Unmute Perspectives
AI: Saviour or Destroyer? Make the Case.
A tool that helps someone do more can also change who gets paid, who makes decisions and who carries the risk. “Saviour” and “destroyer” are deliberately bold positions. The interesting part is the reason behind yours.
Which AI, doing what?
A writing assistant, a medical decision system and a synthetic video are different applications. An argument that treats them as one thing is difficult to test. Start by identifying the task and the people affected. Then distinguish what a system can do in a demonstration from what it can do reliably in everyday use. Capability is only part of the question; access, oversight and the distribution of benefits matter too.
The optimistic case.
AI can be an assistant rather than a replacement: helping someone find information, draft a response or learn a task. Research on the introduction of an AI assistant to customer-support work found improvements in productivity, with effects varying between workers. That gives the optimistic case a concrete example, while leaving open what happens in other jobs. [1] The stronger version of this position asks how people can share in the gains and retain enough understanding to check the work.
The sceptical case.
An answer that sounds confident can still be wrong. NIST’s generative-AI risk profile identifies confabulation among the risks organisations need to manage. [2] A fluent interface can make checking feel less necessary precisely when it remains important. Scepticism can also concern power: who controls the system, who may challenge a decision and what happens when the provider changes its terms. These are questions about deployment and accountability, not a claim that every use is harmful.
The challenge to both sides.
Optimists need to explain what should happen when a tool fails and who bears that cost. Sceptics need to distinguish a risk that can be reduced from a use they believe should not happen at all. Neither a striking success nor a spectacular failure settles every application. A measured position may support one use and reject another. That is a more demanding argument than simply changing the label from good to bad.
Define an acceptable trade-off.
Try making your position operational. What error rate would be acceptable for this task? Who should review the output? What should a person be able to opt out of? What evidence would show that the promised benefit reaches users or workers rather than only the supplier? Your comment does not need to predict the entire future. A specific example, a clear standard and a serious response to an objection will give the discussion more to work with.
Make your case.
- Which use of AI best supports your view, and why?
- What decision should always remain accountable to a person?
- What evidence would make you more optimistic—or more cautious?
This is a conversation about reasons, not a popularity contest. There is no automated winner.
Read the context.
Sources support the specific facts above. The questions and interpretations are Unmute’s editorial framing.
- NBER · Generative AI at Work ↗
Research on a customer-support assistant; benefits differed across workers and experience levels.
- NIST · Generative Artificial Intelligence Profile (2024) ↗
A framework for identifying and managing risks, including unreliable generated content.
Interactive preview · Sign-in is simulated. Comments stay on this page only and disappear on reload.
Your customer benefit · Preview
Your opinion belongs here.
You can pin one of your comments in the spotlight. Choose a different comment whenever you like.
Your point of view
AI: progress worth trusting, or a risk worth challenging?
Give us your strongest reason, and the evidence or experience behind it.
Challenge the argument. Respect the person.
Anyone can comment. Continue with Google or email when you’re ready. Your email stays private.
Unmute customers can also pin one comment in the spotlight.
Customer spotlight · Your preview
One active pin. In a live discussion, customer contributions would take turns here.
0 / 1,500 characters
Both choices are optional. In this preview no preference is saved and no email is sent.
On the live site, comments will be reviewed before appearing.