com modules/APIs: - Slack: Central communication channel with the “AI agent team” - Open AI: Message an Assistant, Create Completion, Whisper (Create a transcription) - Airtable: Database for expenses, tasks, ideas, meetings, PDF summaries and research outputs. - Perplexity AI for research - Google Drive: Storage for research summaries, meeting recaps, and other outputs created. - Google Docs - Flux (access via HTTP module): Tool for generating illustrations for stories. - Claude API: PDF analysis and summarization. - Google Calendar: for calendar overview and to create events - Google Mail: to create email draft and to send emails with content created (children story) - Fireflies: to transcribe meetings ·      Flow Control: Router, Iterator ·      Tools: Text aggregator, Sleep module - Data store: to check for meetings (transcripts from Fireflies) already processed  Additional (AI) tools used apart from Make.com: ·      Slack as communication channel with “Personal AI Agent team” ·      Fireflies AI for recording and transcribing meetings ·      Airtable (Data, Forms, Automation, Scripts to call a webhook in Make, Airtable AI for Classification) as main database for all data and topics ·      Retell AI for phone assistant setup Potential future improvements ·      Improve some prompts to get even better results (as the initial project was done in German language, the English translations might not always be perfect for all prompts etc.) ·      Any future task that could be automated can be added at any time Relevant Links: ·      Airable Databases: o  Research Database § https://airtable.com/appwkqCstn8LkQ1v3/shrd8wKYVRv1CD7Dt o  To-Do-List and Notes and Ideas Data Base § https://airtable.com/appatVeN1FtyTBI3j/shrxHyXWWG7oKQV7q o  Expense Tracking and Invoice Database § https://airtable.com/appqyRWmCFzQnQVc8/shrTdeG9OuHH2eRnt o  Meeting notes and recap Database § https://airtable.com/appz0eGQ90BSluCIy/shrp7ubfj4UUMvZ4S o  Phone Assistant Database § https://airtable.com/applEeDPMPiqtzlsu/shrjeSxzzERdaT7Ek o  Contact List Database for Calendar Manager § https://airtable.com/apppX5viKgnym0mrm/shrDukzT5Y4rgd4GA o  PDF Summaries § https://airtable.com/appSaWakSbZqR3t93/shrzQIBgSOotZBeBu ·      Airtable form for adding invoices o  https://airtable.com/appqyRWmCFzQnQVc8/pagAztl71DlPOhYBD/form ·      Link to Loom video with Demo o  https://www.loom.com/share/ea56b8c2c0364dcf8b1ccbe909afaaac Learnings and Conclusion Since I am not self-employed, do not have my own business, or work as an independent consultant who could implement AI projects directly with clients right away, it was important to me to find a way to apply and implement what I have learned in the AI Mastery program. It was challenging to complete the AI Mastery program and apply what I have learned alongside my full-time job. While I couldn’t always attend live, I made sure to catch up with the recordings afterward. Without the enthusiasm and joy, the brilliant and engaging sessions from the trainers, and Vyk’s fantastic office hours, it would not have been possible to stay committed and spend so much additional time to learn and apply what we have learned. This project allowed me to better understand the tools and deepen my knowledge. It helped me put a significant portion of what we have learned in the program into practice by creating an incredibly useful and efficient system for me (and also for my sister and two friends, as I’ve already set it up or promised to set it up for them in a similar way but customized to their fit). This system has made me become more organized, efficient, and productive. I’ve learned that almost everything can be automated with Make.com and that the combination of automation and AI holds enormous potential and power. I have learned how to create my own “AI agent team” with Make, which I feel gives me more control over the output compared to autonomous AI agent teams (at least for now, but I might also become better in creating autonomous Agent team with tools like Flowise, Lindy or n8n….). With the help of AI, I was able to create also complex formulas in Make or Airtable and to create script code in Airtable which is calling a webhook in Make or to create a Custom function (JSON schema) in retell AI to call a webhook in Make. I have also learned how to organize myself in a better way using Airtable (I now manage almost everything in my personal life through Airtable) and how to use Airtable AI effectively. Through continuous testing, I have learned how to improve and optimize my prompts, gaining a better understanding and knowledge of which types of prompts work better or worse and how to achieve the best output. I am fascinated by what is possible with AI and how much fun it can be to implement ideas and see the results. I started last summer with zero knowledge in Make.com or Airtable, and I’ve since become a huge fan of these tools (thanks to Vyk 😉). I’m incredibly grateful to the entire AI Mastery program team for this amazing journey. Thanks to all of you involved in the program! You have inspired me with your enthusiasm for AI. It is an honor to have been a part of it. I will for sure continue to use AI and build Automations and I am looking forward to learning more about it.