Share from 0:00
0:00
/

The Innovators: How Keeyu uses AI to stop customer complaints before they happen

Co-founder and CTO Tahir Rauf explains how his Australian startup is using AI agents to fix e-commerce problems before customers even know something has gone wrong.

Most companies treat customer complaints as something to resolve after the fact. A package arrives late, an order goes missing or a customer receives the wrong product. The customer contacts support, someone opens a ticket and the company begins figuring out what happened.

Tahir Rauf thinks this process is fundamentally backwards.

Rauf is the CTO and co-founder of Keeyu, an Australian startup building AI agents that identify and resolve problems in e-commerce operations before they turn into customer complaints.

On this episode of Innovators, I spoke with Rauf about why he chose to tackle fulfillment, how his company is using AI to automate business operations and why industry-specific data may be the most important advantage an AI startup can have.

“Why does the customer complaint have to come? Why can’t we stop it?” Rauf asked.

That question became the foundation of Keeyu.

According to Rauf, there are more than 200 patterns that can lead to a problem with an online order. Many have little to do with the shipping company, even though carriers often receive the blame.

For example, an online store might show that five shirts are available. A customer places an order, but when the request reaches the warehouse, the inventory isn’t actually there. The customer eventually receives a delay notification or discovers that the order cannot be fulfilled.

The problem began long before the customer noticed anything was wrong.

Keeyu connects to the systems involved in processing an order, including e-commerce platforms, warehouses, enterprise resource planning software and shipping carriers. Its AI agents monitor those systems, identify problems and take action to resolve them.

Rauf said the company currently has about 30 customers and claims to resolve more than 90 percent of the issues its system detects early. Those figures are company-reported, but they suggest there is demand for a product that goes beyond simply analyzing customer service tickets.

The company’s approach also reflects a broader change in enterprise software. Rather than building tools that tell employees what went wrong, startups are increasingly developing agents that can take action without waiting for someone to intervene.

For Rauf, however, the real advantage isn’t simply the AI.

It’s the data.

As generative AI makes software easier to build, traditional software companies face a growing problem. Features that once took months to develop can now be replicated much more quickly. Rauf believes the advantage for companies like Keeyu comes from their ability to collect, organize and understand information across complicated business systems.

“A lot of effort that we spend on is making sure that the underlying data is clean,” he said.

Keeyu’s system combines information from multiple sources to understand the full context of an order. A general-purpose AI model might be able to write code that checks inventory, but it doesn’t automatically understand the relationships between a retailer’s warehouse, fulfillment provider, shipping carrier and customer support systems.

That knowledge has to be built into the product.

Rauf came to this problem after spending about 20 years working in enterprise technology, including a decade focused on e-commerce. He and his co-founders had repeatedly encountered the same operational problems while working for different businesses.

Eventually, they realized that building individual solutions for each employer made little sense when so many companies faced the same issues.

“Why don’t we build an engine instead of solving the problem everywhere in isolation?” he recalled.

The result was Keeyu, a platform intended to work across different retailers and their existing software.

One of the company’s biggest challenges was getting customers onto the platform.

When Keeyu started three years ago, implementation could take one or two months. The company had to coordinate with warehouse teams, e-commerce departments, ERP administrators, carriers and other service providers. Each integration required technical work and access to the right systems.

Today, Rauf says Keeyu has reduced that process to a few hours through what the company calls agentic onboarding.

The system uses AI to identify and configure much of the required integration work automatically. According to Rauf, customers can begin using the platform on their first day.

This matters because complicated implementation has traditionally been one of the biggest obstacles to adopting enterprise software. A product might promise significant savings, but if installing it requires months of meetings and technical work, many businesses will never get started.

Keeyu is also using AI in its own sales operations. Rauf said the company has refined its ideal customer profile and uses AI agents to help identify and reach potential customers. Alongside its Australian sales operation, Keeyu has recently established a sales team in the United States.

The company is also pursuing partnerships with the software providers that already serve online retailers, including e-commerce platforms, warehouse systems, help desks and shipping carriers.

Rauf sees these relationships as central to the company’s growth because integration is at the heart of what Keeyu does.

But perhaps the most interesting part of our conversation concerned the future of AI software itself.

Rauf believes that general-purpose AI tools are useful, but they also create problems when employees build their own solutions without common rules or security controls.

In a company with 20 employees, for example, each person might create a different AI workflow. Without shared standards, those systems can become difficult to manage and may expose sensitive company or customer information.

He argues that vertical AI companies, those focused on a particular industry, can solve this problem by providing a common framework with the necessary integrations, rules and controls already in place.

Instead of asking every retailer to build its own AI fulfillment system, Keeyu wants to provide one that already understands how e-commerce works.

It’s a compelling argument at a time when nearly every software company is trying to add AI features. The ability to generate code is becoming less valuable on its own. Knowing what problems to solve, understanding the systems involved and having access to reliable operational data may prove far more difficult to replicate.

Keeyu still has to demonstrate that its approach can work at scale. Thirty customers and promising early performance figures are a start, but enterprise buyers will want evidence that the system can consistently identify problems, take the right actions and avoid creating new ones.

Still, Rauf is addressing a genuine problem. Most customer service technology is designed to help companies respond to unhappy customers. Keeyu wants to prevent those customers from becoming unhappy in the first place.

And if the company can make that work reliably, it could change how online retailers think about customer service.

Learn more about Keeyu at keeyu.com.

Listen to the full episode of Innovators for my conversation with Tahir Rauf about AI agents, e-commerce operations and building a defensible software business in an age when anyone can generate code.

Discussion about this video

User's avatar

Ready for more?