Building a consumer brand with AI
Three years building a brand without knowing how. How AI changed marketing, compliance, creative, and operations, and where it goes next.
👋 Hi, I'm Shantanu. I run Xandro Lab, a longevity science brand based in Singapore. Every Sunday I write these notes - part building diary, part thinking out loud. Sometimes it is marketing or operations. Sometimes it is health and performance.
Three years ago, I did not know how to run a company. I had spent years at Myntra running revenue operations across large business units, but that was a big organisation with specialists for everything. Here, it was just me, and I had to be all of those specialists at once.
I did not know how to write marketing copy. I did not know how to make an e-commerce website. I had built websites as college projects, but nothing like this. I did not know compliance or regulatory requirements. I did not know how to make a landing page that actually converted, or how to put together a presentation that looked professional, or how to brief a designer when I did not have a clear picture in my head.
And I was supposed to do all of this while also building a product, finding suppliers, managing cash flow, and figuring out who our customers actually were.
The timelines made it worse. Everything came with minimum 2-3 week estimate. A landing page revision: two weeks. A product description rewrite: one week. A compliance check for a new ingredient: 2-3 weeks. I remember doing the math early on and thinking: at this pace, in two years we will still be in the same place.
I am going to say something plainly. A significant part of why Xandro has gotten to where it is today is because of AI. It allowed me to punch above what my skill set and budget would have allowed. If AI was not there, I think I would have failed much earlier. The execution would have been impossible.
That is an uncomfortable thing to say. But I think it is the most useful thing I can tell you.
Today’s reading -
How I stopped spending 8 hours writing a single blog post
Regulatory work, compliance, and market entry
25 landing pages in one month, versus a few pages in 3 years
From expensive shoots and stock images to building a film series
From watching 200 session recordings to reading a summary in an hour
Agents, automation, and AI that acts without being asked
AI is not replacing the team. But it is changing what matters.
#1 The beginning: ChatGPT and the basics
I started using ChatGPT in late 2023 for the most basic things. Drafting copy so I was not starting from a blank page. Rewriting product descriptions in a different tone. Grammar checks. Getting a first version of anything written quickly.
Even that changed the pace of work immediately. Tasks that used to take me three hours started taking 15 minutes. The blank page problem, which is paralyzing, disappeared almost overnight.
The blog posts are the clearest example I have.
I used to spend close to seven or eight hours on each one. Sometimes more. First figuring out what to write about, which itself would take a long time. Then writing it word by word, sentence by sentence. Then revising. Then revising again. By Sunday evening I had written one post and done nothing else with my day.
Now I can write at 9:30 PM, after a full day of events. Throughout the day I drop voice notes into my phone. A thought during the run. Ideas while talking to someone. By the time I sit down at night, I have most of the raw material. Then I structure it, refine it, and the actual writing will take maybe 1-2 hours instead of a full day.
If the post is not good, it is because the thinking was not strong enough, not because I ran out of time. That is a completely different problem to have.
#2 How it spread beyond marketing
What I did not expect was how quickly AI stopped being just a marketing tool and started running through everything.
Regulatory work is the example I keep coming back to. Supplement compliance is one of the most specialist-dependent parts of this business. Ingredient approvals, label formats, market-specific restrictions, supplement fact tables. Every market has different rules and getting it wrong has consequences.
When we started working with our Korean manufacturing partner on complex formulations, including a methylated multivitamin and an LXP weight management formula, we were dealing with the most complicated supplement fact tables I have ever seen. These were not simple capsule blends. They were vitamin pre-mixes compressed into tablets, and traditional Korean herbs that were not commonly used in Singapore, in dosing structures I had not worked with before.
The regulatory team was stressed. I was stressed. The format was completely new to all of us. AI made the back and forth manageable. We could check formats, flag potential issues, understand ingredient interactions, translate complex base formulas into something that could pass review. Not perfectly, and not in one pass. But at a speed that would have been impossible if we were waiting for consultants at each step.
#3 The landing page moment
Two years ago, building a landing page was a production. You needed copy, then a Figma design, then a developer to turn it into an actual webpage, and the whole thing took weeks. We tried. We produced only a few - it took me several months to get the website where it is today.
