Get Closer to the Pain
The case for bespoke intelligence over off-the-shelf platforms
Tony Fadell, the designer behind the iPod and Nest thermostat, has a principle: ‘Get closer to the pain’. Don’t observe from a distance. Don’t survey. Sit in it. Let the frustration become yours until the solution becomes obvious.
For me, the pain was sitting across the kitchen table.
My wife Christine runs TiNE, a bespoke jewellery business. She designs one-of-a-kind pieces, often drawing on the cultural heritage of the wearer. Filipino, Hawaiian, Pacific Island traditions. Hawaiian heirloom motifs. Samoan geometric patterns. Every commission is a conversation about meaning, memory, and identity.
She’s extraordinary at that work. Patient, intuitive, deeply attentive to what people are really asking for when they say “I want something special.”
What she shouldn’t have to be is a systems administrator.
I’d been building her digital platform the way I knew how. Twenty-five years of creative technology experience, fifteen of them at AKQA London. Headless CMS. Analytics. Booking APIs. Hosting. Marketing automation. The full MACH architecture stack. Fifteen SaaS products, cunningly connected. I was quietly impressed with myself at the orchestration.
Then I’d walk Christine through it. Here’s your CMS for content. Here’s Stripe for payments. Here’s Shippo for shipping labels. Here’s your analytics. Here’s your email platform.
And I’d watch the light leave her eyes.
“This is all too much. I just want to focus on my work. I can’t learn all these dashboards - Make it stop.”
The Last Mile Nobody Walks
Here’s where someone usually says: “Why not just use Shopify?”
Fair question. And for many businesses, Shopify is a great answer. But there are two problems for businesses like Christine’s.
First, the economics. Shopify’s transaction fees are a percentage of every sale. When you’re selling high-value, low-volume bespoke pieces, that percentage adds up fast. A commission worth several thousand pounds loses a meaningful chunk to platform fees on every transaction. For a small business running on tight margins, that’s not a rounding error. It’s rent.
Second, and more fundamentally, Shopify gives you a catalogue. And to be fair, an increasingly sophisticated one. Their recent AI work is impressive - Sidekick is evolving into a genuine store management assistant, Agentic Storefronts are pushing products into AI conversations on ChatGPT and Perplexity. They’re innovating fast, credit where it’s due, it’s a great platform.
But here’s the thing. TiNE will have a catalogue too. Ready-to-order pieces with custom options, materials, sizing. That’s real e-commerce and it matters. The difference is that for a business like Christine’s, the catalogue is the base layer, not the whole offering. Above it sits a bespoke commissioning process, a multi-week conversation about heritage, occasion, symbolism, budget, materials, where the “product” doesn’t exist until the relationship produces it. That’s the higher-value, higher-touch part of the business. And no platform models it, because platforms are built to scale the transactional layer more than the relational one.
What a business like this actually needs is a system that supports both modes under one roof. The catalogue for browsers and ready-to-buy customers. The intelligent conversational layer for commissions. Shopify will give you the first, and Sidekick is genuinely making that first mode easier to manage. But a general-purpose store assistant doesn’t know what a bespoke commission is. It doesn’t understand the emotional weight of a heritage motif or the workflow of a jeweller who needs three weeks, multiple conversations and working with CAD designers and goldsmiths before a piece takes shape. That domain intelligence only comes from building for the specific business, not the platform.
A ring might represent reconnection with a cultural identity. The choice between a giver’s heritage and a receiver’s heritage in a design carries meaning the system should respect, not flatten. That understanding only comes from the last mile. Sitting down with the business owner. Mapping every friction point, every context switch, every moment where a creative person has to stop being creative and start being a systems administrator. This is the work that platforms can’t do and agencies struggle with at this scale. Traditionally, the economics haven’t supported it. A five-person jewellery studio or a family-run holiday estate can’t afford what an agency charges for bespoke systems. So they get squeezed into platforms that weren’t built for them, and everyone pretends it’s fine.
But the economics are changing. AI-native development tooling is fundamentally shifting what a small, focused studio can deliver. Work that used to require a team of six and a six-month timeline can now be done by a practitioner with deep domain knowledge and the right AI-assisted pipeline. The economy of scale has flipped. Bespoke doesn’t have to mean expensive any more. It has to mean close. Close to the problem. Close to the client. Close to the pain.
That’s where Sinjin Studio lives. In the last mile. For businesses too distinctive for off-the-shelf platforms and traditionally too small for custom builds. The hard yards. The kitchen table conversations. The bespoke intelligence layer that only comes from genuinely understanding what makes each business different.
The Principle
So Sinjin Studio will never just build a website.
We build sites that are beautiful for humans to navigate. Fluent for AI agents to understand. And backed by an intelligent business layer that manages the digital operations so the owner doesn’t have to.
Not auto-generating content. Not offloading creative decisions to a machine. Helping humans make human-generated content with AI assistance. Empowerment, not dependency. Not replacing responsibility, reducing friction.
