Tahir Rauf on Vertical AI Agents for E-Commerce

October 24, 2026
Tahir Rauf · Keeyu
Co-Founder & CTO · Keeyu
AI InterConnect Podcast · Published October 24, 2026

Tahir Rauf is the co-founder and CTO of Keeyu, which builds AI agents for e-commerce operations that help retailers identify and resolve post-purchase problems before they reach the customer. He tells Laurent Cochet how learning to program on paper shaped his thinking, the four principles he believes make agents reliable, where vertical AI fits, and why he thinks the SaaS products that succeed will be those that add AI capability to those principles.

Programming on paper

Laurent notes that Rauf grew up in Pakistan and later moved to Australia to work in technology. Rauf says that when he was entering college he had no access to a computer, so he started writing programs on paper, turning exercises such as reversing an array of integers into a kind of game.

The benefit, he says, was clear thinking. Software engineering is all about thinking through scenarios and solving them in your head first, so when he finally got access to a computer at university, he was already good at it.

Complaints start upstream

Drawing on two decades of work with businesses ranging from Mastercard's loyalty division and one of the biggest grocery chains in the ANZ region to mid-market companies and SMEs, Rauf sees the same pattern: a customer complaint is usually the result of something that broke upstream. An order passes from a platform like Shopify or Magento through warehouse management systems, carrier platforms and order management systems before it reaches the customer.

Keeyu does two things: it identifies those problems and then fixes them. Help desk systems, he says, make it easy for customers to complain and for teams to solve a problem after it has occurred. Keeyu flips that into what he calls proactive customer care, detecting and resolving issues such as late deliveries, wrong items or out-of-stock orders so the customer is never affected.

By his estimate, the fit starts with businesses handling at least five hundred orders a month that run both an e-commerce platform and a help desk, since that is where fragmented systems start to miss things.

Four principles for agents businesses can trust

Rauf says any agentic system needs four principles to be reliable. The first is context: giving the agent enough information to do the job. Too much adds cost and risks oversharing private customer data; too little leaves the agent unable to act. Keeyu relies on an integration engine that passes a refined amount of data. The second is boundaries, set together with the customer, so an agent cannot, for instance, cancel a thousand orders.

The third is the path to escalation. Keeyu builds a risk profile for each issue it detects: a simple order of thirty or forty dollars might be canceled autonomously, while a high-impact order brings a human in to approve it. The fourth is accountability, meaning the agent can roll back or fix what it got wrong and shows people what it has done without burying them in detail.

Keeyu is not meant to replace people, he says. New customers start with most actions requiring human approval, and the agent takes control step by step as they get comfortable. In the ideal scenario, he says, users end up orchestrating agents and approving only high-impact operations. For enterprises, the bigger concern is data integrity and sovereignty, so Keeyu built a governance layer that tokenizes data before it reaches the underlying model, removing details that could identify a customer.

Where vertical AI fits and what happens to SaaS

Rauf places vertical AI between two extremes: SaaS, where users perform every action themselves, and generic agents and LLMs that take a lot of work to behave as intended. Vertical solutions combine agentic and LLM capabilities with industry expertise and a governance layer. Before they became Keeyu customers, he says, many businesses built their own siloed applications with tools like ChatGPT, which risked exposing secrets and customer information.

He argues traditional SaaS is becoming less useful because software development is now effectively free. A non-technical partner at Keeyu built a full web and mobile quiz app for an all-hands meeting on her own, work he says would have taken multiple developers two to three months a year earlier. He does not expect SaaS to disappear, though: in his view, the SaaS products that succeed will be those that bring in the four principles while adding AI capability.

Smaller teams, fewer features and the US market

Keeyu is now expanding beyond Australia into the US, where his co-founder, who looks after sales and investment, has built a team and partnerships with distribution channels. Rauf names Puma, recently onboarded, and the supplements company IM8 among its enterprise customers, and says Keeyu is trying to make the agent easy enough for enterprises to use without the company deploying its own engineers to them.

With twice the resources, he would add sales staff rather than engineers. Most code is now written by LLMs, he says, and with 100 engineers it becomes harder to manage that output, while a smaller team can see one month of work done in a day.

The same logic shapes product decisions. Keeyu spent time building and publishing a Shopify app, only to find customers preferred its own platform. His lesson, and his advice to founders, is to start with a problem customers will pay to solve and build only the features that matter to them.

Watch the full conversation

Watch the full conversation above. Tahir Rauf says LinkedIn is the best place to reach him.

Watch on YouTube →

AI InterConnect hosts conversations with founders, investors and operators building with AI. Explore more episodes on the AI InterConnect podcast page.

Back to Blog

AI InterConnect

AI InterConnect connects founders, CEOs, investors and technology leaders through conversations, market intelligence and growth infrastructure.

Explore our podcast, research and GTM systems to identify opportunities, build authority and accelerate business growth.