The Night Shift That Never Sleeps: How Carl James Juevesano Builds a Shopify Store and Teaches Its Agents to Sell

It is 2 AM, and the store is open. The store is always open. That is the strange bargain at the heart of ecommerce: the lights never go off, the door never locks, and a customer can walk in at any hour of any day, from any time zone on Earth, ready to buy something. What the store cannot do, most of the time, is answer. A physical shop keeps human hours. It opens, it closes, and nobody expects the person behind the counter to be standing there at two in the morning. An online store makes the opposite promise and then quietly breaks it. The shelves are stocked, the checkout works, the products gleam under perfect digital light, and the one thing a hesitant customer actually needs, a reply, is asleep in another time zone.

The questions that arrive at night are not exotic. Where is my order. Does this come in a larger size. Can I still change the delivery address. If I order today, will it reach me by Friday. Each one is small, and each one is a door quietly closing. A customer who asks a question is a customer in the middle of deciding, and a decision left alone until morning usually decides itself in the worst way: the tab gets closed, the cart gets abandoned, the moment passes. Stores do not lose these sales dramatically. They lose them silently, one unanswered question at a time, in the hours when nobody is watching.

This is a story about a man who works on that exact problem for a living, and about the unfashionable way he got good at it. His name is Carl James Juevesano. He is an SEO specialist and web developer from Iligan City in the Philippines, with more than eight years of experience, and these days a growing part of his work is configuring AI customer support agents for Shopify stores: agents that answer product questions and guide customers toward checkout when no team member is online. But the interesting thing about Carl is not the agents. It is everything underneath them. The agents sit on top of eight years of search rankings, websites, email campaigns, and automation plumbing, and that foundation is the whole story. AI, in his practice, is not a costume he put on last year. It is a layer he added, slowly and deliberately, to a craft he already had.

The night problem, from first principles

Start with the arithmetic of an online store. A store sells during the hours its customers are awake and thinking about buying. Its staff, if it has staff at all, works a fraction of those hours. For a small store, the gap is enormous: the shop floor is open for twenty four hours and staffed for perhaps eight, which means two thirds of the store's life happens with nobody home. Large companies paper over this with support teams on three continents and follow the sun scheduling. Small stores, the kind that make up most of ecommerce, cannot. For them the night gap is not an edge case. It is the majority of the day.

Now add geography. The internet did not just make stores open all night; it made them open everywhere. A store built in one country routinely sells to customers in a dozen others, and every time zone it sells into adds another night that happens during somebody else's day. A customer browsing at a comfortable 8 PM in their own city may be knocking on the door at 3 AM in the store owner's city. Nothing about this is a failure. It is the basic physics of selling online. The failure is only in pretending that a contact form and a promise to reply within one business day is an answer to it.

There is a second, quieter half to the problem, and it matters just as much. The questions that go unanswered at night were never only support questions. Where is my order is support. But does this come in a larger size, asked by someone with a credit card in hand, is a sales conversation wearing a support costume. Stores tend to staff and think about these two kinds of conversation separately, with support as a cost to minimize and sales as a prize to chase. At 2 AM they collapse into the same thing: a person, mid decision, asking the store to help them finish deciding. Whoever or whatever answers that person is doing support and selling at the same time, whether the store planned it that way or not.

Hold those two ideas together, the staffing gap and the support sales blur, and you have the real shape of the problem Carl works on. The answer he configures for Shopify stores is an agent: software that knows the store's products, answers the recurring questions instantly, and walks the customer toward checkout while the humans sleep. Done badly, this is the robotic chat window everyone hates, the one that loops cheerfully and answers nothing. Done well, it is something else entirely, and the difference between badly and well turns out to be the most interesting part of his practice. We will get there. First, the man.

Meet Carl: eight years of the unglamorous stack

Carl James Juevesano works from Iligan City, in the northern part of Mindanao in the Philippines. His trade, stated plainly on his AI Pro profile, is SEO and web development, with more than eight years of experience, and with email marketing and automation alongside. He claimed profile number 4 in the E2 Discord, which is a small detail that says something real: he showed up early, and he is available to start immediately. His skill list reads like the org chart of a small digital agency compressed into one person: SEO, Shopify, WordPress, AI support agents, workflow automation, web design, email marketing, Canva, ChatGPT, and Claude.

