Stop Using AI as a Search Engine

Most owners use AI the way they use a search engine: type something vague, skim the generic answer, conclude the tools are overhyped. The owners seeing real returns run it differently — as our co-founder Keshav Mohabir put it at an investor-day talk, treat it like “a team member who needs clear direction.” Four strategies, field-tested.

1. Brief it like a junior teammate

Bad prompts share a shape: general, broad, and with no clear outcome attached. “Create a marketing plan” produces a plan for nobody in particular. Picture instead a brilliant graduate — sharp, well-read, and completely new to your business. Told to “handle marketing,” they hand back generic, directionless work. Given client profiles, objectives and constraints, they hand back work that moves numbers. The model is the same; the management is not.

• Name the outcome first. If you cannot say what “done” looks like, neither can the machine.
• Brief it like a person. Context, audience, constraints, examples of good — the same packet you would give a new hire.
• Kill the guesswork. Specific beats clever, every time.

2. Make it ask before it answers

The cheapest upgrade to any prompt is a single instruction: ask me clarifying questions before you produce anything. Answer those questions properly — paste in the brand document, the winning examples, the half-finished draft — and the output snaps into focus. Instead of “help with social media,” hand over the brand voice and ask for a set number of LinkedIn posts for a specific launch, aimed at a specific reader, building on drafts you already have.

The order matters: you remain the strategist. The system exists to sharpen and extend what you provide, not to think in your place.

3. Hand it the operations, keep the strategy

Worked this way, the leverage compounds. Keshav shipped a beach-conditions ranking app in three days and a wildfire-alert system in a week — scopes that traditionally cost a development team months. The step beyond prompting is agents: systems that run whole workflows in the background — a social campaign end to end, timesheets becoming invoices — with a human supervising exceptions. None of this is about replacing people. It is about clearing repetitive load so leadership hours go where only leadership can: strategy, partnerships, scale.

• List your most time-consuming manual processes.
• Move the data collection and analysis across first.
• Reinvest the recovered hours in decisions only you can make.

4. Practise on your own life first

Leaders who demand business results from AI on day one skip the apprenticeship. Fluency comes from personal, low-stakes use: meal plans on a real budget, research on things you actually care about. For Keshav, the practice turned serious. When his wife’s labour hit complications, he used AI to research the scenarios ahead — timelines, C-section procedure, what happens when a newborn is not breathing at delivery — feeding it full context and demanding specifics. When exactly that emergency arrived, he had already rehearsed it: he stayed calm, supported his wife, and let the medical team work. Their son was born healthy.

The skill that got him through that night is the same one that ships applications in days: complete context, precise questions, acting on the answer. Learn it where the stakes are yours; apply it where the stakes are the business.

The gap is leadership, not technology

You would not expect a new hire to deliver exceptional work without direction, and the machine is no different. Your results will match the clarity of your instructions. So: which repetitive task do you hand over this week — and what context would the system need to do it properly?

Adapted from an article first published on LinkedIn.