A jeweller’s site that understands what a bespoke commission is. A holiday rental that responds to natural language enquiries about wedding planning. Every system shaped to the business it serves, not the other way around.
A Conversational Admin
So I’m building Christine a single, intelligent, conversational admin interface. One place to manage the entire business.
Instead of logging into Stripe to check an order, then switching to Shippo to manage shipments, then opening the CMS to update a product, then checking analytics, she just talks to her admin.
“Show me today’s orders.”
“Create a shipping label for Gerry’s order #4521.”
“Update the price of the Lava Necklace to £585 for 9ct gold.”
“Help me with some Insta posts.”
“How’s next month looking?”
One conversation. Thirty-plus tools working behind the scenes. Every action validated, every price server-derived, every shipping rate recalculated from source.
Intelligence That Acts
The key word is “acts”. This isn’t a chatbot that just tells you things. It’s an interface where AI has agency over the business tools, within guardrails.
Destructive operations require confirmation. Prices are always fetched from Stripe, never from the UI. Shipping rates are recalculated from Shippo at the point of need, regardless of what the display showed. The security model assumes the client is compromised.
Within those guardrails, the AI orchestrates complex, bespoke workflows that used to eat hours.
Adding a new jewellery piece used to mean: create content in the CMS, upload images and set focal points, configure variants in Stripe, set shipping weights, verify pricing, publish. Six tools, twenty minutes, fifteen context switches, errors.
Now it’s a guided conversation. The admin walks through each phase while a persistent workflow panel tracks what’s been collected. Images, variants, pricing, shipping data, all accumulated in real time as the conversation progresses, with the ability to pick up later and resume where you left off. Validated and ready for publishing.
The more I build it, the more I notice it converging on a familiar shape: navigation on the left, conversation in the centre for input and decisions, an interactive panel on the right for insights and actions. Three columns. If that sounds like Claude Desktop, it’s no coincidence. What I’m building is essentially a bespoke, domain-specific version of the same pattern. And I suspect the bigger version of this idea doesn’t live as a standalone app at all - it lives inside something like Claude Desktop for Business or Cowork, with an open market for developers to build custom components. If anyone at Anthropic sees the potential - hello 👋.
The Emotional Intelligence Layer
This is the part I’m most genuinely excited about.
When a customer explores bespoke commissioning on TiNE’s site, the AI doesn’t just list options or present a product configurator. It understands what design choices communicate.
A centre stone isn’t just a stone. It’s a message. An emerald might say permanence. A sapphire might say trust. The choice matters, and the AI helps articulate why.
A heritage motif isn’t decoration. It says “I see you. I understand where you come from.” When a customer is choosing between incorporating the giver’s heritage or the receiver’s in a ring design, that carries emotional weight. The system helps surface that consideration rather than reducing it to a dropdown menu.
Take a real scenario. A customer wants to commission an engagement ring for their partner, who they knew loved the ocean. They know they want something meaningful but they don’t have the vocabulary for what “meaningful” looks like in jewellery. The AI asks about the occasion, the relationship, what they want the piece to say. It suggests pearls and organic designs, not a traditional engagement ring with diamonds. It surfaces questions the customer hadn’t thought to ask: “Would you like the design to reference your partner’s heritage, your own, or both?”
How does it do this? The system is trained on real commissioning journeys and tuned to think beyond the obvious, to proactively suggest unexpected directions while always being transparent about practical considerations like care and durability. It reads how the customer is responding and calibrates accordingly, gauging their appetite for creative risk. That’s bespoke work on the AI layer itself, carefully managing what the system knows and when it surfaces it.
By the time the customer sits down with Christine for the real consultation, they’ve already found language for what they feel. They arrive with better questions. The conversation is richer. The commission is more considered.
Empowerment, not replacement. The AI makes the human interaction better, not redundant.
How It Gets Built
The development process itself embodies the same principle: AI that empowers, doesn’t replace. I use an AI-assisted development pipeline that connects structured specifications to implementation through Linear and Claude Code, with human direction at every stage.
That’s its own story. I’ve covered the full methodology, the structured spec format, the quality gates, and the execution modes, in a companion piece: The Cobbler’s Cycle.
Focus on What You Do Best. Be Guided Through the Rest.
That’s the promise.
An artisan shouldn’t need to understand webhook configurations or cache invalidation strategies. They should be able to say “I made a new piece” and have the system handle the rest, content, pricing, shipping, publishing, guided by conversation, validated by code.
The gap between what platforms offer and what small businesses actually need is enormous. The platforms solve for scale. The agencies solve for budget. Nobody solves for the independent maker, the small studio, the family business with a distinctive proposition and no time to learn fifteen dashboards. Until now, the economics made that gap permanent. AI-native tooling is closing it.
What’s the biggest context-switch tax in your business? The thing that pulls you out of the work you’re actually good at? I’d genuinely like to know.
Get closer to the pain. That’s where the good work is.