AI Sandbox
Airwallex processes billions of dollars across 150+ countries for over 100,000 businesses. At that scale, the operational challenges are real, complex, and consequential. We are addressing some challenges internally, while making others accessible to innovative AI startups. As a proof-of-concept program, AI Sandbox puts your product in front of ambitious problem statements. You will get the opportunity to work on live Airwallex challenges, along with real data, real stakes, and a real path to enterprise deployment. Let’s build together.
The Mission
As part of the Airwallex for Startups program to solve real-world internal challenges with proactive AI-first solutions.
Internal Challenges
We identify high-impact operational bottlenecks within Airwallex that require intelligent automation.
Proactive Approach
We don't just wait for solutions. We actively seek out the best AI-first startups to build them.
Enterprise Path
Successful pilots are fast-tracked through our internal procurement and security frameworks.
The 5 Program Themes
1. GTM Intelligence & Revenue Productivity
The best sales teams in the world are still operating on instinct.
B2B sales has always been a craft. The best reps know which accounts to prioritise, when to reach out, how to tailor a pitch, and when to walk away. They build this intuition over years through pattern recognition, relationship-building, and a lot of wasted calls that taught them something. With time and experience, that instinct lives in their heads, but not in the system. Ultimately, we’re looking for ways to compress ramp time from new hire to top performer.
What we're looking for:
Autonomous account research and deal profiling that surfaces what matters before a rep picks up the phone
Smarter territory and account prioritisation based on trusted signals
Conversation intelligence that coaches in real time and makes every rep better, faster
Outreach that adapts based on what's actually converting
2. Customer Intelligence & Retention
Most companies only see a fraction of the value sitting in their customer base.
Every customer interaction generates signals such as usage patterns, transaction behaviour, support sentiment, feature adoption. Together, they tell you who's ready for more, who's about to churn, and who needs a different experience entirely. But that signal is scattered across systems that don't talk to each other. The companies that will win at customer lifetime value in the next five years will be the ones that read the full picture and act on it in real time.
What we're looking for:
Lifecycle intelligence that maps each customer's trajectory and recommends the next best action
Churn prediction models that combine product usage, payment behaviour, and support data to intervene early
Personalised engagement triggers that adapt messaging, offers, and touchpoints based on where a customer actually is, not where a playbook assumes they are
Real-time alerting that routes the right insight to the right team at the right moment
3. Financial Operations & Treasury Automation
Traditional ERP tools break down at the complexity of global firms.
Reconciliation becomes a sprawling, manual exercise when you're operating across dozens of entities, currencies, and banking relationships simultaneously. The data exists, and the systems are increasingly connected. What's missing is the intelligence layer that sits above them. One that understands the structure of a complex, global business and can reason across it.
What we're looking for:
Multi-entity cash reconciliation that actually works across currencies and jurisdictions
Automated variance analysis that surfaces what matters, not everything
Liquidity forecasting and management beyond spreadsheet models
Intelligent T&E compliance that catches issues before they compound
4. Financial Crime & Compliance Automation
Rule-based detection and manual review can't keep up with the volume and sophistication of modern financial crime.
Every licensed financial institution runs compliance operations at scale. AML screening, transaction monitoring, merchant risk assessment, and more. The regulatory bar keeps rising and the cost of getting it wrong is becoming more severe, but the tools haven't fundamentally changed. AI has the potential to fundamentally restructure this by moving from static rules and manual review to adaptive, intelligent systems that triage automatically, investigate autonomously, and escalate only what genuinely requires human judgment.
What we're looking for:
Autonomous triage and resolution for AML and transaction monitoring alerts
Smarter name screening that reduces false positives without increasing risk
Merchant risk scoring that adapts as patterns evolve
Systems that can explain their reasoning to regulators, not just flag anomalies
5. GTM Intelligence & Revenue Productivity
Every legal team is drowning in the same work. Almost none of it requires a lawyer.
Contract review. NDA redlining. Regulatory policy drafting. Legal research across jurisdictions. These are the tasks that consume the majority of in-house legal time at any fast-growing company. The challenge isn't that legal AI doesn't exist. It's that most of it is built for law firms and optimised for billable hour workflows. A legal AI platform purpose-built for regulated industries is still largely unbuilt.
What we're looking for:
NDA and vendor contract redlining that understands fintech-specific terms and risk
Regulatory policy drafting across multiple jurisdictions
Multi-jurisdiction legal research that actually synthesises, not just searches
Contract lifecycle management built for speed and scale
What you get
- Real data and systems. Work with production-representative data and the internal teams who know it best.
- A dedicated internal sponsor. Every selected startup is paired with an Airwallex leader who owns the problem area and has the authority to champion your solution internally.
- A defined PoC timeline. Structured engagement with clear milestones and alignment on what success looks like from day one.
- A path to deployment. If the PoC delivers, there's a defined route to a paid pilot and vendor contract.
- Airwallex for Startups perks. Selected startups get access to Airwallex's startup program including credits, global account infrastructure, and the tools to scale your own business across borders.
- Visibility. Work with a company processing billions globally. A successful engagement is a proof point that opens doors beyond Airwallex.
How it works.
Application & Selection
Submit your product's fit for a specific challenge theme. Airwallex will assess your team's strength and product maturity to determine if you're the right partner for the problem.
Scoping the PoC
Work directly with an internal sponsor to define the Proof of Concept (PoC). This includes setting success criteria, data requirements, integration points, and a firm timeline.
Build & Test
Execute your PoC against real-world Airwallex challenges. Throughout this phase, you will receive ongoing feedback from the team that owns the problem.
Evaluation & Integration
Results are assessed against the agreed success criteria. If the PoC is successful, the process moves forward toward potential full-scale integration.
Ready to build something real?
AI Sandbox applications are now open for a limited time. Tell us which challenge you'd take on and why your team is the one to solve it.
You're a good fit if:
- You're an AI-first startup with a working product
- Your solution maps clearly to one of the five challenge themes
- You can commit engineering time to a structured PoC
- You're looking for a design partner who'll push your product, not just validate it.
We're open to startups at any stage, but you need to be ready to deploy — not just demo.
