Outsource Automation, Not Intelligence

By Dr. Mayank Gupta, Partner, Collaborate and Eight

Does your AI usage save time; or surrender judgement?

If you have ever asked yourself that, here’s a quick test.

How did you choose a restaurant last time you went out to dine? Did you ask a friend? Or did you open an app, scan the star rating, and book the first 4.5-star spot that popped up?

It’s a safe assumption that you opened the app. It’s instantaneous, easy and most times, it works.

But sometimes it doesn’t. You walk in. The vibe doesn’t match the photos and the food, hmmm.. average! And you think: How did this get 4.5 stars?

And maybe then you realise, you didn’t just outsource a decision, but also the research for the decision.

And that’s the exact same trap we’re falling into with AI at work.

Did we underestimate a cognitive experience in 2010?

In 2010, cognitive scientist Doug Hofstadter while using Google Translate to translate his own writing, noticed something: the tool kept offering suggestions and alternatives that were grammatically correct. He found the translations inappropriate for his endeavour. They were more inferences rather than literal translations - the ones he needed.

He was angry at the suggestions, and made this known to one of his student in a personal correspondence. He wanted machines to do what they are programmed to do, not spurt out suggestions when uncalled for. But he stayed in control. He used Google Translate as a translator - what its sole  purpose is -; not a decision-maker.

Fast-forward to 2026. 

We ask AI agents to summarise reports, draft client emails, analyse competitors, even recommend hiring filters. And when the output looks polished, we nod and move on. In this, when we forget to ask ourselves - did we really need that? 

That's how we end up outsourcing thinking along with automation.

The Restaurant Review Trap - once again, now at Work

Let’s stay with the restaurant example, because it reveals something important about how we are making choices today: hurriedly. 

In 2010, you might have called a friend and asked: “Is that new Italian place you mentioned any good?”

Your friend would have said: “Depends. Are you going for a date or a work dinner? Pizza or Pasta”

That’s intelligence -contextual, tailored. It asks follow-up questions and makes you brainstorm.

Today, you scroll through 200 reviews, glance at the average score, and most of us pick the highest number. That’s automation. It’s efficient, but it strips away nuances. 

And it comes with an ironic twist: we are learning to spot which reviews are actually human by looking for flaw - typos, over-the-top rants, messy grammar, no em dash please. Polished, perfect language now feels suspicious! We are inverting our standards just to prove we’re reading real people.

Now apply that to your business.

When your team uses AI to draft a proposal, do they treat it as a first pass—or a final answer?

When they summarise a competitor’s annual report, do they check the source data—or trust the bullet points?

When they use AI to screen candidates, do they review the shortlist with fresh eyes—or forward it to HR without question?

If you’re doing the second, you’re reading the 4.5-star review and skipping the actual menu.

What to Do Instead: Three Questions Before Every AI Prompt

The solution isn’t to stop using AI. That’s unrealistic and unnecessary.

The solution is to stay in control - the way Hofstadter did with Google Translate in 2010. By defining the problem and your expectations beforehand. By not letting AI lead you astray. 

Next time you or your team open an AI tool, pause and ask:

1: Do I know what I actually want?

If you can’t articulate the outcome before you prompt, you’re letting the AI define your goal. That’s backwards. Write down your objective first.

2: Am I being challenged while reviewing this output?

If the response feels smooth, obvious, and frictionless, that’s a warning sign. Good AI output should make you think, not just nod.

3: Would I defend this output if challenged?

Imagine your boss or a client asks: “Why did you recommend this?” If you can’t explain the logic behind the AI’s suggestion - because you don’t know where the data came from or the assumptions behind it - then you’ve outsourced more than automation. You’ve outsourced accountability.

4: Answer “no” to any of the above three, and you’re not using AI. AI is using you.

A Practical Reframe

Automation is: “Draft this email, and I’ll edit it.” Intelligence is: “Draft this email, and I’ll decide if it’s right.”

Automation is: “Summarise this report, and I’ll verify the key numbers.” Intelligence is: “Summarise this report, and I’ll interpret what it means for our strategy.”

Automation saves time. Intelligence saves judgment. And in a world where everyone has access to the same tools, judgment is your only real competitive advantage.

Over to You

We run workshops on exactly this: helping teams use AI without losing their edge. It’s not about banning AI agents - they are part of the future, if not yours then probably your competitors. We help you steer through this difficult terrain comfortably from the driver’s seat.

If that sounds useful - or if you’ve got a story about when AI led you (or your team) astray - drop a comment or DM me. No pitch. Just a conversation.

Because honestly? Let AI generate, and people decide. 

Let’s talk more.