Inception and Visa test whether AI agents can be trusted with real money

Inception, the Abu Dhabi–based AI company backed by G42, has announced a strategic partnership with Visa aimed at accelerating what the companies describe as agentic commerce across Central and Eastern Europe, the Middle East, and Africa (CEMEA).

At a surface level, the announcement is about integrating Inception’s AI systems with Visa Intelligent Commerce, allowing software agents to discover products, make purchasing decisions, and complete payments autonomously. But beneath that framing is a more consequential question: whether financial systems are ready to treat AI agents as legitimate actors inside payment networks.

Until now, most AI-led commerce has stopped short of execution. Recommendation engines suggest what to buy, chatbots guide users through choices, and automation helps with internal workflows. Payments, however, remain tightly controlled, deliberately manual, and surrounded by layers of verification.

This partnership is an attempt to change that boundary.

Moving payments closer to automation

Under the agreement, Inception will integrate Visa Intelligent Commerce into its agentic systems, combining its domain-specific large language models and enterprise AI components with Visa’s payment infrastructure. That includes tokenisation, authentication mechanisms, and agent identity frameworks designed to allow non-human actors to transact securely on Visa’s network.

The emphasis from both companies is on controlled experimentation rather than rapid rollout. The collaboration includes co-development, pilot programmes, and client demonstrations, suggesting that neither side sees agent-driven payments as ready for widespread deployment.

Ashish Koshy, CEO of Inception, framed the partnership around trust rather than speed.

“Agentic commerce is reshaping how transactions happen, and working with Visa ensures that these AI-driven journeys are built on a trusted payments foundation,” he said, adding that the goal is to equip merchants and platforms with tools that can support intelligent commerce at scale.

Visa’s position reflects a similar caution. “Trust will be the core requirement for these experiences to scale,” said Godfrey Sullivan, Head of Product and Solutions for CEMEA at Visa. “By working together, we are enabling organisations across the region to test, learn, and roll out secure agent-driven shopping and payment experiences”.

Why this matters beyond the press release

Allowing AI agents to initiate and complete payments is not just a technical shift. It changes how responsibility, risk, and accountability are distributed across commerce systems.

If an AI agent makes a purchasing decision, who is liable when something goes wrong? The merchant, the platform, the model developer, or the payment network? These questions are unresolved, and they become harder in regions like CEMEA, where regulatory approaches to AI and digital payments vary widely.

For Gulf economies pushing automation and productivity, agentic commerce fits neatly into broader digital transformation agendas. For emerging markets, it raises concerns around oversight, consumer protection, and system resilience.

Visa’s involvement suggests that global payment networks are beginning to plan for a future where transactions are not always initiated by humans. Inception’s role reflects a parallel trend in enterprise AI, where automation is moving from analysis and recommendation into execution.

A cautious step, not a conclusion

This partnership does not signal that AI agents are about to start spending freely across global commerce systems. The language of pilots and readiness efforts makes that clear.

What it does signal is a shift in where experimentation is happening. Instead of building agentic systems outside financial infrastructure and asking payments to catch up later, companies like Inception are trying to embed these systems directly into existing networks.

Whether that approach succeeds will depend less on AI capability and more on governance, controls, and trust. Payments have historically been the last layer to automate for a reason.

This collaboration is an early test of whether that reason still holds.