Agentic Commerce Is Here: How AI Agents Can Transact, Book, Pay, and Operate
I demonstrated eight ways an AI assistant can help teams arrange services, gather information, and prepare work for review rather than simply answer questions.
AI Summit Seoul · South Korea · 19 August 2026
Eight Useful Applications
The demonstrations follow everyday tasks from the initial request to something a person can use or check. Business analysis used sample data, and the expert brief included a synthetic voice note; these were demonstrations, not findings about a real business.
- Bookings and payments: arrange a reservation across email and calendar, or buy a report and check the delivered file.
- Travel planning: fit meetings, dietary needs, and mobility constraints into an itinerary.
- Opportunity research: check public tenders and support programs for suitability, deadlines, and eligibility.
- Business questions: use approved company data to prepare answers for review.
- Company knowledge: find relevant internal documents and flag conflicting policies.
- Expert preparation: collect facts, documents, and open questions into a brief for a professional.
- Finance review: compare accounting records and group discrepancies for the finance team.
- Testing: check saved examples without making another booking, payment, or record change.
Coordinating Services and Checking the Result
Delegation Across Booking Services
An OpenClaw coordinator delegated work to specialist agents. Chope, Gmail, and Google Calendar supported the reservation path, with time zone checks connecting the booking to its communication and calendar context.
Research with Bounded Data Access
GeBIZ and Enterprise Singapore supplied opportunity sources for checking suitability, deadlines, and eligibility. Business analysis used approved BigQuery queries, while Notion retrieval stayed within approved documents and exposed conflicting policies. Public research and company retrieval therefore used different information boundaries instead of a single unrestricted search.
Reconciliation Before Finance Review
The finance demonstration reconciled Xero records against the trial balance and grouped unresolved differences for review. The analysis used sample data, and the expert brief used a synthetic voice note.
Replay Without Repeating External Actions
Saved cases were replayed without making another booking, payment, or record change. This separated checking behavior from repeating its external effects. The demonstrated release checks covered seven scenarios, 14 observed turns, and 10 reconciliation checks; these counts describe the internal demonstration checks.