Thought Behind Things

Building Conviyo with zero investors and zero burn

Anas Shaikh explains how Conviyo grew from a single restaurant's AI agent into a conversational-commerce platform, and why he kept the business at zero burn.

  • Ep 549
  • Oct 31, 2026
  • 7 min read

A series on AI, and a case study in building alone

Muzamil Hasan opens by placing this conversation inside a longer run of episodes on AI. Earlier guests were often at a very early stage, with a few customers and some initial validation of their work, and more senior operators had joined the series too. One conversation he calls a personal favourite was with the team at Blink, where the argument was that if a business builds a solid digital foundation underneath, automation and intelligence become much easier to layer on top. That sets up why Anas Shaikh is here: he runs a company that does this for restaurants, and it has been built by a very small team.

From hospitality to a string of small ventures

Anas starts his career at Dream Resort, one of the leading resorts in Pakistan, working in a junior sales role and reaching management level within a year. From there he experimented with entrepreneurship. He describes a software house he closed within six months, then an augmented-reality project from 2018 that he says was “very early” and that he also shut down.

He then consulted for startups on fundraising while connected to venture capital in the Gulf, and in late 2019 launched an e-commerce platform in Hyderabad. COVID arrived in the middle of that venture. The team built exclusive online stores for brands and, by his account, went from about 19 orders a day in December to roughly 700 a day by March from its own warehouse. “That was my first sort of success out on the journey,” he says.

Later he joined a startup in a senior operating role, then a QR wallet business, and then Chiku, where he says he was briefly the only person responsible for the whole operation. He describes merchants growing from around 120 to around 1,500 in four months, and about 5,000 at the peak before that company was acquired. A loyalty app he worked on afterward, he says, burned through his savings, and that cost is part of why he later took the cost of AI far more seriously.

How the first agent came together

The founding moment is specific. Anas spent June and July learning AI at night. A mentor, Rajab, suggested he look at n8n, and within about 24 hours he had built a first AI agent connected to Shopify and sent a video of it to his mentor, who told him to start selling it. The first sale came through a friend who ran a restaurant brand and agreed to test the product for three days.

Anas is candid about how the early product was built. “It’s like building a bridge while not being an engineer,” he says of the workflow-building stage. When things broke, he initially tried to fix them himself, until his mentor pushed him to bring in engineers. He calls that an ego adjustment, and it is when he stopped working on his earlier venture and focused on the agent.

He also draws a sharp line about how building has changed. “Anyone can build,” he says, and the bottleneck has moved elsewhere.

Distribution is the hard part now

Anas makes his central argument here. He says building a product has become cheap, and that distribution is what still matters. He compares it to the media industry fifteen years ago, when cheap cameras, cheap editing and social platforms made cheap distribution possible. “Quacks like me could come in and build shitty media products,” he says, and business models were built on top of them.

He adds that distribution is still hard without relationships. His own company, he says, benefited from strong partners, including Blink, and from integrations with most of the systems Pakistani businesses already use. Muzamil pushes back that many AI founders still struggle with distribution, and Anas agrees the difficulty is real even if the cost of reaching customers has come down.

Why zero burn

A recurring thread is Anas’s rejection of the burn-and-grow model. He says it drags companies down, and that it does not work well in Pakistan. He describes a moment when Conviyo had Google credits and began spending them quickly: the team burned through around $5,000 in credits within two months, and the cost of a single conversation had risen to $5.80. That led to a close look at the infrastructure, and he says the team brought the per-conversation cost down to around seven cents.

He is also clear about what cheap capital hides. “In many ways you were subsidising your customers,” he says of the early months, when the real cost of the service was masked by credits. A mentor, he says, told him early to focus on unit economics and understand what each component costs.

Anas says the company has kept growth organic and has deliberately not chased outside money. He presents this as discipline rather than a boast.

Conversation as an operating layer

Anas explains how the agent works across WhatsApp, Instagram, Facebook, TikTok and Snapchat. A customer reaches out, the agent identifies the nearest branch, searches the menu, builds a card, and pushes the order to the relevant branch, without a human stepping in unless a trigger the business set off fires. He says the agent speaks more than 90 languages, in text and voice, and gives an example of a Peshawar burger chain whose customers send voice notes in Pashto.

His view of the category is that conversation should be treated as a layer, closer to an operating system than a single chatbot. “Right now people are looking at conversational as a technology,” he says. “I’m looking at it as an ecosystem,” with apps built on top. He sees the value in how closely the company works with customers to solve their problems, rather than in any one tool. “What works is how closely you are working with the customer to solve their problems,” he says.

Meta’s pricing change and the web chat workaround

The conversation turns to pricing. Anas says Meta’s announced WhatsApp pricing, which he describes as high relative to the region, forced a rethink across the industry. He notes that a single message can cost around ₹5, and that a full conversation could cost far more once friction points stack up.

Conviyo’s response was to deploy a web chat interface that customers reach from a link, so the conversation can happen outside WhatsApp’s charged messaging layer. The web experience was built to closely resemble WhatsApp. Anas says that early on, customers who reached out on WhatsApp often never opened the web link: in one restaurant chain’s data, 54 percent of customers who reached out on WhatsApp never opened the web chat link. He describes that as a sign the company was hurting the businesses it was meant to help. The response, he says, is a WhatsApp-based flow that feels like a native app inside the messaging platform. He is careful not to overstate the rollout, noting that the company is still moving toward it.

Who the platform is for

Anas is explicit that the product is not built for everyone. The company began by working with everyone, which he says caused trouble, and then shifted toward tier-one brands with multiple branches and teams that understand AI. Some customers who had been on other platforms moved over when WhatsApp’s pricing changed.

He also describes expansion beyond restaurants. He says the company has begun working in clinics and doctors’ practices, where reservation management is a major use case, and that it has integrated with a large pharmacy chain. It is also working with a couple of customers in Saudi Arabia, including one that sells IV drips through Shopify and sends a booking link over WhatsApp once an order is accepted.

Self-serve onboarding and what remains open

Near the end, Anas describes the next steps. The company plans to launch a free trial and a self-onboarding flow in which a business connects its Shopify or Blink account, its WhatsApp, and its other channels, and the AI sets up its menu and upselling logic automatically. For now, sign-ups are moderated and reviewed, because the company is integrating payments and wants to get recovery right, which he says is one of the biggest problems in Pakistan.

Muzamil closes by thanking Anas and inviting viewers to share their views in the comments. The conversation is candid about what is unfinished: the self-serve model is not yet public, the web chat is not fully rolled out, and Anas says the company is still learning what works in its market. The throughline is a product built on cheap distribution, a refusal to burn capital, and a focus on the end customer’s experience rather than the technology behind it.

Build with me in the Inner Circle.

You don't need to code. You need a problem you know well, and a room of people building beside you.

Muzamil Hasan speaking on stage