How I Use A.I.

I studied communication with data analytics, and the habit that stuck is building the measurement alongside the thing itself. So here is the dashboard one of my content engines reports into. It says plainly what it has measured, and a zero gets printed as a zero, because pretending is worse than a small number.

A dashboard that admits where it is blind is one you can trust when it shows you something good. Have a click around: the tabs work, and so does the platform rail. The design is the one that really runs; every figure here is invented, for representation only.

Built for a cybersecurity client in Singapore. The challenge: convincing someone who is not even cyber aware that they might need cybersecurity, without getting technical. I could spend hours explaining why your business is vulnerable; you would nod, agree, and move on with your day. (We all know we should wear seat belts. Plenty of us still don't.)

Or… I can drop you into the middle of a cyber attack for a few minutes and let you figure it out for yourself.

Ask ChatGPT a question and it answers by citing a handful of sources. Answer Engine Optimisation and Generative Engine Optimisation are the work of becoming one of those sources: getting a brand cited by A.I. The method is simple to say and slower to do: measure what the engines say about a brand, build the content that changes the answer, then measure again. Try it yourself; run the search below.

Agentic workflows sound complicated, and underneath they all come down to the same three things: a task broken down properly, each A.I. agent given thorough instructions and one narrow job, and a refined sequence for the handoffs. Say you want to publish quickly without losing what makes your content yours. Below is roughly what that looks like.

Content is only the example. The same discipline built the agent that runs my own research and drafting, and it has no authority to send anything on my behalf. I like my A.I. capable, and on a leash.

Generating an HTML page with A.I. is the easy part; a well-designed website is fitted to the requirements of your particular business. The two builds below are both mine, end to end, and they could hardly be more different. Click between them and I will tell you the story behind each.

Bottleat is a hand-built online shop for a friend’s gourmet sauces: three sessions, a generous amount of coffee, and a prefilled WhatsApp link doing the work of a paid payment gateway. He had quit his full-time job to chase what he loves, so every decision was shaped around him: brand visuals built on the man himself, a deliberately simple UI for an older clientele, copy that sounds like him. We built it sharing a laugh and enjoying being alive, at a fraction of what consultation rounds would cost. Making it simple takes twice the amount of thought.

sourishsinghroy.com is designed and built end to end, including the machine-readable layer that decides whether AI engines can quote you. I wanted a portfolio that is a bold depiction of what I can actually do; too many I had seen were a lousy Canva PDF of accomplishments. It began as a rough wireframe in Framer and never stopped evolving, and that steady iteration is what modern digital marketing runs on. It comes naturally to me.