80% of Fintechs have adopted AI in marketing, but only 29% got results. Here’s the kit that closes the gap.

80% of fintechs have implemented AI across at least one key area. But only 29% say it delivered real results. That gap is not a technology problem. As most teams do not know which workflow to tackle first, or how to connect AI to data that actually means something.

The finance teams closing that gap aren’t doing anything fancy. They are simply connecting tools they already have to attribution data they already own, starting with one use case, and building from there. Here is what that looks like in practice — and a complete kit to help you do the same.

Why most finance teams are stuck despite high AI adoption

The problem is not effort. 62% of marketers cite data quality and fragmentation as their top barrier to AI success (IAB State of Data 2025). AI is only as good as the signals you feed it. If those signals are fragmented, self-reported, or siloed across platforms, AI does not fix the problem — it compounds it.

There is also a speed problem. Banks take an average of 14 months to implement AI. Fraudsters can deploy new tactics in minutes. The teams that move fastest do not wait for perfect conditions. They pick one clear workflow, connect it to clean attribution data, and start.

What finance teams are already doing

Teams across Southeast Asia are running AI on their AppsFlyer data today. Not in pilots. In live workflows with measurable outcomes. Here are two examples:

GCash: monitoring 40+ products without growing the team

GCash is one of the most downloaded finance apps in Southeast Asia, running 40+ unique products and processing millions of data rows every day. Their performance marketing team is lean — and monitoring every metric across every campaign manually was creating blind spots.

They activated Agent Hub within the AppsFlyer platform to continuously monitor attribution data, flagging unusual conversion patterns from media sources before they escalate. One agent runs deep research analysis across campaigns to identify partners where clicks are rising, but lower-funnel conversions like installs and registrations are not. A second agent checks for configuration issues: misaligned attribution windows, toggled settings, anything that could corrupt the data before it reaches a decision.

“What Agent Hub has done for us is essentially give us an immediate response on outlier changes, so we can be very agile in the changes we make to protect the campaign and protect the budget.”

Kimberly Ong, Performance Marketing Technology and Transformation, GCash

The result: 3+ hours freed every week, and fraud anomalies caught earlier than manual review allowed.

Flip: executive-ready insights with a team of three

Flip is one of Indonesia’s leading money transfer apps, with 15+ financial products inside a single app. Their performance marketing team of three handles everything: strategy, execution, and reporting for multiple stakeholders — product marketers and business owners who need performance data but do not use attribution dashboards in their day-to-day.

The ad hoc requests were the breaking point. Business owners would come to the team directly asking for a performance review of a specific product in a specific time frame. Each request consumed time the team did not have.

Their solution: connect AppsFlyer MCP to Claude and build a shared project that any stakeholder can query directly. Inside the project, they loaded standardized instructions, product definitions, conversion benchmarks, and audience context so Claude understands what good performance looks like for each product — and communicates it in plain language, not performance marketing jargon.

“Our ideal vision is to give the stakeholders access to an AI that can act as an extension of our team and give them the reports and insights they need with the same standardized knowledge as our own performance team.”

Asriana Septari, Sr. Digital Marketing Manager, Flip

With AppsFlyer’s MCP, the manual reporting bottleneck is eliminated  and stakeholder-ready insights are delivered automatically.

Want to hear the full stories directly from the teams? Watch the on-demand webinar.

What the teams getting results have in common

The apps that are effectively using AppsFlyer’s AI solutions did not start with a big rollout. They started with one workflow — anomaly detection, funnel reporting, or fraud flagging — and expanded once they had proof it worked.

The most common mistake teams make is waiting for the perfect data setup before trying anything. The teams seeing results did not wait. They used the data they already had, connected it to AI tools already available in their existing package, and ran one prompt on real data.

Ultimately, it’s all about the right signals. AI compounds the value of your data — but only when it pulls from a trusted, connected source. That is the foundation on which everything else builds.

How to start: the Finance AI Marketing Starter Kit

AppsFlyer built a practical kit specifically for finance marketing teams. It’s a working document you open, pick one prompt from, and run on your own data.

What’s inside the kit:

• Ready-to-run prompts: for funnel quality analysis, anomaly investigation, budget reallocation, and fraud signal detection: copy, fill in the brackets, and run in Claude, ChatGPT, or any LLM your team uses
• Real peer examples from: GCash, Flip, and others, showing exactly what they built and what it produced
• A 30-day plan: four weeks, one clear goal per week, from running your first prompt to building a reusable team workflow
• A guide to internal alignment: what to tell your developer, security team, and legal team to move fast without friction

AI Assistant, Agent Hub, and AppsFlyer MCP are all included in your existing package and are currently in beta. There is no added cost to start.

Download the Finance AI Starter Kit

The foundation is already there

The finance teams moving fastest right now are not waiting for a better model or a bigger budget. They are starting with what they have, properly connecting it, and running one prompt on real data.

The tools will keep changing. The AI models will update. But your attribution data, clean, connected, and AI-ready, is the foundation that makes all of it work. That is where to start.