Back
AI Operations Engineer, CX
Full-time
Team: CX Operations \| CX AI Location: Singapore, Hong Kong, Philippines, APAC. You will need meaningful daily overlap with Hong Kong time (UTC\+8). Reports to: AI Operations Lead, CX
About Reap
Reap is a leading global payment technology provider that enables financial connectivity and access for businesses worldwide. By merging traditional finance with digital assets, bridging disparate economies, and connecting key financial players, we are transforming the financial landscape into a more interconnected and interoperable space for efficient money movement.
With stablecoin-enabled corporate cards, payout solutions, and expense management tools, we streamline financial operations and empower businesses to scale. Our APIs enable businesses to embed finance into their own products and services, from issuing Visa cards to facilitating cross-border payments.
Reap is supported by a strong network of investors, including Acorn Pacific Ventures, Arcadia Funds, HashKey Capital, Hustle Fund, Fresco Capital, Abacus Ventures, and Payment Asia.
About The Role
Our customers get support from an AI agent. Not a chatbot with a decision tree — a genuine agent that reads account state through live tool calls, follows structured workflows, and resolves real financial-services problems end to end.
You will own how well it works.
The single number that matters is Independently Resolved Rate (IRR): the proportion of customer conversations the AI agent resolves completely on its own, without a human stepping in. It sits at 39% today with 3 months of uptime so far. We are going to 80% across every channel, and we are going to get there without degrading the quality of a single customer interaction.
That is the job. Find out why the agent fails, fix the cause, and build the tooling that makes the next fix faster than the last.
This is a hands-on role in a small team. You will not be handing recommendations to someone else to implement — you will be on the platform, writing the workflow, testing it, and shipping it to production. If you like owning a metric and having the keys to move it, this is a very good seat.
An honest word about the shape of the work. The centre of this job is iterative: root cause analysis, QA, workflow rewrites, testing, shipping, measuring — again and again, until the number moves and stays moved. That loop is the craft, and we need someone who genuinely enjoys it. Around that core, the role deliberately grows outward: you will build AI-powered tools for the wider CX and Operations organisation, and you will own the improvement strategy for your business line — not just execute one handed to you. Expect roughly 80% of your time in the iterative loop and 20% in building and strategic work at the start, shifting toward 70/30 as your first business line reaches business-as-usual. If you want a mostly-strategic seat, this is not it. If you want a seat where strategy is earned from the work, keep reading.
What You Will Own
Analyse
Dig into ticket-level and aggregate performance data to find where and why the AI agent fails to resolve independently.
Build and maintain the reporting that makes IRR movement legible — by business line, by channel, by intent, week over week.
Run root cause analysis on failed resolutions and turn findings into a prioritised, evidence-backed improvement backlog.
Improve and ship
Design, write, and deploy natural language workflows that raise resolution rates while holding our standards for tone, accuracy and compliance.
Tune guardrails, business contexts, and escalation logic so the agent knows both what to do and when to stop.
Test rigorously before production — simulation runs, scenario testing, staged rollout, and a rollback plan that actually works.
Measure the effect of every change. If a shipped improvement did not move the number, say so and iterate.
Keep it current
Own the freshness and structure of the knowledge sources the agent reasons over. In a company that ships this fast, stale knowledge is the quietest and most expensive source of failed resolutions.
Maintain the configuration registry so that anyone can understand what is live, why, and who changed it last.
Collaborate on the tool layer
Work with Platform Engineering on the MCP integrations that give the agent access to account, card, transaction and payment data.
Specify what the agent needs, identify gaps and data quality problems in existing tools, and pressure-test new tool surfaces before they reach production.
Write and debug low-code tool functions in JavaScript to extend agent capability where a platform integration does not exist yet.
Build for CX and Operations, not just the agent
Build the internal automation and AI-powered tooling that makes the improvement loop faster: issue-trend monitoring, automated QA of AI-handled tickets, knowledge-staleness alerting, and outage-to-agent awareness so the agent knows about an incident before customers start asking.
Extend that same capability to the wider CX and Operations teams — internal AI tools, workflow automations, and data products that remove manual effort well beyond the support agent itself. Part of the point of this role is reducing our dependency on engineering capacity for CX tooling: where today an internal tool waits in a platform queue, tomorrow you build it.
Treat your own productivity, and your team's, as an engineering problem.
Own the direction, not just the queue
Set and own the improvement roadmap for your business line — you write the plan, defend the prioritisation, and report the results to CX leadership.
Contribute to where the AI programme goes next: expansion sequencing across business lines, what we measure and why, and where AI can serve CX and Operations beyond support tickets.
From your second quarter, co-shape the playbook for the next business line expansion, informed by what worked and what didn't on your first.
Your first six months
Months 1–3 — Own a Product, Reap Business Account
Ramp on Lorikeet, our MCP tool surface, our workflow standards and our ticket taxonomy.
Take end-to-end ownership of the AI agent for our Reap Business Account line: workflows, knowledge, guardrails, tooling, performance.
Drive IRR on that line from 40% toward 60%, evidenced by shipped improvements and week-over-week movement, not one-off spikes.
Establish the operating rhythm — a repeatable weekly loop of analyse, diagnose, ship, measure — and own the improvement roadmap that the loop feeds.
Months 4–6 — Extend and harden
Reach and hold 60% IRR on Reap Business Account, and take it to business-as-usual: documented, monitored, stable without daily intervention.
Extend the operating model to our other business lines, adapting rather than copying — and co-author the expansion playbook as you go.
Take the tooling you built in the first phase from minimum viable to genuinely production-grade — reliable, documented, and used by oth