However, in the last month I have generated close to 25 landing pages. Every one talks to a different angle on Protocol X. Some speak to doctors. Some to athletes. Some to recovery. Some to cognitive performance. Different entry points, different hooks, same product. And all it took was a few hours of notes and directions. I got ready made HTML pages. Let me share some examples below (these are not live yet).
Protocol X currently has one public-facing page. We are about to release around 20 different angle-specific versions of it.
That volume was not possible before. The bottleneck is now thinking, not production. If I can be clear about the angle and the audience, the execution follows fast. That is a completely different constraint to be working against.
Last week there was a sudden need for an updated investor presentation. It took less than 6 hours to finish. I knew the story, I just needed it shaped properly. I do not have strong presentation design skills. AI handled the design. I handled the story. We started at 80% of the way there and I spent the remaining time getting the voice right.
That 80/20 idea has become how I think about everything now. I am not starting from zero. I am starting from a draft, and the real work is the last 20%.
#4 What is happening on the creative side
This one surprised me more than I expected.
We used to either shoot everything or use stock images. Shoots are expensive and slow. Stock images are generic and everyone recognises them. Today if I need a product visual with a specific mood or background, I take a good base photo, run it through an AI tool, and rebuild the environment around it. If I want to build something from a moodboard, that is now possible without a full production team.
We are working on two major content workstreams right now. One is high-volume TikTok content, with a target of around 50 videos a week. The other is a project that I care about a lot, which is building an AI universe around longevity. A film series - imagining what it really looks and feels like to age powerfully.
What would have required tens of thousands of dollars in production can now be done with creative direction, the right tools, and iteration. The barrier is no longer money. It is vision.
#5 Understanding what is actually working
One area where AI has saved a quiet but significant amount of time is data.
I used to watch session recordings on Microsoft Clarity to understand what was happening on the website. If someone spent ten minutes on a page, I would watch all ten minutes. Drop-off points, scroll depth, rage clicks, abandonment moments. It was thorough but completely unsustainable. 200 sessions would have taken me days.
Clarity now summarises all of it. I can understand what is breaking, what is confusing, where users are dropping off, in a fraction of the time. 200 sessions in under an hour. That changes how fast I can find the problems and how clearly I can brief the team on what to fix.
The same shift is happening with email performance. Running analysis faster, generating hypotheses about what is underperforming, coming to the team with a brief rather than a pile of raw data. From analysis to new copy generation is a maybe an 15 min work. In fact, I could have ready made new emails every morning based on performance of individual campaigns.
#6 Where this is going
Everything I have described so far is still me using AI as a very powerful assistant. The next phase is different.
I am thinking about agents. Inventory forecasting that runs on a schedule. Weekly performance summaries that arrive already structured. For simple operational problems like incorrect pricing, direct action without me needing to initiate it. The logic exists. I have not fully built it yet. But it is the direction.
I also think we are not far from website experiences that personalise in real time based on what a visitor is doing. A lot of what currently requires manual testing and campaign iteration will start happening automatically. The front end of the business will start making more decisions on its own.
AI is moving from telling you what to do, to doing it.
#7 What it means for the people around me
I want to be honest about this because I think it is easy to get it wrong in both directions. AI is not replacing the team. The work requires judgment, relationships, creative direction, and the ability to make decisions under uncertainty. None of that is changing.
But what matters has shifted. The people who are most valuable now are the ones who can direct AI effectively. The ones who can write a brief that gets 80% of a landing page done in one pass. The ones who can look at a data summary and know what question to ask next. The ones who can take an AI-generated first draft and make it actually sound like the brand.
I call it - targeted, high volume, quality outputs.
That is a different skill set from three years ago. I think about this when I hire and when I design how we work.
P.S I started Xandro not knowing how to do most of what running a company requires. I am still not an expert in most of it. I am a generalist who has learned to move fast by knowing which tools to use and when.
AI gave me access to a level of execution that my budget and skill set alone would not have reached. The playing field has shifted, and I think it is shifting in the direction of people who are willing to learn fast and direct clearly rather than just the ones with the most money and the most specialists.
The tools that exist in two years will make what I am describing here look like the beginning. I am moving as fast as I can to stay ahead of it.
Until next Sunday,
Shan