Notice what is not on that list: anything fashionable for its own sake. There is no metaverse, no blockchain, no proprietary framework with a clever name. There is the stack of things a small business actually needs to sell things on the internet, learned in the order a small business needs them. Can people find the store (SEO). Does the store exist and work (web development, Shopify, WordPress). When people find it, does it look trustworthy (web design, Canva). When they leave, can you bring them back (email marketing). When the same task happens for the hundredth time, can a machine do it (workflow automation). It is a deeply unglamorous list, and it is exactly the list that matters, because glamour was never what kept a store alive.

His past web work follows a pattern he has built his practice on: SEO and site work for businesses in Australia and the United States. Sit with that for a moment, because it explains everything else in this story. A developer in the Philippines serving clients in Australia and the US is living the time zone problem from the inside. His clients' customers are awake, shopping, and asking questions during hours when he is asleep, and his clients are awake to check his work during hours that overlap only partly with his own day. The international client pattern is not just a market he serves. It is the condition his whole working life is organized around. When such a person starts configuring agents that answer customers at night, he is not chasing a trend. He is solving his own recurring Tuesday.

There is a temptation, when profiling anyone who works with AI, to skip straight to the AI. Resist it here, because Carl's story makes no sense without the long runway. Eight years of SEO means eight years of watching how people actually search, what they actually click, and which words on a page actually move a stranger to trust a store with money. Eight years of websites means knowing, in his hands and not just in theory, how a store is put together: themes, products, collections, checkout, the hundred small settings that decide whether buying feels easy or feels like paperwork. Eight years of email marketing means understanding the second visit, the customer who came back because a message arrived at the right time. An AI agent dropped into a store by someone with that history is a different object from the same agent dropped in by someone who discovered ecommerce last spring. The tool is the same. The judgment around it is not, and judgment, as we are about to see, is the actual product.

The time zone is the territory

It helps to understand where Carl sits, geographically and economically, because his practice sits at the crossing of two facts. The first fact is Iligan City: a real city, in a real province, in a country whose skilled remote workforce has spent two decades learning to serve businesses on the far side of the planet. The Philippines did not stumble into that role. It built it, shift by shift, in call centers and support desks and design studios, until working while your clients sleep became a national competence. The second fact is that Carl's version of that competence is the independent one. Not a seat in a call center, but a practice: his own clients, his own stack, his own name on the work, built on SEO and site work for businesses in Australia and the US.

Think about what the international pattern does to a person's instincts. When your clients are in Australia, their business day starts a few hours ahead of yours, close enough to feel familiar. When your clients are in the United States, their day is nearly your night flipped upside down. Serving both teaches a lesson no course can: the internet's customers do not share a clock, and any system you build that quietly assumes they do will fail them at scale, politely, every single day. The store that only answers during its owner's waking hours is not a store with a support policy. It is a store with a geography problem it has decided not to solve.

It also teaches the reverse lesson, the one about trust. A client in Sydney or Chicago who hands their store to a developer in Iligan City is trusting someone they may never meet in person, across a distance that makes casual oversight impossible. The only currency that survives that distance is visible, checkable work: rankings that move, pages that load, stores that take orders, agents that answer correctly at hours when the client can test them and the developer is asleep. That last detail matters more than it might seem. An AI support agent configured by a remote developer is testable by the client at the exact moment of its greatest claimed value, in the dead of the developer's night. It either works or it does not, and both parties know it by morning. Carl's practice grew inside that discipline of checkable work, which is one more reason his current build is a live store anyone can visit, rather than a gallery of screenshots anyone could stage.

The live build: a trade show display store, mid construction

The best way to understand a builder is to look at the thing currently on his workbench, half finished, with the tools still out. On Carl's workbench right now is a live Shopify store for custom trade show displays. Backdrop displays. Booth solutions. The big printed walls and counters and banners that turn an empty rectangle of convention center floor into a company's temporary headquarters. It is worth pausing on how different this is from the stores that fill most tutorials. This is not a demo theme with three placeholder products and lorem ipsum where the prices should be. It is a real storefront for exhibitors, people whose buying is driven by event dates and freight deadlines, and it is being built out, right now, from the stock theme it started as.

