Reposting About Personal Agents + Coloring Book Update

I still believe the technology to have robots make calls for you is coming soon

I’m on vacation this week, so I’m reposting an old (but still relevant post). As NBC used to say in the ‘90s, if you haven’t seen it, it’s new to you.

Before I get to that, I wanted to provide a brief update from last week’s post about creating an AI coloring book. I got an outpouring of responses from readers, ranging from commiserating over their own challenges with AI-generated code to one person who printed out the steampunk owl and gave it to their child to color with. I am still trying to get my book on Amazon, but also still running into technical issues despite ChatGPT assuring me that one attempt was the “Third (and Final)” and another one was “Bullet Proof.” I’m still working and hope to get up next week.

Finally, next week I’ll be doing another mailbag column. I’ll do a post on Monday to collect questions, but if you have any burning questions for me, send them to me or post them below.

Now, here’s an older post that’s still worth reading:

At OpenAI’s developer event last November in London, they did a cool demo in which a user asks ChatGPT to order pies. (They then proceeded to give everyone in the audience some savory pie.) This demo foreshadows a powerful new use case: AI voice agents who work on our behalf, handling everyday tasks and eliminating some of life’s most frustrating interactions. When this becomes widely available, how will businesses respond? And how should investors think about companies adapting to a world where customer interactions are increasingly AI-driven from both sides?

In this post, we’ll explore what this future could look like, the challenges it poses for companies, and the potential impact on profitability and customer experience.

What Could Voice Agents Do For You?

Picture any moment in the last few months where you’ve been on hold. Now imagine that you never have to do that again. (I’m guessing 100% of you are smiling right now.) Instead, you can send your personal bot to do the call for you. In fact, you may have already used an early version of this feature. If you use Google to make a dinner reservation, one way that they do it is to make an automated call to a restaurant.

What kinds of activities could these agents do:

• Arguing with Insurance: Instead of battling with an agent over a denied claim, your AI could gather the necessary information, provide evidence, and negotiate on your behalf

• Making Appointments: Booking doctor’s appointments or finding a plumber would be as simple as telling your AI to do it. The AI would call around, coordinate with your availability and put it on your calendar

• Negotiating Bills: Have you ever tried to lower your phone or internet bill? Now imagine telling your AI, “Get me a better rate,” and it handles the back-and-forth with your provider

• Modifying An Order: From jeans to concert tickets, your AI could take your call to handle follow-ups like returns or changing the date of an event

When you think about all the routine calls and emails that could be offloaded to a smart assistant, it’s easy to see the appeal. So, how long will it take for people to start using this?

Thinking through the adoption curve

If an AI Voice Agent app became available today, how long would it take before it’s ubiquitous? One good analogy is online restaurant ordering. The first major platforms like Seamless and Grubhub launched in 2004, and by the late 2000s 20-30% of pizza orders were online.¹

Restaurant reservations are similar with OpenTable having made 1M reservations by 2002 and 70M by 2008.² ³

So, let’s say it might take 3-5 years before a critical mass of customers adopt this technology.

But what happens when this becomes the norm? How do companies respond when their people are no longer talking to actual customers but rather to their bot representatives?

How Will Companies Adapt To Inbound Robocalls?

I’ve spoken with multiple firms and investors who are grappling with this question already. The potential for AI-to-AI communication raises several fundamental questions for businesses, with positive and negative implications for investors:

1. Do Companies Still Need People to Speak with Bots? 
Obviously, customer service departments are staffed by humans trained to communicate with other humans. But if they’re now speaking to AI agents, do you need humans on the other end? Could an AI voice bot speak to the customer’s bot on behalf of the company instead? This approach could be more cost-effective and consistent, improving margins by lowering the cost to serve. For investors, this could mean increased profitability, particularly in service-heavy sectors.  But, that won’t happen right away. It’s easier for a bot to argue that an insurer should pay my doctor’s bill than for the bot on the other end to decide whether to pay it.

2. Bot-to-Bot Interactions: Cheaper Than The Alternative 
When two AI agents are tasked with handling both sides of a “customer” interaction, is a voice call really the most efficient way to do it? AI agents could communicate directly through a back-end system, bypassing the limitations and costs of spoken language and speeding up the transaction. For companies, this could lead to lower operational costs. However, while investors might welcome the potential for cost savings, they’ll also need to consider how companies stay engaged with customers.

3. The Risk of Losing Customer Touchpoints 
One of the primary challenges with bot-led interactions is that they remove a direct opportunity to talk to the customer. When companies replace human interactions with AI agents, they potentially lose the opportunity to build relationships, cross-sell, and upsell. While this may streamline operations and boost short-term margins, it could have longer-term consequences if customers feel less connected to the brand. If humans embrace AI agents, then how do companies talk to them?

4. Negotiating With the Best 
The savings from reducing the number of human agents on the other side may be necessary because the bots could be good negotiators which will impact margins. First of all, as these agents become more widespread, people will send them to complain about more things. Today some people might let a minor overcharge or unexplained fee go by unchecked to avoid hours on the phone, but once they just have to send their AI agent after it, customers will be more likely to pursue these opportunities. Secondly, the bots will probably be better negotiators than the average human, learning from tens or hundreds of thousands of interactions which techniques are most effective. Companies will likely have to give some revenue back to their customers, negatively impacting margins.

5. Another Alignment Challenge
Going back to my post about regulating LLMs, if this becomes popular, LLMs or the companies that build these agent applications will need to add some new guardrails. They need to prevent the bots from engaging in deceptive behavior in negotiating. For example, when calling to try to reduce the phone bill, bots can’t pretend that the customer is very ill or desperately needs money or something. This will be an important engineering challenge to get right for this whole model to be possible.

Conclusion

As AI voice agents become more capable and more widely used, we’ll likely see businesses across sectors experiment with different approaches. It will take a while for this technology to become widespread. Some may choose to keep human agents in the loop, providing a necessary human touch for more complex or sensitive issues, while others may move aggressively toward AI-first customer service models. In all cases, businesses will need to remain adaptable as AI technology—and consumer expectations—continue to evolve.

My recommendation for companies is that when bot calls are 1-2% of volume, that’s a good time to start experimenting with how to handle them effectively. That could mean piloting approaches by early 2026, but better to do that than be caught by surprise by an army of polite but firm bots asking for a discount.

Note: The opinions expressed in this article are my own and do not represent the views of Bain & Company.

1. https://adage.com/article/digital/pizza-giants-customers-click-call-delivery/136087

2. https://ecommons.cornell.edu/server/api/core/bitstreams/2d775832-a1a6-4343-bb51-24d9db56b68d/content

3. I used Perplexity to find these obscure 20 year old statistics in a few minutes. We really live in the future.