Kinectify’s use of Azure OpenAI to draft suspicious-activity narratives puts generative AI inside one of gaming’s most sensitive workflows, but CEO Joseph Martin’s central assurance is also the line that determines whether the product is a compliance aid or a compliance liability: the software may assemble and write the case, but a trained investigator must make the decision and approve the filing.
That distinction matters more than the usual AI marketing gloss. A Suspicious Activity Report in the US, or a Suspicious Matter Report in Australia, is not a customer-service reply that can be corrected after the fact. It is a regulated submission based on a judgment that a transaction or pattern may involve crime. FinCEN’s casino guidance requires risk-based controls to identify, analyze and report suspicious activity, while AUSTRAC expects reporting entities to review relevant material generated by their monitoring programs. The accountable operator—not its cloud vendor and not its language model—owns that judgment.
In an interview with iTWire, Martin said Kinectify’s model is used to turn structured investigative facts into an executive summary or draft report. Microsoft’s own Kinectify customer story confirms that the company uses Azure OpenAI Service for executive investigation summaries and government filings, particularly SARs. Microsoft also says Kinectify’s Azure platform can process as many as 40 million transactions daily.
The consequential detail is what Microsoft’s account does not say: it does not independently describe Kinectify’s human-review controls, the fields passed to the model, or whether the generated narrative is technically prevented from introducing unsupported material. Martin’s statement that the filing remains a draft is therefore the key operational claim—and one customers should verify in product configuration, audit logs and contract language before treating AI-generated narratives as production-ready compliance records.
Kinectify’s pitch is grounded in a plausible division of labor. A gaming operation can deterministically collect and correlate customer identity data, machine activity, cash-ins, cash-outs, ATM use, loyalty records and prior investigative notes. A large language model can then convert selected facts into readable prose. The human investigator reviews the evidence, decides whether suspicion is reasonable, edits the narrative and approves the submission.
That workflow is materially different from asking an LLM to decide whether a patron is laundering money. Martin told iTWire that Kinectify supplies the model with structured facts rather than asking it to discover knowledge from its broader training data. If implemented rigorously, that reduces one obvious risk: a model inventing a transaction, customer attribute or investigative conclusion that was never in the case record.
But reduces is not eliminates. Grounding controls do not automatically solve the problem of wrong source data, missing context, faulty entity matching, stale customer records, or a model that overstates the implication of an otherwise accurate transaction pattern. The most useful audit question is not whether the model is “hallucinating” in the abstract. It is whether every claim in the proposed narrative can be traced back to a specific transaction, record, alert, investigator finding or attached document.
FinCEN’s casino guidance makes the importance of that traceability clear. It asks casinos to maintain supporting records and to ensure SAR narratives clearly explain who, what, when, where, why and how. A polished AI narrative that cannot be traced back to evidence is worse than a clumsy manual draft: it creates a false appearance of confidence in a filing that may later be examined by regulators, auditors or law enforcement.
Kinectify says its investigators can edit the drafts and that actions are logged. For buyers, the practical test is whether the system preserves the original generated text, the source evidence, every edit, the approving user and the exact final version filed. Without that chain, “human in the loop” becomes a slogan rather than a defensible control.
Those statements can coexist if “fully automated” means automated preparation and workflow rather than autonomous suspicion determinations. But the wording leaves too much room for interpretation in a regulated setting. A casino compliance officer evaluating the platform should insist on a precise answer: Can the software submit a SAR or SMR without an authorized human explicitly approving the matter and the final narrative? If it can, does the vendor recommend that workflow? If it cannot, the marketplace language should be read as sales shorthand rather than a description of an unattended filing process.
The distinction is not theoretical. FinCEN places the obligation on casinos and card clubs to establish and implement risk-based internal controls. Its guidance also stresses documentation of decision-making, including decisions not to file. Kinectify can accelerate the evidence-gathering and writing stages, but it cannot transfer regulatory accountability away from the licensed or regulated business.
Martin’s resistance to fully autonomous “agentic” reporting is therefore more credible than the broad marketplace phrasing. He described agents as tools for gathering public-records information, retrieving documents, categorizing material and assembling an investigation package. That is where automation has a clear operational payoff: it removes repetitive research and clerical work while leaving the conclusion with the person whose organization is exposed if the conclusion is wrong.