That phrase, built out from the stock theme it started as, deserves a closer look, because it describes how almost every real store actually gets made. Nobody serious starts from nothing. You start from a theme, a prefabricated storefront that looks respectable in the demo and slightly wrong for your actual business in a hundred specific ways. The work of building a store is the long walk from that generic starting point to a place that fits one particular business: its products, its customers, its way of taking an order. Watch what this particular store needed. Artwork upload, so that an exhibitor can hand over the graphics that will be printed across a ten foot backdrop. Live chat, so that a buyer with a show next month can ask a human (or, at night, an agent) whether the timeline works. These are not decorations. For a display store, artwork upload is the product's front door, because a backdrop without the customer's artwork is just expensive cloth. Every store has one or two features like this, features that look minor in a feature list and are actually the reason the business can operate at all. Finding them, and building them properly, is the job.

Now look at the fingerprints AI left on this build, because they are exactly where a thoughtful practitioner would put them. Many of the store's banners and images were created with ChatGPT prompts. Think about what that means for a display company: the product is visual, the store has to sell printed visuals with visuals, and the traditional path to a full set of polished banner imagery runs through photoshoots, stock libraries, and design hours. Prompted images collapse that path. But notice what they do not collapse: someone still has to know what a trade show backdrop should look like, judge whether a generated image would survive being printed eight feet wide, and reject the ones that would not. The prompt is the easy part. The eye is the job.

The second fingerprint is deeper. AI coding assistance built features and design elements that were not in the original theme. This is the quiet revolution hiding inside routine store builds, and it is worth stating plainly. For years, the distance between a stock theme and the store a business actually needed was measured in developer hours, and small businesses routinely lived with the gap, apologizing for the workarounds. AI coding assistance shrinks that distance dramatically, but only for someone who can do three things at once: describe the missing feature precisely, read the code that comes back well enough to know if it is sound, and integrate it into a live store without breaking checkout. Carl uses AI for coding assistance and troubleshooting as standing tools of his trade. On this store, that habit stopped being a productivity trick and became the difference between the theme the store started with and the store the business needed.

Keep the honesty of the profile in view here. The store is live, and it is also still being built out. Those two facts together are the truth about nearly every working store on the internet, and about nearly every working builder. A store is not a statue that gets unveiled. It is a garden that gets tended, and the tending never really ends, because the business keeps changing what it sells and the customers keep changing how they ask. Carl's store, mid construction in public, is not an unfinished portfolio piece. It is a fair picture of the real thing.

The agent is a configuration, not a model

Here is the claim at the center of Carl's practice, and it runs against the grain of how AI support tools are usually sold. Configuring an AI customer support agent is configuration and workflow work, not model training. Nobody sensible is training a model for a Shopify store. The models already exist, they are excellent at language, and they are identical for everyone. What differs from store to store, wildly, is everything wrapped around the model: what it knows about this store's products, which questions it should answer on the spot, which ones it should guide gently toward checkout, and which ones it must not touch at all. That wrapping is the work. The agent is a configuration, and configuration is a craft.

Break the craft into its parts and it stops looking like magic and starts looking like what it is: a series of judgments. First, the agent has to know the store. Not the internet's idea of the products, but this store's products: their sizes, their options, their prices, their quirks, what ships quickly and what does not. A customer asking about a backdrop display does not need a lecture on printing in general; they need to know about this backdrop, this material, this turnaround. Second, the agent has to know the store's policies and facts well enough to answer the recurring questions instantly and identically every time, because the night shift's great advantage is consistency. A human at the end of a long day gives slightly different answers on Tuesday and Friday. A well configured agent gives the store's real answer every single time.

Third, and most important, the agent has to know where its answers should stop. This is the line in Carl's thinking that elevates the whole practice. There are questions an agent should answer completely, questions it should answer partly while guiding the customer toward checkout, and questions it should hand to a human in the morning, carrying a clean summary of what was asked. A store that lets an agent improvise on refunds, complaints, and exceptions has not automated its support; it has automated its apologies. A store that draws the boundary well gets the opposite: customers feel answered, the agent sells while it helps, and the humans wake up to a short list of conversations that genuinely needed a human. The boundary is not a limitation of the system. The boundary is the system. Drawing it well requires knowing the business, the products, and the customers, which is to say it requires exactly the eight years standing behind the configuration.

