In the middle of last year we did something that, in retrospect, was the most expensive thing the studio has ever done: we cancelled all of our retainer engagements and gave ourselves six months to ship one product of our own. The product was JuuLinkAI. The six months became eight. The bank account became thin. Here is an honest account of what those eight months actually looked like.
Why we did it
Client work, when it goes well, is a perfectly good business. You scope a thing, you do the thing, you invoice for the thing, you do another thing. It is predictable. It pays the rent. We had been doing it for years and we were grateful for it.
It also has two failure modes that compound over time. The first is that you are always renting someone else's strategy. Even when the relationship is good and the client is a friend, you are paid to execute on a vision you did not author. The second is that the skills you build are linear: you get faster at delivering what people ask for. You get no faster at deciding what to deliver.
We wanted to learn the second skill. The only honest way to do that is to be on the hook for both halves of the work — to decide and to deliver, with no one else to blame. So we stopped saying yes.
The first month: pretending we knew what we were doing
We had a sketch of the product on a piece of paper. Link-in-bio for Malaysian small businesses, with AI-generated reviews. That was the whole brief. Reading it now, it is striking how much we did not know — what the merchant actually needed, what the AI feature would be used for, whether the pricing would clear hosting cost.
We spent the first month building. This was, in hindsight, a mistake. We should have spent it talking to merchants. We had the engineering muscle, so we built — three weeks of clean code with no audience.
The lesson we now repeat to ourselves: do not start with the thing you know how to do. Start with the thing you do not know how to do. Building was the easy part. Understanding the customer was the hard part. Doing the hard part second is a way of postponing the failure.
The second month: the first painful conversations
In month two, we left the studio. We walked into coffee shops, bubble-tea kiosks, tailor shops, kopitiams, and asked the owners if they would talk to us about how they used the internet for their business.
Most said no. The ones who said yes were generous beyond what we had any right to expect. We learned things in those conversations that no amount of internal whiteboarding would have produced. A few examples:
- Almost every merchant we spoke to used WhatsApp Business as their primary customer channel. Instagram was secondary. A few used TikTok. Almost none used a website.
- The thing they wanted help with was not posting more. It waslooking professional on the platforms they were already on. The metric they cared about was "does my profile look like a real shop?"
- The reviews problem we thought we were solving — "not enough social proof" — was real, but the merchants framed it differently. They did not want fake reviews. They wanted help translating the genuine things customers said in conversation into the kind of phrasing that read well on Google.
That last insight changed the product. We had been planning to generate reviews from a prompt. We pivoted to generating reviews from a description of the business plus a list of things the merchant could remember real customers saying. The output is the same shape, but the input is anchored in reality. The merchants responded much better to that framing.
The third and fourth months: the rebuild
We threw away most of what we had built in month one. This was painful — we had been proud of the architecture — but the architecture had been pointing at a product no one was asking for.
The rebuild was faster because we now knew what mattered. We cut features. We cut the analytics dashboard. We cut the multi-user team accounts. We cut the "custom domain" feature that we had thought was differentiating and that not a single merchant had mentioned. The product got smaller and, paradoxically, more useful.
Cutting features feels like loss until you ship. Then it feels like clarity.
The fifth month: the launch that wasn't a launch
We had planned a launch. We had drafted a Product Hunt post. We had a queue of tweets. We had a press email going out to a few Malaysian tech blogs.
We did none of it. Instead, we walked back into the same coffee shops we had interviewed in month two and offered our friends free accounts. Eight merchants signed up that week. Two of them invited their merchant friends. Within three weeks, we had thirty active users, all from word of mouth.
The instinct to do a big launch is, we now think, mostly vanity. For a small product with a small audience, the best launch is a slow accumulation of users who actually use the thing. The Product Hunt post would have produced a spike of free-tier signups and nothing else.
The sixth, seventh, eighth months: the unglamorous middle
Most of the work, after launch, was unglamorous. Fixing bugs. Translating the interface. Negotiating with our cloud provider about an unexpectedly large invoice. Responding to merchant WhatsApps at 9pm on a Sunday. Writing the privacy policy. Filing a business registration update. Talking to our accountant about SST.
There is no chapter of a startup book about this stuff. It is where most of the time actually goes. It is also, we think now, where most of the learning lives. You cannot understand a product until you have answered the support emails. The support emails are the product, told in reverse.
What we know now
We are still very small. JuuLinkAI is still very small. We are not about to write a victory lap post. But there are a few things we believe now that we did not believe eight months ago:
- Client work is not the enemy of product work. We have started taking client engagements again, selectively. The income from client work is what funded the product work. They coexist if you let them.
- You cannot research your way to a product. But you cannot build your way to one either. The actual process is alternating — a week of building, a week of listening, a week of building, a week of listening. The cadence is the work.
- The first version should embarrass you a little.We waited too long to show people. If we had shown them in week two, we would have got to the right product faster.
- The bank account is a tool, not a scoreboard.We made the mistake, early, of measuring the studio's health by the bank account. The bank account is downstream of the work. Fix the work, and the bank account fixes itself — slowly.
A note to other studios thinking about this
If you are running a service business and you are considering making the switch — do it, but plan for it to take twice as long and cost twice as much as you expect. Keep at least one paying client through the transition. Do not announce the switch publicly; just start. The announcement is a way of forcing yourself, and forcing yourself is rarely a good substitute for wanting to.
Eight months in, the studio is still here. The product is small but real. We are answering this question — "was it worth it?" — with a tentative yes. Ask us again in a year.