My AI Daily Planning Test Cancelled Half My To-Do List
I gave Claude my real 12-task to-do list and asked for honest AI daily planning. It deleted four tasks outright and I overruled one of its calls.
Work
July 23, 2026 · 5 min read
I gave myself one afternoon and a budget of zero to build a freelance CRM with AI. By dinner I had a working spreadsheet that tracks every lead, every open proposal, and every unpaid invoice in one view. What I did not have is the thing every "AI CRM in a day" article promises: automation that runs without me touching it. This is for freelancers and solo consultants who keep their client list half in their head and half across three inboxes, and want one honest place to see it all. By the end you'll have the exact prompt that builds the skeleton, the tools that are genuinely free, and the one wall that stops the day cold. Fair warning: the wall isn't where I expected it.
A freelance CRM is one place to see every lead, every open proposal, and every invoice you're still chasing. That's it. The enterprise version forecasts sales teams and quarterly numbers; the freelance version answers three questions on a Monday morning: who owes me a reply, who owes me money, and which deal is about to go cold. If you've already let AI plan your workday, this is the client-facing half of the same idea. I wanted those three answers without paying $20 a month or learning software I'd quit using by Friday.
The plan was to skip signing up for a dedicated CRM and have AI design a spreadsheet I'd own outright. Every dedicated tool wants an account before it shows you anything, and the free ones paywall the exact feature you came for. HubSpot's CRM is genuinely free forever, but as of July 2026 the free plan caps you at 1,000 contacts, two users, one deal pipeline, and zero workflow automation. The automation lives in Sales Hub, which starts at $20 per user a month.
So the free tier hands you a place to type names and nothing that moves on its own. Let's do the math: if AI could design a spreadsheet that does the same tracking job, that's $240 a year I keep in my pocket. That was the bet for the afternoon.
Ask an AI to design a freelance CRM and a single prompt gets you a usable spreadsheet skeleton. Here's the exact one I used.
Design a simple CRM I can run in one spreadsheet as a freelancer. Give me one tab with these columns: Client, Contact, Stage (Lead, Proposal sent, Won, Lost, Paid), Value, Last contact, Next action, Next action date, Notes. Make Stage a dropdown. Then give me one formula that flags any row where Next action date is already in the past.
What comes back is genuinely usable: nine columns, a Stage dropdown with five options, and a formula that turns overdue rows into a warning when the next-action date slips behind today. Paste it into Google Sheets, Excel, or a free Airtable base and the skeleton is standing in the time it takes to drink a coffee. I ran the same prompt twice and got the same structure both times, and that's the point. It's boring and repeatable, which is exactly what you want from plumbing.
The build broke at one spot, and it's the same spot for everyone: automation. The skeleton holds a list. It won't email a client when a proposal goes quiet, log a call by itself, or link this month's invoice back to last month's lead. That last part, records in one tab actually connecting to another, is where a spreadsheet stops being a CRM.
I'm not the first to hit this. A team at enable.services ran the same experiment in January 2026 with ChatGPT and Gemini, and walked into the identical wall. The AI couldn't create a live Google Sheet, only an Excel file. The free access rate-limited them to one request an hour. And neither model could make the records genuinely link across tabs. Their verdict was blunt: a CRM built on a spreadsheet, even with AI as capable as ChatGPT and Gemini, is still just a spreadsheet. Both models even told them to go buy real CRM software.
I couldn't close that gap either, and I'll be straight about why. Connecting email, firing an automation, or importing a real client list means signing up for a tool and, past the free tier, paying for it. I didn't stand up a live automated pipeline for this piece, so I won't pretend an automation fired when it didn't. What I can tell you is exactly where the free road ends.
A spreadsheet CRM is enough right up until you're juggling fifteen live conversations at once. One solopreneur CRM guide put the tipping point well: a spreadsheet works fine at three leads, and falls apart at fifteen conversations, four proposals, two delayed payments, and one client who only replies after nine business days of silence. That's the line.
Below it, the AI-built sheet wins on cost, hands down. Above it, you want real record links, and a free Airtable base (1,000 records per base, relational tables) or Notion gets you there without costing anything yet. Here's how the free roads compare before anyone reaches for a credit card.
| Setup | Cost | Automation | Good until |
|---|---|---|---|
| AI-built spreadsheet | $0 | none | ~15 active clients |
| Airtable free | $0 | 100 runs/month | 1,000 records per base |
| HubSpot free | $0 | none (paywalled) | 1,000 contacts, 1 pipeline |
| HubSpot Starter | $20/user/mo | full workflows | you've outgrown solo |
I ended the day with a $0 spreadsheet that answers my three Monday questions and a clear map of what AI won't do for free yet. The skeleton took an afternoon and cost nothing: any spreadsheet, one AI assistant, difficulty a 2 out of 5. The automation I actually wanted, the part that would make it feel like software, would have cost a signup and a monthly bill, and no AI closed that gap on its own. If you want to see what real automation takes once you're ready to cross that line, our walkthrough on how to set up your first AI agent is the honest next step, and our roundup of AI tools that are actually free covers what else you can run at $0 first.
Yes for the skeleton, and no for expecting automation in a day on a zero budget. The AI-built spreadsheet is the fastest cheap CRM a freelancer can stand up today, and for most solo careers it's genuinely enough. What I'd drop is the story that AI runs my whole pipeline by tonight. That part isn't free, it isn't one afternoon, and often it isn't even worth it until you're busy enough to feel the pain. Get the list right first. Bottom line: build the sheet today, and buy the tool the week it starts costing you sleep.
I gave Claude my real 12-task to-do list and asked for honest AI daily planning. It deleted four tasks outright and I overruled one of its calls.
I ran a day of errands with an AI browser agent. It nailed the finding half. The buying half is exactly where you keep your hands on the wheel.
I tried to work faster with AI by handing it every task in my workday. What was done before lunch, what nearly cost me a client, what AI couldn't touch.