Notice, too, what the agent is for, in the way Carl sets it up. The agents on his Shopify projects answer product questions and guide customers toward checkout when no team member is online. The goal is not to replace the team. It is to hold the conversation until a human returns. That phrase, hold the conversation, is the honest center of the whole night shift idea. The agent's triumph is not closing every sale at 3 AM. It is making sure the customer who arrived at 3 AM still feels welcomed, informed, and inclined to finish the purchase, tonight with guidance, or tomorrow with a human who picks up the thread instead of starting cold. An agent that holds conversations turns the store's deadest hours into its quietest asset. An agent that tries to be the whole store turns them into a liability with good grammar.

There is one more quiet consequence worth naming. When an agent handles the recurring questions, the humans stop spending their mornings on where is my order and start spending them on the conversations that actually need judgment, empathy, and authority. The night shift does not shrink the team's work so much as it purifies it. And because the agent has been logging what customers ask in the dark, the store wakes up knowing things it never knew before: which product confuses people, which page fails to answer the obvious question, which size chart nobody can find. The night shift, set up by someone who also does the SEO and the site work, feeds its lessons straight back into the store. That loop, from night question to morning fix, is where we go next, by way of the habit that makes all of it possible.

Anatomy of a held conversation

It is worth slowing down and walking through what a well configured night actually looks like, minute by minute, because the difference between the agent everyone hates and the agent Carl describes is visible in the small steps. A visitor lands on a product page at an hour when the store's team is asleep. Something is unclear: a size, a material, a delivery window, a detail the page states somewhere they have not found. They open the chat and ask. The first thing the agent owes them is speed, not as a gimmick but as a courtesy: an instant, accurate answer in the store's own facts, the answer the page should have given them. This is the tier where agents shine, and where most night questions live, because most night questions are ordinary questions asked at an extraordinary hour.

Suppose the question sits next to a purchase, as product questions so often do. The customer is not merely curious; they are deciding. Here the configuration earns its keep a second time. The agent answers the question and then does the gentle, legitimate thing a good shop assistant does: it points at the next step. The size exists, here it is, and here is the way to checkout. Guidance, not pressure. The material describing Carl's setups is precise about this pairing, answering product questions and guiding customers toward checkout, because the two acts are really one act seen from two sides: remove the uncertainty, then show the door the uncertainty was blocking.

Now the hard case, the one that separates configurations from toys. The question is an exception: a complaint, a damaged delivery, a refund demand, a situation tangled in feelings and money. This is where the boundary drawn in advance does its work. The agent does not improvise generosity with the store's money and does not stonewall the customer either. It acknowledges, it gathers the facts cleanly, and it holds the conversation like a good night porter holds a message: this will be with a human in the morning, and here is exactly what will happen next. The customer goes to sleep feeling heard rather than handled, and the team starts its day with a prepared summary instead of a cold complaint. Three tiers, one conversation: answer what is known, guide what is ripening, and hand over what is human. None of it requires a new model. All of it requires a person who knows the store well enough to decide, in advance, which tier every kind of question belongs to. That deciding is the service, and it is why the configuration is a craft practiced by someone with eight years of context rather than a switch flipped by someone with none.

AI as a habit, not a costume

Every profession has a tell that separates the people who live in a tool from the people who are visiting it, and for AI the tell is time. Carl has used ChatGPT for about three years, on a paid subscription, and Claude for about a year. Read that the way you would read any other line in a tradesman's history. Three years of paying, monthly, out of his own practice's pocket, for the top shelf version of the tool. Years of daily reps, predating the current wave of enthusiasm by a comfortable margin. A person does not keep a subscription alive for three years because a tool is fashionable. He keeps it because Tuesday is easier with it than without it, and he has had a thousand Tuesdays to make sure.

