What Happens When You Can Delegate the Hassle?
By Ramon B. Nuéz Jr.
I had a delayed Amazon order and an AT&T bill I wanted to lower. I don't enjoy dealing with either situation. They take time, and with AT&T, I also wondered whether I was getting the best deal or being steered toward the plan the company wanted me to choose.
I was going to do what I usually do: call customer service. Then I thought, could OpenAI's Dot help me?
My expectation was that Dot would advise me on what to say. I would carry the conversation, and it would help me get a better outcome.
The experiment started with AT&T. This is the opening exchange, edited for clarity:
Me: I'm logged into my AT&T account. Take a look and let me know where you might find savings. Please don't take any action yet. It would be useful to chat with customer service about lowering my bill, but for now, just review the options.
Dot: Yes. Log into AT&T in Chrome and tell me when you're ready. I'll review the bill line by line for plan costs, device installments, insurance, add-ons, and available discounts. Then I'll show you the savings options before changing anything.
Me: I'm already logged in. Take a look and let me know where I could save. Please don't take any action yet.
Then I asked whether Dot could negotiate for me through text chat. It said yes.
Watching the exchange was fascinating. The AT&T representative asked questions, made suggestions, and verified information. Dot asked its own questions, discussed the options, and confirmed assumptions. Meanwhile, I was talking with Dot in ChatGPT, asking questions and requesting clarification.
There were two conversations happening at once. Dot carried the exchange with AT&T, and I directed Dot.
I remember the representative commenting that I was the most polite person they had spoken with that day. That amused me, given that Dot was handling the support conversation.
Dot reviewed plans, usage, credits, and payment options, then worked through the details with the support representative.
I initially authorized research only. Later, I approved changes to three lines while leaving another line unchanged. I also wanted to keep paying by credit card and protect existing device promotions. A lower bill that lost those benefits could have missed the point.
These were available plan options. The useful work was finding them, asking about the conditions, and checking whether they fit what I wanted.
By the end, I had approved plan changes that were quoted to lower my monthly bill. The changes were scheduled for November 2, so I still need to confirm the savings in the final billing details.
That felt amazing. I had expected help preparing what to say. Instead, I had help carrying out the conversation.
Could Dot handle the Amazon issue, too?
I also wondered how Dot would handle the problem with my Amazon order. This task had a different goal: resolve a delayed order and arrange a refund.
The situation was simpler than the AT&T review. An order for toothpaste and dog shampoo was delayed. An attempt to cancel it had failed, and I had placed a replacement order. I wanted to resolve the original order without paying twice or disturbing the replacement.
Dot explained the situation in Amazon's support chat. The representative recorded a refund to my original card, expected within five to seven business days. The replacement order stayed untouched.
Both exchanges happened through website text chat. No phone call was involved.
I explained the problems, set limits, and approved changes. Dot handled the back-and-forth with support. I remained involved, but I could ask Dot for an explanation rather than manage every message myself.
More room for what interests me
I keep coming back to S. B. Divya's Machinehood. What stays with me is Welga's relationship with her personal AI assistant: she gives it a goal, and the goal is achieved. The appeal is being able to delegate a task and continue with her life. 1
My customer-service experience was more hands-on. I was watching, asking questions, and learning what Dot could do. I can't put a number on the time it saved. What I experienced was a shift in my role: I could explain what I wanted and evaluate the options while the agent carried the conversation.
That makes me curious about what happens as I become more comfortable delegating other tasks.
Making an appointment, reviewing a recurring bill, and following up on an order are all things I can do myself. They are also things I would often rather hand off. Each seems small until it joins everything else waiting for attention.
If an assistant can carry more of that work with less supervision, it could give me room to write, build something, or follow a question I'm curious about. It could also make it easier to deal with something I've been putting off.
That's the kind of augmentation that interests me: help doing the work around my interests and obligations, so I have more capacity to pursue them.
How far could this go?
Could a personal agent help someone find a better job? It might research employers, compare offers, prepare questions, or follow up. Could it eventually handle part of a salary negotiation? Possibly. I haven't tested that. A successful customer-service chat gives me a reason to explore the idea, but it doesn't establish that the agent can handle it.
Someone may be good at their work and find the research, paperwork, and conversations around changing jobs difficult. Help with those tasks could make an opportunity easier to pursue.
The judgment becomes more consequential, too. A higher salary might come with less flexibility or less time off. An agent can compare the terms and ask questions. I still have to decide what another week with my family is worth to me.
My AT&T experience contained a smaller version of that decision. I had to define an acceptable result, including the benefits I wanted to keep. I also had to decide when Dot could move from research to action.
Approving every message would pull me back into the work. Giving it broad permission to get the best deal would leave too much undefined. A useful arrangement needs a clear goal, meaningful limits, and a way to bring decisions back to me when those limits aren't enough.
It also needs a final check. A refund recorded in a support chat still has to reach my account. A quoted reduction still has to appear on a bill.
Who gets that extra capacity?
This brings me back to the access question in my previous article. If a more expensive AI service gives someone more capacity to research, build, and follow through, could that help them create a better business or pursue more opportunities?
I think it could. How much depends on whether the assistance is reliable and how much effort it takes to use.
Cost is part of that question. So is the attention required to use the tool. Access alone doesn't tell me how much someone can delegate or how much checking the work will require.
For now, I have two ordinary customer-service exchanges to work from. They showed me that I could hand over meaningful parts of a task while keeping a say in the decisions.
The next thing I want to understand is how much direction that takes as I try other tasks. If I can spend less attention managing the work, I want to see what that leaves room for.
References
- S. B. Divya, Machinehood. Official publisher page. The assistant connection above is my reading of the novel.
