Agentic Millennial.
Non-tech. Non-coder. Windows user. I'm documenting my journey of integrating agentic AI into my daily workflow, and sharing the practical shortcuts I find along the way.
01Codex
Setup
★★★☆☆- I created a ChatGPT account and found Codex in the navigation menu. During setup, I was prompted to install Codex for Windows. This part was not completely straightforward, and I needed to troubleshoot access through the Command Line Interface (CLI).
- Once installed, I customised the workspace settings, including sandbox, personality, and plugin preferences. This helped me create a setup that better matched how I work.
- I also connected tools I use regularly, including GitHub, Cloudflare, Gmail, and Google Drive. These integrations helped Codex interact with the services I rely on and made the workspace feel more connected.
Cost & ROI
★★★★☆- Usage can be checked in Settings under Usage Remaining. This shows current limits, including short-term and weekly usage. Monitoring this section helps me understand how much capacity I have available and whether my current plan is enough for my workload.
- I upgraded to the Plus plan because I was using ChatGPT and Codex regularly for research, content creation, learning technical concepts, and website development. Since upgrading, I have not reached my limits despite using ChatGPT on a near-daily basis.
- Plan details, usage allowances, and pricing can change over time, so Usage Remaining should be treated as the best place to check the latest information for your own account.
Personal Use Cases
★★★★☆-
Website build
I used Codex to build this landing page by connecting GitHub and Cloudflare, then using Codex's agentic function to develop copy, adjust images, and refine the UX/UI.
Pros:Understands most natural language prompts well.
Challenges:Occasionally, changes do not push to GitHub correctly, which can leave the website and repository out of sync.
-
Tip: For longer projects, such as building a website, enable goal mode by typing
/goalin the chat. It helps Codex stay focused on the wider objective and check progress as it works. -
Gmail Inbox Triage
I used Codex to help triage my Gmail inbox by adding labels, creating a daily digest of key actions, and removing emails I no longer needed.
Pros:Understands most natural language prompts well and can develop multiple automations with distinct schedules.
Challenges:When Gmail is connected as a plugin, automations only occur when I am active; this is referred to as a local workflow. Logging off stops automations. There is a workaround which requires development of a cloud Gmail workflow with Gmail API. Watch this space!
02Google AI Studio
Setup
★★★★☆- Google AI Studio can be accessed from aistudio.google.com and saved to desktop for easier access.
- The user interface is the most comprehensive of the tools I have tested so far.
- The left panel includes the Playground, where I can access different models and start building.
- Models are available for app development, image generation, video generation, and other use cases.
- The left panel also includes the Dashboard, where I can check usage, manage API keys, and review rate limits by model.
- The right panel lets me select the model I want to use. For example, image generation requires Nano Banana, while video generation requires Veo.
-
Tip: the bidirectional arrow near the top-right of the page lets me test two models against the same prompt and compare the outputs side by side.
Cost & ROI
★★★★☆- I have only used the Free tier plan in Google AI Studio, which has been sufficient for my personal use cases.
- Google AI Studio is useful for testing Gemini models before committing time or budget to a larger workflow.
- On the left-hand side, click Dashboard to assess Usage and Rate Limits, among other features.
- Usage is tied to model-specific rate limits, which can fluctuate depending on model demand at any given time.
- Rate limits are defined by three metrics:
- RPM (Requests Per Minute): how many times I can send a request to the model within a 60-second window.
- TPM (Tokens Per Minute): the total volume of input and output text or media the model can process per minute.
- RPD (Requests Per Day): the maximum number of prompts I can send in a 24-hour period.
-
Tip: Rate limits are not tied to individual API keys. They are enforced at the Google Cloud Project level, so the whole project shares the same limits. Therefore, it is worth checking the Dashboard before relying on a model for repeated or time-sensitive work.
Personal Use Cases
★★★☆☆-
Website Build Testing
I used the latest Gemini model, Gemini 3.6 Flash, in Google AI Studio to review this website built with HTML, CSS, and JavaScript, with version control managed through GitHub and deployment via Cloudflare.
To perform the audit, I copied the relevant source code from GitHub into AI Studio and asked it to review the website's:
- layout and user experience (UX)
- content and copy
- accessibility
- performance
- security best practices
AI Studio identified areas for improvement and generated updated code snippets that could be copied directly back into GitHub before redeploying the website.
Tip: Enable Auto-save, or manually save the prompt, so it is stored in Google Drive. This makes it easier to revisit and reuse the same prompt for future website reviews without recreating it from scratch.