What does he use them for? The list on his profile is refreshingly unglamorous: coding assistance, troubleshooting, content, and website design ideas. Four workhorse uses. Coding assistance, the store builds and theme features we have already seen. Troubleshooting, which deserves more respect than it gets, because troubleshooting is where AI meets reality at its least forgiving: something is broken, the error is specific, and a plausible sounding answer that is wrong is worse than no answer at all. A practitioner who uses AI for troubleshooting has learned, through direct and sometimes painful experience, how to interrogate it, test what it claims, and keep his own diagnosis in charge. Content, the raw material of SEO and email. And website design ideas, the top of the funnel for the design work, where a tool that can generate ten directions in a minute makes the human's taste more valuable, not less, because now the scarce skill is choosing.

Contrast this with the costume version, which anyone who hires freelancers has met. The costume version discovered AI recently, lists every tool by name, and generates work that looks fluent until the first hard question arrives. The difference is not talent and it is not intelligence. It is reps and stakes. Carl layered AI onto eight years of fundamentals, which means he can see when the machine is wrong, because he knows what right looks like without it. He did the work by hand first. That ordering, fundamentals first and acceleration second, is the single most reliable pattern in every credible story about AI making someone better at their job, and it is worth engraving, because the reverse ordering produces people who can generate answers and cannot evaluate them, which is a dangerous kind of fast.

There is also something to be said for the modesty of the setup. Two tools, deeply used, beat fifteen tools sampled. ChatGPT, paid, for years. Claude, for about a year, alongside it. That is it. No sprawling stack, no tool of the week. The practitioners who get the most from AI tend to look boring from the outside, in the same way a carpenter's favorite hammer looks boring. The excitement is in the workbench, not the toolbox.

The loop: found, welcomed, answered, brought back

Now assemble the whole practice and look at it as one machine, because that is how a store owner experiences hiring Carl. A customer begins as a stranger with a question typed into a search engine. SEO is the craft of being the answer that stranger finds, and it is the top of everything Carl does: eight years of learning which pages deserve to rank and how to build them so they do. The stranger clicks, and lands in a store. Web development and design decide whether that store earns trust in the first ten seconds, and features like artwork upload and live chat decide whether the store fits the business it serves. The stranger hesitates, at noon or at 2 AM, and asks a question. The agent answers it, accurately, in the store's own facts, guiding toward checkout, or holding the conversation gracefully until a human returns. The customer buys, or almost buys, and weeks later an email arrives with the right thing at the right moment, and the loop turns again.

Found, welcomed, answered, brought back. Four crafts, one loop, and in Carl's case, one person. This is the part of the story that should make small business owners sit up, because the loop is usually sold in pieces by different specialists who never meet: an SEO here, a developer there, a chatbot vendor with a subscription, an email tool with another. Each piece works, roughly, and the seams between them leak money in ways nobody owns. The person who can hold the whole loop at once is not four times as useful as the person who holds one piece. He is useful in a different way entirely, because he can see the leak wherever it is. When the night questions reveal that a product page confuses people, the same hands fix the page. When search traffic arrives asking for something the store does not explain, the same hands write the explanation and teach it to the agent. The loop improves as a loop, and that compound effect, small fixes feeding each other month after month, is what eight years of fundamentals plus three years of AI habits actually buys.

It also explains why the night shift framing undersells itself. The agent is the most visible piece, the one that feels like the future arriving. But in the machine as a whole, the agent is simply the piece that closes the last open gap in the day. SEO already covered the finding. The store already covered the welcoming. Email already covered the return. The only hours left undefended were the dark ones, and those are the hours Carl's agents now stand watch over. The night shift that never sleeps is not a robot replacing a store's people. It is the final shift in a store that finally, fully, keeps its own promise: open means open, and someone, or something well taught by someone, is always there to answer.

Lessons worth stealing

For store owners, Carl's practice compresses into a handful of moves that any store can make, with or without hiring him.

