Do you HAVE to provide Human Customer Service?
12 Aug 2025
5.5 min read
The U.S. Senate is considering a bill that could reshape how companies handle customer service. If passed, businesses would be required to offer customers the option to speak to a human agent located in the U.S. – no matter if their first interaction was with AI or an overseas team.
If passed, the bill would require:
AI agents to clearly identify themselves as AI.
Businesses to give 120 days’ notice before outsourcing customer support.
Agents to immediately disclose their location.
Companies that outsource to be listed publicly for five years – unless jobs return to the U.S.
The goal of the bill is to try and protect customer service jobs. We’re not going to comment on whether the bill will pass or whether it would achieve its aims, if passed.
It appears the bill is aiming at companies who use overseas outsourcing more than those who use AI, but the effect will be similar. If firms suddenly have to pivot to US-based agents, then that will drive costs up, but some of that could be mitigated by using AI more effectively.
The big question is whether customers really want this. The statistics suggest that when asked to choose between humans and AI, they pick humans. But those statistics assume that both things are equal. So in situations where the quality of response or outcome are equal, surely most consumers will choose the path that takes the least time. That means AI.
Taking aim at bad chatbots
We’ve said it before: chatbots suck. Well, most of them do.
This is because many of them use ‘deflection’ as their main metric. But there is good deflection (the customer got their answer), and bad deflection (the customer got so annoyed they left), and most platforms don’t distinguish between them.
Throughout the history of chatbots, they have mostly been glorified search engines. You ask a question, it looks for keywords and tries to return the most appropriate article. You as the consumer then have to trawl through the article looking for the relevant answer.
So bad chatbots exist for one purpose only: to get in the way. They prevent customers speaking to agents because that makes customer service expensive.
The best chatbots are built to try and answer as much as possible, but always to step aside and handover to a human if that is the right thing to do.
Will native-only customer support provide a worse service?
Moving all support onshore may protect jobs, but it also risks cutting after-hours coverage – potentially frustrating customers who expect 24/7 help. Demanding U.S.-based agents will raise the costs for firms, and many won’t accept the rising costs. So the overall experience for the customer is going to go down.
That is, unless firms use AI effectively to pick up the slack given that it can run 24/7.
Making your chatbot useful
So one way to combat the intentions of this bill is to ensure that your AI Agent is actually useful, so people like interacting with it.
The easiest way to start is to put yourself in your customers’ shoes and behave as though you don’t know anything about your brand. Then ask yourself, how easy is it to solve a problem with the chatbot?
The other thing to do is to really examine the metrics you see about your chatbot. If you are looking at deflection rate, are you also measuring whether customers got the answer they wanted? What else can you see within the conversations customers are having with your bot?
One common blocker is that many AI Agents lack the integrations, and particularly the deep integrations to be able to take actions and do anything useful. This means tracking down orders, starting returns, changing order details and so on.
Here’s your mini-checklist:
Test the bot as if you’re a first-time customer.
Measure how many clicks, or how much time it takes to reach the right conclusion.
Measure both deflection and resolution rates.
Add deep integrations for real actions (order lookups, returns, cancellations).
What to do when someone insists on speaking to a human agent
Ultimately you should let customers speak to a human agent if that’s what they want. But, you can still use AI to help them get a faster response.
Imagine a customer insists on speaking to a human. Your AI agent could say something like:
Absolutely, I will pass this case along to one of my human colleagues. They will be online at 9am, which is in 12 hours’ time.
To make the process quicker, could you describe the problem, and I can summarise this for my colleague.
Here you’ve reminded them that it will be a while until a human comes online, but have offered a step that will make things quicker.
If the customer describes the problem, it may then be something that an AI agent can solve. For example:
OK, so I understand that you are trying to track down your order. I may be able to help with that. Could you give me your order number so I can look it up with the carrier?
Suddenly you are using AI in a way that may help the customer, and possibly give them a resolution, and ultimately a good deflection.
If it’s something that is outside the scope of the AI, then you can still use the AI to gather as much information as possible. This could include asking for photographs of damaged products, or proof of purchase, or anything that an agent would need to resolve the case. You can also use image recognition to verify if these photos or documents are acceptable.
Now the human agent has all the details at their fingertips and can solve the problem as quickly as possible, but AI has played an important role.
So …yes, you should always provide human customer service
At DigitalGenius, we have always been clear that you should always provide human customer service on some level. However good AI is at doing customer service, there will always be situations where humans can handle things better, or can unpick complex situations.
Sometimes that is about showing sensitivity or using judgement. AI can be trained to get there, but sometimes there can be a highly-emotionally charged situation that you not want an AI fumbling.
For some brands, it’s important that there is a continuity of care between a human agent and customer, so one agent understands the context that each customer has.
But the brands who will be most successful will be the ones who strike the best balance between humans and AI.
The best CX blends:
Access – Customers can reach a human when they need to.
Automation – AI handles quick, repetitive, or simple tasks.
Assistance – AI supports human agents with context and prep work.