03Relay
Setup
★★★★☆- I created a free account with Relay.app.
- The free plan included monthly AI credits and workflow steps, which let me test automations before committing to a paid plan.
- I used natural language prompts to build workflows.
- Relay generated visual flowcharts showing each step, which could be adjusted and tested before connecting them fully to apps such as Gmail.
Cost & ROI
★★☆☆☆- Relay offers Free, Professional, Team, and Enterprise plans.
- Pricing can vary depending on the number of workflow steps required.
- Each workflow step may use one or more AI credits, and credits can run out quickly. On the free plan, my credits ran out after three days of automations.
- I have not used a paid plan; instead, I used Relay to understand the workflow structure before rebuilding some automations in Codex.
Personal Use Cases
★★★☆☆-
Gmail Inbox Triage
I used Relay.app to triage my Gmail inbox by adding labels, creating a daily digest of key actions, and deleting emails I no longer needed.
Pros:Easily connects to Gmail and understands most natural language prompts.
Supports automations with custom schedules and does not require me to be logged in.
Allows workflows to be manually tested and paused from the dashboard.
Customer support responded quickly when I had issues.
Challenges:Relay can be case-sensitive. For example, "Newsletter" and "NEWSLETTER" may be treated as different Gmail labels, which can cause workflow errors.
-
Tip: Use the same spelling and capitalisation every time you reference labels, folders, or workflow names in Gmail and potentially other email clients.
04NotebookLM
Setup
★★★★☆- I accessed NotebookLM through my Google account at notebooklm.google.com. Google is now increasingly referring to the product as Gemini Notebook.
- I created a new notebook, which works like a focused workspace for one topic, project, or research area.
- I added source material into the notebook, such as PDFs, websites, Google Docs, Google Slides, YouTube videos, and audio files.
- Once sources were uploaded, NotebookLM generated a source-grounded workspace where I could ask questions, summarise information, and extract useful insights.
- The tool is particularly useful because responses are based on the sources added to the notebook, rather than general internet-style answers.
- The setup was straightforward because it did not require coding, technical configuration, or installation on Windows.
-
Tip: NotebookLM works best when the source material is clear, relevant, and well organised. The quality of the output depends heavily on the quality of the documents or links added.
Cost & ROI
★★★★☆- I have only used the free version of NotebookLM so far, which has been sufficient for testing personal research and note-taking workflows. The free tier is generous enough to create notebooks, upload source material, ask questions, and generate useful summaries without needing to pay upfront.
- Paid access is available through Google AI plans rather than as a simple standalone NotebookLM subscription. These plans increase usage limits and unlock broader Gemini features across Google products.
- From a cost and ROI perspective, NotebookLM is most valuable when working with long-form source material, such as PDFs, websites, webinar notes, Google Docs, or research documents. I would only consider upgrading if I regularly hit the free usage limits or needed to work with larger notebooks and more frequent AI-generated outputs.
Personal Use Cases
★★★☆☆-
Webinar Learning
I have uploaded webinar transcripts into NotebookLM and asked it to summarise the key themes, then break down technical terms into plain, digestible English.
-
PowerPoint Presentation Updates
I have uploaded legacy presentations from similar projects into NotebookLM and used it to identify themes, refresh copy, and align the content with newer industry trends.
-
Social Media Post Development
I have used NotebookLM to generate LinkedIn-style post ideas and carousel outlines from existing notes, transcripts, and research material.
05Notion
Setup
★★★★☆- I signed up for Notion using my Gmail account.
- In Notion, content is organised into "blocks", which can be combined to create pages, tables, notes, and workspaces.
- Because I signed up with Gmail, Notion Calendar automatically synced with my Google Calendar.
Cost & ROI
★★☆☆☆- Plans can be managed in Settings under Access & Billing.
- I have only used the Free plan so far and found some functionality limited, especially around integrations.
- I found it frustrating that many app features are paywalled. As a new user, being able to test relevant features properly on the free plan is important before deciding whether a paid upgrade is worthwhile.
- Paid options include Plus, Business, and Enterprise, with monthly and annual billing options.
Personal Use Cases
★★★☆☆-
Webinar note taking & Social Media Post Development
I used Notion to record and transcribe live webinars from my calendar using the AI meeting notes feature. Notion also generated summaries, which I could duplicate, edit, and repurpose for email or social media.
Pros:Simple recording flow: click Record, then Stop. The transcript and AI summary are created automatically once the recording ends.
Challenges:On the Free plan, I was only able to record five one-hour webinars before running out of AI meeting notes.