That architecture is sensible for gaming AML. A slot floor generates behavior signals that do not resemble a simple bank ledger: wagers, session duration, repeated cash activity, player-account behavior and movements across properties can all be relevant to an investigation. Compute demand rises sharply when an operator scores large populations, searches for patterns and connects activity across venues.
The company says it has detected 43% more suspicious activity, achieved 96% faster decisioning and reduced investigation time from six to eight hours to about 45 minutes. Those are vendor-reported outcomes published in Microsoft’s customer story, not independently audited performance benchmarks. They should be treated as evidence that a deployment delivered improvements for Kinectify’s customer base, not as a guaranteed outcome for every casino, club or bank.
The more durable claim is that cloud elasticity helps with batch scoring and enterprise-wide correlation. On-premises equipment can run AML systems, but it forces a venue to purchase for peaks, refresh hardware, secure the stack and maintain specialized operational skills. Azure shifts much of that infrastructure work to Microsoft’s platform. It does not, however, settle data-governance questions such as retention periods, tenant isolation, cross-border access, incident notification, model-input handling or whether a customer can export its complete investigative record when changing vendors.
For Australian customers, data residency also needs more than an assurance that workloads run in Australian data centers. Boards should establish where backups, logs, support access, disaster-recovery environments and AI-processing components reside; who can access them; and how those commitments are enforced in the agreement. Kinectify says it provides customers with a SOC 2 Type 2 report annually. That is a useful assurance artifact, but it is not a substitute for evaluating the exact control scope, exceptions and customer responsibilities described in the report.
The immediate Australian example is AUSTRAC’s revised transaction-reporting regime. New threshold transaction report and suspicious matter report forms became available on July 1, 2026, and the new SMR bulk-file schema is identified by AUSTRAC as SMR version 3.0. Newly regulated entities under the expanded AML/CTF regime had to be ready to report from that date, and the enrolment deadline of July 29, 2026 has now passed.
However, the story’s implied urgency needs one important qualification. AUSTRAC gives entities that were already enrolled on March 30, 2026 a transition period: they may move to the new forms any time from July 1, 2026 through March 30, 2029. Existing venues do not all face a forced immediate cutover to the new SMR format. Newer entities do, and established operators may still have strong reasons to migrate early, but a vendor’s ability to ship support for a July 2026 form change is not itself proof that every current customer had to deploy it in July.
That is the kind of detail buyers should press vendors on. Ask whether the platform supports both AUSTRAC reporting paths where transitional arrangements apply, how it validates the selected schema, what happens when a report is rejected, and whether report templates can be versioned by property, entity and effective date.
That environment creates a market for faster monitoring and reporting software, particularly among venues without large internal compliance or infrastructure teams. It also creates an incentive to confuse speed with compliance. Faster alert disposition is valuable only if the alerts are better prioritized, the evidence is complete, the human review is meaningful and the organization can explain why it filed—or did not file—each report.
Kinectify’s strongest case is not that AI can replace AML analysts. It is that an Azure-based platform can remove the mechanical work that consumes analysts’ time: pulling records, connecting activity, populating fields, preparing a chronology and drafting a readable narrative. Martin is right to keep the human at the final decision point, because that is the moment when the legal and regulatory risk crystallizes.
For IT teams and compliance leaders, the procurement standard should be straightforward: do not buy an “AI SAR” feature. Buy a controlled investigation system that happens to use AI for drafting, and demand proof that the final report remains evidence-linked, reviewable, editable and attributable to a named human approver.
In an interview with iTWire, Martin said Kinectify’s model is used to turn structured investigative facts into an executive summary or draft report. Microsoft’s own Kinectify customer story confirms that the company uses Azure OpenAI Service for executive investigation summaries and government filings, particularly SARs. Microsoft also says Kinectify’s Azure platform can process as many as 40 million transactions daily.
The consequential detail is what Microsoft’s account does not say: it does not independently describe Kinectify’s human-review controls, the fields passed to the model, or whether the generated narrative is technically prevented from introducing unsupported material. Martin’s statement that the filing remains a draft is therefore the key operational claim—and one customers should verify in product configuration, audit logs and contract language before treating AI-generated narratives as production-ready compliance records.
