ServiceNow AI Agents for Scam Response: Governed with AI Control Tower and ServiceNow CSM

Bill Martin · 4 hours ago

Hi everyone,

Our team, Somtum Intelligence from MFEC, built InterveneNow, a scam response prototype on ServiceNow AI, and was named 1st Runner Up at Asia HackNow 2026 in Singapore, placing second out of 70 teams across Asia Pacific. I want to share how we combined ServiceNow AI agents, ServiceNow CSM and ServiceNow AI Control Tower so that AI governance is built in from the start, not added at the end.

The problem: online scams move faster than response

Online scams and financial fraud are growing, and victims lose money in minutes. Today, the work after someone tries to report a scam is split across many teams: intake, containment, tracing money mule accounts and cross border referral. By the time the case reaches the right person, the money has moved. Fraud prevention needs one connected, governed journey.

What we built with ServiceNow AI agents

The demo video below has no narration, so here is what you are seeing:

Report a scam: A citizen opens the scam reporting portal and chats with Now Assist in ServiceNow Virtual Agent. A guided topic asks which transactions were unrecognised and whether the card has been blocked, then creates a case in ServiceNow CSM.
AI reads the report: In the CSM Workspace, Now Assist summarises the case and an AI step extracts the scam facts: amount, receiving account, transaction date, scam type, warning signs and missing information.
Contain: A ServiceNow AI agent recommends whether to hold the receiving account, with its reasons, a confidence level and a repeat victim check against earlier cases. It drafts the hold request but does not send it. The officer has to approve first.
Trace and Customer 360: A Financial Crime Analyst gets Customer 360 insights from the Now Assist panel, and an AI fund trace follows the money hop by hop through money mule accounts.
Refer to partner agency: AI drafts a cross border referral so the funds can be frozen in the destination country. The officer reviews and approves it before it goes out.
Resolve: AI writes the resolution notes and a plain language update for the victim. The update covers what was held, the reference number, refund timelines and advice on staying safe.
AI risk management in ServiceNow AI Control Tower: The containment agent is registered as an AI asset with its purpose, models, tools and prompts. It goes through a review, evaluation and approval playbook and receives a High risk classification, because it can affect a citizen's money.
Close the loop: A Compliance Officer reviews the full audit trail of what the AI recommended and what each human approved.

AI agent governance by design

Human in the loop AI: Every ServiceNow AI agent that touches money or an external party only recommends and drafts. Nothing runs until the officer approves.
Explainable AI: Each recommendation comes with its reasons and a confidence level, so officers can see why the AI suggests what it suggests.
AI governance before deployment: Agents are risk classified and approved in ServiceNow AI Control Tower before they are used, not after.
Plain language for citizens: Victims never see internal jargon. Updates are written for a stressed person, not an analyst.
Reuse before building: We used out of box ServiceNow CSM, ServiceNow Virtual Agent, Now Assist and AI Control Tower wherever possible, and only built custom AI agents for steps the platform doesn't cover.

What is real and what is simulated

This is a hackathon prototype. Bank, police and cross border partner integrations are simulated, and the case uses fictional data. What we proved is the end to end journey, the human approval points and the AI governance model on ServiceNow AI.

Lessons learned

Start from the victim's journey, then map ServiceNow AI capabilities onto each step. That keeps the solution focused on outcomes, not features.
Put the approval gates into the design from day one. Adding them later is much harder.
Register agents in ServiceNow AI Control Tower early. It makes the AI risk management conversation with judges, customers and regulators much easier.

Thank you

Thank you to my teammates Oil, for the citizen portal, slides and branding, and Khaw, for the backend and automation. Thank you also to MFEC for backing the team, and to the ServiceNow Asia HackNow 2026 organisers, mentors and judges.

I would love feedback from anyone building ServiceNow AI agents for financial services, public sector or AI governance use cases. How would you extend this pattern in your organisation?