Antom upgrades AI Copilot to streamline merchant payment operations
Ant International's Antom has rolled out major upgrades to Antom Copilot, its AI agent for merchant payments and operations. The agent now lives inside the Antom Merchant Portal and automates routine work while handling more complex operational flows with context from real cases.
For operations leaders, this points to faster onboarding, fewer support tickets, tighter risk oversight, and more predictable payment outcomes. The pitch is simple: give teams a capable agent that gets things done across the payment lifecycle, then measure the lift.
What's new in Antom Copilot
- End-to-end support across onboarding, integration, dispute resolution, risk management, and payment success rate operations.
- Domain-specific intelligence that works like a virtual team of payment specialists to reduce manual effort and costs.
- Stronger multi-turn dialogue and memory to resolve issues across multi-step operational environments.
- Proactive signal monitoring to flag risks and opportunities before merchants raise them.
- Upgraded agentic RAG architecture with dynamic SOP planning. If you need a primer on RAG, see this overview from IBM here.
- Reported outcomes: 90% reduction in merchant payment integration time and a 46% boost in dispute-handling efficiency (per Antom).
Why operations teams should care
- Onboarding and integration timelines shrink, freeing engineers and support from repeat handoffs.
- Dispute queues get thinner as the agent resolves multi-step cases with fewer escalations.
- Risk monitoring tightens with earlier alerts and clearer playbooks.
- Payment success rate work becomes more systematic, with SOP-driven tuning and follow-up.
How to put it to work in your org
- Enable Copilot in the Antom Merchant Portal and grant least-privilege access for ops, risk, and support roles.
- Map current SOPs into the agent's dynamic SOP planning. Start with onboarding, disputes, and payout exceptions.
- Set hard KPIs: integration lead time, dispute turnaround, payment success rate, auto-resolve rate, manual-override rate, MTTR.
- Run a 30-day pilot on a limited merchant cohort. Compare baseline vs. agent-assisted results weekly.
- Instrument logs: keep every agent action auditable (inputs, SOP step chosen, outcome, human approvals).
- Define guardrails: PII handling, escalation thresholds, high-risk scenario routing, and human-in-the-loop triggers.
- Train the team on prompts that reflect real tickets and edge cases. Refresh SOPs monthly based on agent transcripts.
Architecture notes
Antom Copilot runs on an upgraded agentic retrieval-augmented generation stack with dynamic SOP planning. In practice, that means the agent retrieves the right policy or guide, plans a step-by-step workflow, executes, and asks for human approval where needed.
This approach cuts context switching for operators and enforces consistent decision paths. It also makes post-incident reviews cleaner because every step is traceable.
EPOS360 for SMEs
Antom also launched EPOS360, an AI-based application that brings together POS systems, banking, payments, lending, and growth support in one place. For lean ops teams, a single console reduces vendor sprawl and consolidates reporting and controls.
Governance checklist
- Access control: role-based permissions, audit trails, and separation of duties.
- Data: redact PII in training logs, define retention windows, and restrict export paths.
- Quality: weekly review of agent actions with spot checks on declines, disputes, and chargeback flows.
- Continuity: document fallbacks for agent outages, including manual SOPs and escalation trees.
Strategy from Ant International
Antom Copilot sits within Ant International's broader push into agentic payment technologies and AI-driven commerce. The company reports training and tuning with deep industry cases and new agent protocols for efficient task orchestration.
Jiangming Yang, chief innovation officer at Ant International, said: "Ant International is focusing on building domain-specific agentic AI that solves global payment complexity for merchants. We train and tune AI agents using deep industry know-how and a large set of anonymised complex cases, and evolve new agent protocols to ensure agents solve tasks in efficient orchestration."
Want to upskill your ops team?
If you're building AI-driven workflows and SOPs, this curated track can help your team get practical fast: AI Automation Certification.
Your membership also unlocks:
