Government Abstract. Baran Ozkan explains how AI-native techniques, false-positive discount, and workflow readability are redefining how establishments scale regulated operations with out dropping audit defensibility.
Monetary crime compliance is shifting from rule-heavy oversight to operational infrastructure. As fintech and banking techniques scale in complexity, establishments are being pressured to rethink how monitoring, investigations, and audit readiness operate in actual time.
Baran Ozkan, co-founder and CEO of Flagright, works on the intersection of transaction monitoring structure, AI-native compliance tooling, and operational workflow design. His perspective comes from constructing inside high-growth monetary techniques the place alert quantity, investigation pace, and regulatory defensibility collide.
This dialog focuses on operational readability, false-positive discount, AI-native compliance techniques, and what it takes to scale regulated infrastructure with out dropping audit self-discipline.
AITJ: Baran, what was the tipping level that impressed you to construct Flagright and the way did your previous roles at TransferGo and Forto form that imaginative and prescient?
The tipping level was an extended, irritating seek for a real-time, risk-based transaction monitoring resolution once I was main product at TransferGo. I used to be near the fact of how compliance groups function, particularly when a enterprise is scaling quick throughout corridors and buyer segments. What I noticed was an enormous hole between what legacy distributors promised and what really labored in manufacturing. Integrations have been gradual, tooling was inflexible, alert high quality was poor, and when groups wanted to alter one thing rapidly, they typically couldn’t with out prolonged vendor cycles. That’s when it clicked for me that compliance tooling was holding fintechs again, not as a result of groups didn’t care, however as a result of the infrastructure was not designed for contemporary monetary merchandise.
TransferGo formed my view of the issue from a fintech operator perspective. If you’re shifting cash for actual prospects, you can not afford brittle techniques that create huge queues of low high quality alerts. You want controls that may adapt as your product modifications whereas nonetheless staying defensible for regulators and companions. That rigidity between development and management is everlasting, and I needed to construct a platform that lets firms scale with out feeling like compliance is a continuing drag on the enterprise.
Forto formed my view from a techniques perspective. In logistics, you study rapidly that actual world operations are messy and information is rarely excellent. You might be constructing actual time techniques which have to remain dependable beneath strain, and also you want observability, clear information fashions, and clear workflows if you need selections to be repeatable. That have strengthened a core perception I carried into Flagright: the reply just isn’t extra guidelines or extra alerts. The reply is operational readability and an structure that makes good selections potential at scale.
The 180 web page Monetary Crime Compliance Handbook you launched is bold. What’s one perception in that handbook you consider might essentially shift how compliance groups function as we speak?
One perception I hold coming again to is that almost all compliance applications don’t fail as a result of groups have no idea the foundations. They fail as a result of the working system behind this system is weak. Groups might have insurance policies, coaching, and a set of controls, however they lack the sensible workflows, proof trails, and suggestions loops that make this system resilient each day.
Within the handbook we tried to translate regulation into operations. Which means being specific about what a very good investigation appears to be like like, what info ought to be captured, how selections are documented, how high quality assurance is run, and the way reporting is ready in a repeatable means. When groups deal with compliance as an working system quite than a guidelines, two issues occur. First, they cut back noise, as a result of controls grow to be measurable and tunable. Second, they grow to be audit prepared by design, as a result of proof is produced naturally via the workflow quite than retroactively assembled when an audit is looming.
If I needed to summarize it in a single sentence, it’s this: compliance maturity is much less about figuring out extra rules and extra about constructing a system that produces constant selections and defensible proof at actual world pace.
You typically talk about operational readability over regulatory overload. How do you outline operational readability and why is it lacking in most fincrime methods?
Operational readability means a compliance group can reply a couple of easy questions at any time limit with out guesswork.
- What dangers are we really seeing in our enterprise proper now
- Which controls are firing and why
- What’s the high quality of these alerts
- How lengthy does it take to research and resolve them
- The place are the bottlenecks
- What modified this week and what influence did it have
When a group has operational readability, they don’t seem to be reacting emotionally to quantity. They’ll see sign and development. They’ll defend why they made selections. They’ll tune controls primarily based on proof, not instinct.
It’s lacking as a result of the business has traditionally optimised for regulatory protection, not operational efficiency. Many stacks are fragmented, with one software for monitoring, one other for investigations, one other for reporting, and plenty of guide glue in between. Knowledge will get siloed. Guidelines get modified with out good testing. Fashions get deployed with out clear reasoning. The result’s a program which will look nice on paper, however feels chaotic in follow.
My view is that one of the best compliance groups function like excessive performing product and engineering groups. They measure, they iterate, they take a look at on actual information, and so they constantly enhance. Operational readability is what permits that mindset to work in compliance.
Are you able to stroll us via the journey of serving to a shopper cut back their false optimistic price and what made that transformation potential?
Let me floor this in a public instance. In a single shopper deployment, false positives dropped from 99.1 % to fifteen.3 %, which is an 83.8 % discount, and investigation time fell by greater than half. What made that transformation potential was not a single magic mannequin. It was a mixture of knowledge, workflow, and management design working collectively.
First, we targeted on controllability. The shopper wanted to have the ability to change detection logic with out ready weeks for vendor intervention. They have been working throughout a number of jurisdictions, and when fraud patterns shift, time issues. We gave them the power to create and modify customized guidelines independently, and to check configurations on reside manufacturing information with out impacting actual alerts. That capacity to experiment safely is a giant a part of how groups cut back noise with out rising danger.
Second, we targeted on transparency. If an alert fires, an analyst wants to know the explanation rapidly. On this case, the shopper had expertise with proprietary techniques that behaved like a black field, which made it arduous to justify selections. By making triggers and context clearer, analysts can work quicker and high quality improves as a result of selections are simpler to validate.
Third, we targeted on workflow effectivity. When groups have to leap between techniques to know one case, investigation occasions inflate and high quality suffers. Unifying monitoring and case administration helps analysts transfer via circumstances with much less friction and extra context, which straight improves throughput and consistency.
What this says about legacy techniques is that many are constructed round vendor management and static assumptions. They typically make change costly and gradual, and so they produce alert quantity that appears like protection however doesn’t translate into actual danger discount. When groups reclaim management, measure outcomes, and iterate on reside information, efficiency improves quick.
Flagright positions itself as an AI native resolution for compliance. How do you differentiate AI native from AI enabled and what benefits does that ship?
To me, AI enabled typically means AI has been bolted onto an present product. You may need a chatbot interface, a mannequin that runs as an non-compulsory characteristic, or a workflow that also is determined by guide work for many outcomes. The inspiration stays the identical, and AI is a layer on prime.
AI native means AI is designed into the system from the start. It impacts how information is structured, how selections are made, how workflows are automated, and the way outputs are ruled. It additionally means the system is constructed to study from investigations, as a result of the suggestions loop is the place the worth compounds.
The actual world benefit just isn’t theoretical. When AI is native, it might probably cut back low worth guide work in investigations, floor context that analysts would in any other case spend time assembling, and assist groups produce constant narratives and proof trails. It additionally permits groups to scale with out merely hiring extra individuals to clear alerts. In compliance, scale with out dropping defensibility is the entire sport.
With the handbook overlaying regional breakdowns throughout the EU, US, APAC, Center East, and Africa, what was essentially the most shocking nuance you encountered?
Probably the most shocking nuance was how a lot of the true complexity comes from variations in operational expectations, not from the headline rules. Many areas share the identical broad objectives, and world requirements affect them, however the best way groups are anticipated to operationalize these objectives varies.
For instance, reporting regimes can differ in what’s required, how info is structured, and the interior proof groups must retain to defend the choice. Knowledge privateness expectations additionally create very sensible design constraints, particularly if you find yourself working throughout areas and must steadiness info sharing with buyer protections.
What that taught me is that the successful strategy is to not construct a separate compliance program for each area from scratch. It’s to construct a core working mannequin that’s constant, after which make the regional variations specific in workflows, documentation, and management configuration. That’s what makes scaling potential with out turning compliance right into a patchwork.
Out of your perspective as each a builder and an investor, what indicators do you search for in startups fixing advanced regulated issues?
The primary sign is respect for the area. In regulated industries, the product isn’t just a characteristic set, it’s a promise of reliability and accountability. Founders want to know the operational actuality of the groups they serve, not simply the regulation in summary.
The second sign is proof orientation. I search for groups that may clarify how their resolution will probably be measured in manufacturing, what outcomes they may enhance, and what tradeoffs they’re making. In compliance, there is no such thing as a room for imprecise claims. You must present the way you cut back noise, enhance detection, and strengthen audit readiness.
The third sign is belief maturity. Safety posture, governance, and accountable deployment will not be good to have. They’re the product. If a group treats these as an afterthought, they may battle to win severe prospects.
Lastly, I like founders who’re affected person about distribution. In regulated markets, belief and adoption compound over time, and one of the best founders construct for the long run quite than chasing brief time period hype.
Flagright emphasizes a no code, API first platform. What have been the hardest challenges in sustaining simplicity with out sacrificing depth?
The toughest problem is that simplicity and depth typically pull in reverse instructions. Compliance groups want expressive energy as a result of actual danger is nuanced. On the similar time, if the system appears like a posh engineering software, you lose the individuals who need to function it day-after-day.
Technically, constructing a no code situation builder that’s highly effective and protected is tough. You want guardrails so groups don’t by chance create controls that explode alert quantity or create gaps. You want robust testing functionality so modifications could be validated earlier than they go reside. You want efficiency so the system can rating exercise rapidly, and also you want observability so groups can perceive how modifications have an effect on outcomes.
Philosophically, you must be prepared to say no to complexity that doesn’t create actual worth. A variety of enterprise software program provides options that create choices, not outcomes. We attempt to hold the psychological mannequin clear. A superb management ought to be comprehensible, testable, measurable, and defensible.
When you needed to future proof compliance for the subsequent 10 years, what three improvements would you guess on?
First, steady controls with simulation and measurement. Groups will more and more take a look at modifications on actual information earlier than deploying them, measure influence on false positives and true danger detection, and deal with compliance as a dwelling system.
Second, privateness preserving collaboration throughout establishments. Monetary crime is networked, and protection must grow to be extra networked. Higher methods to share typologies, indicators, and patterns with out exposing delicate information will elevate the baseline for the entire ecosystem.
Third, AI assisted operations with robust governance. AI will assist investigations, triage, documentation, and high quality assurance, however provided that establishments can clarify selections, monitor efficiency, and hold accountability with people. The winners will mix AI functionality with audit readiness by design.
What private philosophies or management ideas do you come back to when dealing with ambiguity or resistance?
I come again to 3 ideas.
Be unreasonably buyer obsessed. If I’m uncertain what to do, I am going speak to the individuals dwelling the issue. Actuality is one of the best technique doc.
Take duty for safety, accuracy, and outcomes. In compliance, belief is fragile. If we ship one thing, we personal it, together with the arduous edge circumstances.
Transfer with urgency, however don’t confuse pace with recklessness. I consider in iteration and motion, but in addition in constructing the proof and governance that make quick techniques protected.
After I face resistance, I attempt to separate sign from noise. If a number of shoppers and practitioners say the identical factor, that’s sign. If the pushback relies on worry of change quite than information, we acknowledge it however we don’t let it drive the roadmap.