A narrative generator cannot own the risk decision
Kinectify’s pitch is grounded in a plausible division of labor. A gaming operation can deterministically collect and correlate customer identity data, machine activity, cash-ins, cash-outs, ATM use, loyalty records and prior investigative notes. A large language model can then convert selected facts into readable prose. The human investigator reviews the evidence, decides whether suspicion is reasonable, edits the narrative and approves the submission.That workflow is materially different from asking an LLM to decide whether a patron is laundering money. Martin told iTWire that Kinectify supplies the model with structured facts rather than asking it to discover knowledge from its broader training data. If implemented rigorously, that reduces one obvious risk: a model inventing a transaction, customer attribute or investigative conclusion that was never in the case record.
But reduces is not eliminates. Grounding controls do not automatically solve the problem of wrong source data, missing context, faulty entity matching, stale customer records, or a model that overstates the implication of an otherwise accurate transaction pattern. The most useful audit question is not whether the model is “hallucinating” in the abstract. It is whether every claim in the proposed narrative can be traced back to a specific transaction, record, alert, investigator finding or attached document.
FinCEN’s casino guidance makes the importance of that traceability clear. It asks casinos to maintain supporting records and to ensure SAR narratives clearly explain who, what, when, where, why and how. A polished AI narrative that cannot be traced back to evidence is worse than a clumsy manual draft: it creates a false appearance of confidence in a filing that may later be examined by regulators, auditors or law enforcement.
Kinectify says its investigators can edit the drafts and that actions are logged. For buyers, the practical test is whether the system preserves the original generated text, the source evidence, every edit, the approving user and the exact final version filed. Without that chain, “human in the loop” becomes a slogan rather than a defensible control.
The marketing language is ahead of the control language
There is a notable tension in Kinectify’s public material. Martin told iTWire that AI does not make regulatory decisions and that a human remains accountable for the final filing. Yet Kinectify’s Microsoft Marketplace listing describes the service as having “fully automated SAR capability” and says customers can generate, review and file SAR forms in milliseconds.Those statements can coexist if “fully automated” means automated preparation and workflow rather than autonomous suspicion determinations. But the wording leaves too much room for interpretation in a regulated setting. A casino compliance officer evaluating the platform should insist on a precise answer: Can the software submit a SAR or SMR without an authorized human explicitly approving the matter and the final narrative? If it can, does the vendor recommend that workflow? If it cannot, the marketplace language should be read as sales shorthand rather than a description of an unattended filing process.
The distinction is not theoretical. FinCEN places the obligation on casinos and card clubs to establish and implement risk-based internal controls. Its guidance also stresses documentation of decision-making, including decisions not to file. Kinectify can accelerate the evidence-gathering and writing stages, but it cannot transfer regulatory accountability away from the licensed or regulated business.
Martin’s resistance to fully autonomous “agentic” reporting is therefore more credible than the broad marketplace phrasing. He described agents as tools for gathering public-records information, retrieving documents, categorizing material and assembling an investigation package. That is where automation has a clear operational payoff: it removes repetitive research and clerical work while leaving the conclusion with the person whose organization is exposed if the conclusion is wrong.
Azure solves scale, not the hard compliance questions
Microsoft’s May 2024 customer story supplies the clearest independently published outline of Kinectify’s cloud architecture. The company uses Azure Kubernetes Service for ingestion and workload scaling, Azure Cosmos DB for operational data, Microsoft Fabric for analytics, Azure AI services and machine-learning tooling. Microsoft says Kinectify used managed services so it could build without maintaining an in-house infrastructure team.That architecture is sensible for gaming AML. A slot floor generates behavior signals that do not resemble a simple bank ledger: wagers, session duration, repeated cash activity, player-account behavior and movements across properties can all be relevant to an investigation. Compute demand rises sharply when an operator scores large populations, searches for patterns and connects activity across venues.
The company says it has detected 43% more suspicious activity, achieved 96% faster decisioning and reduced investigation time from six to eight hours to about 45 minutes. Those are vendor-reported outcomes published in Microsoft’s customer story, not independently audited performance benchmarks. They should be treated as evidence that a deployment delivered improvements for Kinectify’s customer base, not as a guaranteed outcome for every casino, club or bank.
