---
title: "AI Says Yes"
author: "Mani Sandher"
published: "2026-06-10T12:47:14.000+00:00"
modified: "2026-06-10T12:47:45.000+00:00"
source: "https://www.linkedin.com/pulse/ai-says-yes-mani-sandher-ofkle"
cover_image: "https://media.licdn.com/dms/image/v2/D4E12AQH3Y30hf8Kucw/article-cover_image-shrink_720_1280/B4EZ6xX23GIkAY-/0/1781092286367?e=2147483647&v=beta&t=p_50Uow2M-5zPPzEjxDyhJQSRwfmra23_eQoKrSybZk"
---

# AI Says Yes

Source: [https://www.linkedin.com/pulse/ai-says-yes-mani-sandher-ofkle](https://www.linkedin.com/pulse/ai-says-yes-mani-sandher-ofkle)

Published: 2026-06-10T12:47:14.000+00:00

![Cover image](https://media.licdn.com/dms/image/v2/D4E12AQH3Y30hf8Kucw/article-cover_image-shrink_720_1280/B4EZ6xX23GIkAY-/0/1781092286367?e=2147483647&v=beta&t=p_50Uow2M-5zPPzEjxDyhJQSRwfmra23_eQoKrSybZk)

In the comedy series *Little Britain*, Carol Beer is a customer-service worker who types even the most reasonable enquiry into her computer, waits for the answer, and then replies with the now-famous line: "Computer says no."

It was funny because the computer was clearly not the problem. The problem was the human being who had quietly outsourced all judgment to the machine.

Fast forward, and we now have: "AI says yes." Yes, I can write that report. Yes, I can analyse those customers. Yes, I can screen those CVs. Yes, I can update the system. Yes, I can send the email.

Wonderful. Progress. The machine is being helpful.

And much of the time, it really is. AI can save time, reduce drudgery, improve first drafts, widen the search space, spot patterns, summarise complexity, and help people get unstuck. It can be a genuinely useful partner in thought and work.

But "yes" to what exactly?

That is the question that gets lost when AI feels effortless. A helpful answer arrives in seconds. A tidy summary appears. A recommendation sounds reasonable. A draft looks more polished than anything the team would have produced in the first hour. The temptation is to treat fluency as reassurance.

But fluency is not the same as judgment. Speed is not the same as accountability. A confident answer is not the same as a considered one.

That question matters more as AI becomes more agentic. A chatbot that says yes may produce a paragraph. An AI agent that says yes may take steps in a workflow. It may gather information, draw on memory, call tools, update systems, draft messages, trigger processes, or keep working beyond the first response.

In other words, "yes" is no longer just an answer. It can become action.

That is where the old joke flips. Carol Beer was funny because the computer blocked reasonable requests. With AI agents, the risk may be that the system accepts too much, too smoothly, too confidently, and with too little friction.

This does not mean leaders should become fearful or defensive about AI. Quite the opposite. But as AI becomes more capable, leaders and teams need to become more capable too: not necessarily more technical, but more disciplined in how they evaluate AI-supported work.

What assumptions is this answer resting on? What evidence supports it? What has been left out? What would change my mind? Where are the limits? What are the consequences if this is wrong? Is this a draft, a recommendation, a decision, or an action? Does this need human review before it moves any further?

These are not anti-AI questions. They are pro-judgment questions.

They are also very human questions. The sort of questions good leaders, consultants, teachers, coaches, lawyers, doctors, engineers, and experienced professionals ask almost automatically when something consequential is on the table. AI does not remove the need for that discipline. If anything, it raises the standard, because the work can now move faster than our normal habits of scrutiny.

Human accountability depends not just on keeping a human nominally "in the loop", but on preserving the human capacity to understand, challenge, and own the work. If the human is present but cannot evaluate what happened, cannot question the assumptions, cannot see the limits, and cannot meaningfully intervene, then the loop may be more theatre than responsibility.

This is what I call the "handoff-of-trust" problem. AI becomes powerful enough to help, but smooth enough that human judgment can quietly slip out of the loop. Work still happens. Outputs still arrive. Things move faster. But somewhere along the way, the accountable human may no longer be able to say, with confidence: I understand this, I have challenged it, and I am willing to own it.

So the question is not only: can AI do the work? The better question is: can we still answer for the work once it has?

This is exactly the kind of capability AIchemy is designed to develop: not just using AI more fluently, but thinking with AI more responsibly.