From Chaos to Clarity: My Experience with an AI Time Management Strategist
A candid look at what it's really like to hand your calendar over to an AI — the stumbles, the breakthroughs, and what I'd tell anyone thinking about trying it.
🌱 How It Started
Like most people who work in product and project management, I didn't have a time problem — I had a prioritization problem. My task list was technically complete. Every to-do was captured, tagged, and sitting in a Notion database. But every morning I'd open the board and feel the same low-grade dread: Where do I even start?
That's what led me to build and deploy a custom AI agent directly inside my Notion workspace — one wired into my task database, my Google Calendar, and a set of structured instructions about how I actually want to work. I called it the Time Management Strategist.
What followed was several months of iteration, frustration, small wins, and eventually, a genuine shift in how I spend my days.
⚙️ What the Agent Actually Does
At its core, the agent does three things:
- Plans my day or week by pulling tasks from my Notion database, reading their priority, energy level, and estimated duration, then building a realistic time-blocked schedule.
- Exports that schedule to Google Calendar — one event per task, with the task's Notion URL embedded in the location field so I can jump back to context with one tap.
- Runs a nightly sync on weekdays at midnight, automatically cleaning up stale calendar events when task due dates change or tasks get marked done.
Under the hood, it uses an energy-aware scheduling model. Every task in my database carries an energy tag — 🔴 Deep, 🟡 Medium, or 🟢 Light — and the agent slots deep work into my peak focus hours and lighter admin tasks into the afternoon dip. It also protects a hard lunch block from 1–2 PM and flags any day where I've overcommitted before it ever hits my calendar.
🪨 The Challenges
Getting the profile right
The agent is only as good as the context I give it. The first version of my personal profile — work schedule, peak energy hours, recurring commitments, non-negotiables — was vague and optimistic. I told it my peak hours were 8 AM–noon. In reality, I don't hit my stride until closer to 10 AM after I've cleared Slack and had coffee.
It took two or three weeks of the plans technically being correct but feeling wrong before I revisited the profile and tightened it up. The lesson: garbage in, garbage out. The AI isn't psychic — it plans based on what you tell it about yourself.
Trusting the reject
Early on, I had a habit of just accepting plans without pushing back. The agent always ends a proposed schedule with a simple prompt: accept to push it to calendar, or reject to revise. I kept accepting out of some weird social pressure — as if declining the AI's suggestion was impolite.
Eventually I learned to use the reject loop aggressively. "Move the deep work block to Tuesday, I have a client call Monday afternoon." The agent rebuilds the plan in seconds. That back-and-forth is actually where the most value lives.
Overdue task accumulation
This one stung. For a while I was using the agent to plan forward without ever cleaning up the backlog. Tasks that were 3 weeks overdue sat in the database and occasionally surfaced in plans in awkward ways. The agent has a rescheduling feature — it uses a tiered formula to push overdue tasks forward based on how late they are and their energy level — but I wasn't using it.
Once I built a habit of running a weekly reschedule pass (usually Sunday evening), the plans became dramatically more accurate and achievable.
🏆 The Successes
The plan actually fits the day
The single biggest win is mundane but profound: I stopped overbooking myself. The agent counts available hours, accounts for lunch and breaks, and flags overcommitment before it schedules anything. When it tells me a task needs to move to tomorrow, I believe it — because it's done the math.
Energy-matched focus
Pairing task energy level with time-of-day scheduling sounds obvious, but I'd never actually done it consistently before. Now, when I open my calendar at 9 AM and see a deep work block already waiting for me, I don't have to make a decision. I just start. That reduction in decision fatigue compounds over a week in ways that are hard to overstate.
Calendar as a source of truth
The nightly sync changed my relationship with my Google Calendar. Before, my calendar was mostly meetings. Now it's a real picture of my day — meetings and work. When something shifts in Notion (a deadline moves, a task gets done), the calendar updates automatically. I no longer carry the cognitive load of keeping two systems in sync manually.
The weekly review habit
The agent's weekly planning output includes a set of review prompts at the end: What did I accomplish? What didn't get done, and why? What should I prioritize next week? I've answered these prompts more consistently than any journaling habit I've ever tried to build. Having the scaffold already there removes the activation energy.
🔄 How It Has Evolved
Version one of the agent was a simple planning assistant — no calendar integration, no nightly sync, no rescheduling logic. I'd ask it to plan my day, it would output a nice formatted schedule, and I'd manually block my calendar.
Over time I layered in:
- Calendar export (accept/reject flow)
- Energy-based time estimates as fallbacks when duration isn't set
- The midnight sync trigger to keep calendar events current automatically
- Overdue task rescheduling rules with energy-level adjustments
- Hard scheduling constraints — the 1 PM lunch block, the 5 PM cutoff, the post-lunch deep work exception
Each addition came from a real pain point. The sync came from noticing that my calendar was full of ghost events for tasks I'd already finished. The rescheduling logic came from the overdue backlog problem. The energy fallbacks came from tasks I'd never bothered to time-estimate.
The agent today is meaningfully more useful than what I started with — not because I changed the underlying AI, but because I kept refining the instructions based on what wasn't working.
💡 What I'd Tell Someone Starting Out
Start with the profile, not the features. The scheduling is only as useful as the context you give it. Take 20 minutes and fill in your energy hours, work schedule, and non-negotiables with real honesty — not aspirational ones.
Use the reject loop. The accept/reject prompt at the end of every plan isn't a formality. It's the primary interface. Don't rubber-stamp plans that don't feel right.
Clean your backlog regularly. The agent can't plan well if your task database is full of zombie tasks from three months ago. Run a reschedule or archive pass at least weekly.
Let the calendar sync do its job. Once you trust that the nightly cleanup is running, you can stop manually managing calendar events. That trust takes a few days to build — but it's worth it.
🪟 Where Things Stand Now
Most mornings, I open Notion, ask the agent to plan my day, glance at the proposed schedule, and either accept it or tweak one or two blocks. The whole ritual takes under five minutes. My calendar reflects my actual intentions — not just my obligations.
I still have chaotic days. I still have weeks where everything goes sideways. But I no longer start the day staring at an undifferentiated pile of tasks wondering which one to pick up first. That, more than anything, is what I was looking for when I built this thing.
It turns out the value of an AI planner isn't that it plans better than you. It's that it plans consistently — even on the mornings when you don't have the bandwidth to think clearly about it yourself.
Built inside Notion using a custom AI agent connected to a task database and Google Calendar.