• Start with the questions, not the tool. Before configuring anything, collect the questions your store actually receives when nobody is online. That list, usually shorter and more repetitive than owners expect, is the entire job description of your night shift.
• Write down where the agent must stop, before you switch it on. Which questions get full answers, which get guidance toward checkout, and which get a graceful handoff to a human in the morning. The boundary is the product. A store that cannot state its boundary is not ready for an agent, and a vendor who does not ask about it is not ready either.
• Treat a stock theme as a starting point, not a store. Every business has one or two features that look minor and are actually load bearing, the artwork upload of your particular trade. Find yours and build them properly, with AI coding help if you have someone who can judge the code.
• Let the night shift teach the day shift. Read what customers ask in the dark. Every repeated question is a page that failed, a product description that skipped the obvious, a policy that lives nowhere findable. Fix the store with what the agent learns.
• Keep a human thread to return to. The goal is a conversation held, not a human replaced. The stores that do this best make the morning handoff feel like a relay, where the customer never has to start over.

For freelancers and builders, the lessons point the other way, inward, at the shape of a career.

• Fundamentals first, acceleration second. Carl's AI work is credible because eight years of SEO, sites, and email stand underneath it. If you are early in your career, the worst reason to skip the fundamentals is that AI makes them look optional. It makes them more valuable, because they become the part only you can supply.
• Pay for your tools and keep them for years. A paid subscription kept across three years of Tuesdays is what fluency looks like from the outside. Depth with two tools beats tourism across fifteen.
• Sell the loop, not the widget. Anyone can install a chat window. The offer that matters is the whole path a customer travels: found, welcomed, answered, brought back. Learn enough of the adjacent crafts that the seams stop leaking, and you become very hard to replace.
• Work in public, mid build. A live store still being built out says more about a builder than a polished case study, because it is checkable. Carl's workbench has a real store on it. Let yours have something real on it too.

Why this belongs at Starvard

Starvard exists to teach how the world actually works now, and few lessons are more current, or more commonly taught badly, than the one inside Carl's story. The common version of the AI lesson says: learn the newest tool, move fast, automate everything, and do not look back. Carl's eight years say something truer and more useful. Learn a real craft until your judgment is trustworthy. Then let the machines make you faster at the parts that were never the point. The point was always the judgment: knowing what a store needs, knowing what a customer is really asking, and knowing, precisely, where an agent's answers should stop.

His story also belongs here because it is honest about the middle. The store on his workbench is live and unfinished. The agents are configurations, not magic. The AI habit is measured in years of subscription receipts, not in a weekend of enthusiasm. Starvard's pages are full of people building real things in the open, and Carl, profile number 4 in the AI Pro lineup, building a trade show display store in Iligan City for customers on the other side of the world, is exactly the kind of builder this university was made to celebrate. You can read his full background, skills, and availability on his AI Pro profile.

If you are a store owner whose customers keep asking questions in the dark, or a builder somewhere layering new tools onto an honest craft, you do not have to figure this out alone. The builders in this story gather in the E2 Discord, where the AI Pro profiles live and where people like Carl post their work and answer for it. Come and meet them: join the E2 Discord. Bring your hardest 2 AM question. Someone there is awake, and even when they are not, their night shift is.

Watch

Can AI Run Customer Support? I Tested It on 50 Tickets by Ops with Cam: a builder connects AI agents to a Shopify support inbox, runs them against 50 test tickets and a written returns policy, and scores every decision against an answer key written in advance. It is Carl's philosophy with a scoreboard attached: policy and boundaries first, and a human handed anything unclear.

How to Build a Full Shopify Store using AI (Claude Code) by Metics Media: a complete build, from a blank Shopify dashboard to a published store ready to take orders, with an AI coding tool doing the construction while a human directs. Watch it as the far end of the same move Carl makes on the trade show store: features and design that were never in the original theme.

Tech SEO in the Age of AI Search webinar by Lumar: two technical SEO specialists walk through what changes when AI crawlers and AI Overviews join human readers in the audience for your pages, from crawlability to structured data. It is a good map of where Carl's eight year craft is heading next.

Sources

Carl's AI Pro profile: Carl James Juevesano on AI Pro, with his background, skills, and availability, including eight plus years of SEO and web development and his Shopify AI support agent work.

The live build: his Shopify store for custom trade show displays, a real storefront that is still being built out from the stock theme it started as, presented honestly as a work in progress, with banners and images created from ChatGPT prompts and features built with AI coding assistance.

Reporting source: the AI Pro profiles collected from the E2 Discord community, where builders post their work, their tools, and their availability, and where Carl claimed profile number 4.