Building a Make.com Personal Assistant
Building a Make.com Personal Assistant
I keep having ideas about blog topics, product improvements and business opportunities on the move but I had no reliable mechanism to capture them. Often I email myself but then my inbox is clogged up for weeks or they don’t all make sense when I revisit them. Even worse, my family has numerous appointments that never make it into the diary, making scheduling incredibly difficult and results in missed appointments.
By this point I’d already built VisaScopeAI entirely with no-code tools, proving the concept worked. After attending a Make.com workshop about their AI capabilities, I became curious whether these tools could solve this personal frustration. This wasn’t just a technical experiment it was also about trying to find a way to solve problems for myself and my family.
The system needed to work on the go, using whatever method was easiest in the moment. The first decision was choosing the capture method or “trigger”. The tool that would receive my input before passing it to the AI pipeline.
I wanted flexibility, email so I could forward on emails I had received from the school or clients, voice for quick captures whilst walking and messaging for speed. This meant testing several options to find what actually worked with Make.com’s integrations.
Telegram seemed perfect until I discovered Make.com’s integration only works with desktop triggers, which was useless for mobile capture. WhatsApp would have been ideal but connecting it costs €70-300 per month minimum. Far too expensive to validate a personal MVP. Email was pragmatic but connecting Gmail to Make.com required setting up a Google Cloud Project with OAuth configuration which was too complex in this case.
Slack emerged as the winner. The free tier supports both text and audio. It had a solid Make.com integration and works perfectly on mobile. I wasn’t 100% happy with Slack as I had wanted to try a new product I don’t normally use and it didn’t have email integration, but I worked on a solution for that which I will detail.
I took the same incremental approach I use for product development at work proving that each piece works before combining them.
I started with one outcome test. Could I send a text message to Slack and have the AI create a meeting in my Google Calendar? This would validate the core concept that AI could reliably understand natural language and turn “meeting with the accountant Tuesday at 2pm” into an actual calendar event.
Mapping the AI output to the Google Calendar module to behave took more work than I expected. I had to programme the agent to output the exact values that were needed for the module which resulted in me having to provide the exact JSON output required in the prompt. Once this was in place things moved forward quickly.
Next came Notion. I’d never used Notion before, so I worked with Claude and ChatGPT to design a database for all tasks, notes, ideas and references. The concept was straightforward: anything that wasn’t a meeting would land here, automatically tagged and categorised based on the content of my message.
The Notion API connection caught me out. Connecting it to Make.com requires three steps and the third one wasn’t obvious. I needed to explicitly share the specific database with the integration. The authentication token alone isn’t enough. Without this step, everything appears to work, but Make.com can’t actually see your databases.
Finally came voice capture. I added Whisper for transcription, which genuinely surprised me with its speed and accuracy. I could record a voice note whilst walking and it would transcribe perfectly. The challenge was that Make.com doesn’t let you reuse workflow modules across different paths so I had to duplicate the entire AI workflow and routing logic for both text messages and audio files.
The entire build took approximately two to three days of part-time work and I was happy with the quick development of the MVP.
The one thing Slack’s free tier doesn’t support is email-to-channel forwarding. Rather than upgrading or abandoning the feature I built a workaround using a separate Make.com workflow.
The solution was pretty straightforward. I set up a Gmail folder that my workflow watches. When a new email arrives in that folder, Make.com extracts the subject and content, then posts it as a message into my Slack channel. Once it’s in Slack, the original workflow picks it up and processes it exactly like any other message and the AI doesn’t know the difference.
This meant I could forward school emails, client messages, or anything else directly into the system without leaving my email app. No complex OAuth setup, no paid Slack features required.
The system handles most inputs correctly. I can record a voice notes such as “Blog idea, how no-code tools are changing product development” and it appears in my Notion database with appropriate tags. I can type “Meeting with procurement team Tuesday 2pm” in Slack and it creates a calendar event. I can forward an email from my daughter’s school and it captures the appointment.
It’s not perfect. Sometimes the AI misinterprets context, particularly with sarcasm or informal language. There’s still more work to do. I want to try and extract email attachments as text as currently they are ignored. But the success rate is high enough that I actually use it daily. It genuinely saves time rather than creates additional work.
The unexpected benefit: forcing myself to articulate ideas clearly. Because the system requires reasonably structured input, I’ve become better at formulating thoughts precisely.
This wasn’t just about building a personal tool. The implementation process validated an approach I’d been curious about for some time. No-code automation platforms have reached a point where they can solve real problems quickly, provided you accept their constraints.
The entire build came together within a couple of days. Some planning, testing, building and trial and error. AI brings a new dimension to the workflows but they are lightweight and robust and perfect for smaller organisations looking to make efficiency savings and reduce manual work.
I’m continuing to refine the system. There are improvements I want to make, edge cases to handle better, and additional integrations to explore. But it’s working well enough that it’s become part of my daily routine, and that’s exactly what I needed it to do.
Building workflow automation solutions for your organisation? I help companies identify opportunities where intelligent automation delivers disproportionate value.
