TL;DR - Agents actually save time in real world tasks. The AI agent I used is from a SF based startup called wajo.ai.

My move-in date got bumped up by five days. Just like that, six different logistics melted down at once: home insurance, BC Hydro, the U-Haul truck, a city landfill run, mover schedules, and the elevator booking.

Instead of spending my Friday morning on hold listening to elevator music, I typed up my task list and handed it over to my AI assistant. Ninety minutes later: 12 calls placed, four items fully closed, and two callbacks queued up. Here’s the breakdown of what worked, what bombed, and what I learned.

1. Local numbers crush 1-800 lines every time

Out of 12 calls, four never reached a human. Every single failure followed the exact same script: a centralized 1-800 phone tree with no operator path, or a transfer loop that dumped back to the main menu.

Every single one of those was a national or centralised line. Every call that reached a person on the first attempt was either a local branch or a queue I stopped trying to be clever about.

Case in point: BCAA. The AI tried to reach my specific advisor via name directory. The automated system found her, announced she was busy, offered voicemail, and promptly disconnected the call. Attempt #2: skip the cleverness, enter the generic insurance queue, wait 15 minutes, and talk to whoever answers. A rep at the Burnaby office picked up, messaged my advisor internally, and resolved the issue mid-call.

It also turned out there is no direct extension to reach. Every call routes to that office and whoever is free answers. The directory option I had been trying to use could not have worked.

2. The best win came from an obscure municipal contact

Task 5 was the landfill run: get rid of a dining table, a shoe rack, and a kids' plastic sand enclosure before the move out. If the city couldn't commit to a date, get quotes from private junk removal services.

The city contracts bulk pickup to GFL Environmental. Call #1 got the request logged but no confirmation number. Call #2 (different agent) locked it down: Request #226****, scheduled for Wednesday, Oct 1, curbside by 7:30 AM.

Free pickup, fully scheduled three weeks before move day. This was the highest-value win of the morning, and the task I was most convinced would fail.

3. The hard boundary isn't capability—it's identity

Two tasks hit a brick wall that no amount of prompt engineering can solve: security verification.

BC Hydro needed my birthdate. BCAA needed my policy number. These aren't complex problems—they're just identity checks that no third party should be answering on my behalf anyway.

The takeaway: stop asking "What tasks are too complex for AI?" Start asking "Which tasks require my identity?" Everything up to that line—navigating phone trees, diagnosing root causes, waiting on hold—can be delegated. You only step in for the final 30-second handshake.

The BC Hydro call is a good illustration. By the time it came back to me, the diagnosis was already done: don't lead with the move request number, there's a live account with the wrong date, ask for it to be moved, here's the agent who left notes on the file. That is a two-minute call for me instead of a twenty-minute one.

Where time got wasted

Dead messaging channels

Task 9 was an SMS outreach. It failed four times because my phone number isn't on iMessage, creating a silent structural routing loop. Switching to email took 5 minutes once identified, but diagnosing the channel failure cost real time.

Switching to email took five minutes once I said so, including finding her address buried in a March thread in a mailbox nobody had thought to search. The fix was trivial. Establishing that the fix was needed was not.

I logged a bug with the engineers, hopefully this gets resolved.

Connecting ≠ Solved

My accountant picked up on attempt #2 and said he’d "call me directly later." That’s a connection, but not a resolution. Measure success by closed outcomes, not completed calls.

A call connecting is easy to score as a success. It is worth being strict that the unit of success is the answer, not the conversation.

The round in numbers

The Takeaway

Before you hand off a busywork list, challenge your operational assumptions. Delegating to an AI assistant forces clarity on what actually requires your direct involvement versus what is just an avoidable administrative bottleneck.

The assistant succeeded not by being brilliant, but by executing the tedious phone calls and logistical legwork efficiently.