The more durable claim is that cloud elasticity helps with batch scoring and enterprise-wide correlation. On-premises equipment can run AML systems, but it forces a venue to purchase for peaks, refresh hardware, secure the stack and maintain specialized operational skills. Azure shifts much of that infrastructure work to Microsoft’s platform. It does not, however, settle data-governance questions such as retention periods, tenant isolation, cross-border access, incident notification, model-input handling or whether a customer can export its complete investigative record when changing vendors.
For Australian customers, data residency also needs more than an assurance that workloads run in Australian data centers. Boards should establish where backups, logs, support access, disaster-recovery environments and AI-processing components reside; who can access them; and how those commitments are enforced in the agreement. Kinectify says it provides customers with a SOC 2 Type 2 report annually. That is a useful assurance artifact, but it is not a substitute for evaluating the exact control scope, exceptions and customer responsibilities described in the report.
AUSTRAC’s deadline is real, but the migration timetable is longer
Martin’s argument that fortnightly releases are safer than multi-year product cycles has a strong operational basis. Small, testable changes are generally easier to review and roll back than a giant platform replacement. Microsoft says Kinectify deploys new features and enhancements every two weeks, with near-real-time bug fixes.The immediate Australian example is AUSTRAC’s revised transaction-reporting regime. New threshold transaction report and suspicious matter report forms became available on July 1, 2026, and the new SMR bulk-file schema is identified by AUSTRAC as SMR version 3.0. Newly regulated entities under the expanded AML/CTF regime had to be ready to report from that date, and the enrolment deadline of July 29, 2026 has now passed.
However, the story’s implied urgency needs one important qualification. AUSTRAC gives entities that were already enrolled on March 30, 2026 a transition period: they may move to the new forms any time from July 1, 2026 through March 30, 2029. Existing venues do not all face a forced immediate cutover to the new SMR format. Newer entities do, and established operators may still have strong reasons to migrate early, but a vendor’s ability to ship support for a July 2026 form change is not itself proof that every current customer had to deploy it in July.
That is the kind of detail buyers should press vendors on. Ask whether the platform supports both AUSTRAC reporting paths where transitional arrangements apply, how it validates the selected schema, what happens when a report is rejected, and whether report templates can be versioned by property, entity and effective date.
The real product is the evidence trail
Kinectify’s expansion into Australian clubs arrives during an enforcement-sensitive period. AUSTRAC’s successful action against Crown resulted in a A$450 million penalty in 2023, after the company admitted shortcomings including inadequate risk assessments, controls and board oversight. The regulator’s broader AML/CTF expansion has now brought tens of thousands of additional businesses into the regime.That environment creates a market for faster monitoring and reporting software, particularly among venues without large internal compliance or infrastructure teams. It also creates an incentive to confuse speed with compliance. Faster alert disposition is valuable only if the alerts are better prioritized, the evidence is complete, the human review is meaningful and the organization can explain why it filed—or did not file—each report.
Kinectify’s strongest case is not that AI can replace AML analysts. It is that an Azure-based platform can remove the mechanical work that consumes analysts’ time: pulling records, connecting activity, populating fields, preparing a chronology and drafting a readable narrative. Martin is right to keep the human at the final decision point, because that is the moment when the legal and regulatory risk crystallizes.
For IT teams and compliance leaders, the procurement standard should be straightforward: do not buy an “AI SAR” feature. Buy a controlled investigation system that happens to use AI for drafting, and demand proof that the final report remains evidence-linked, reviewable, editable and attributable to a named human approver.
References
- Primary source: itwire.com
Published: 2026-08-02T10:44:03.040000+00:00
Kinectify hunts money launderers with AI on Azure, and its CEO Joseph Martin says the humans aren't going anywhere | iTWire
Every time someone presses the button on a poker machine, it gets logged. Multiply that across a venue's floor, a full year of play, and every cash-in, cash-out, ATM trip and...
itwire.com
- Related coverage: microsoft.com
Kinectify fights financial crime in the gaming world with Microsoft Azure | Microsoft Customer Stories
The modern gaming industry must protect itself against illegitimate activity. Kinectify, an anti-money laundering (AML) risk management technology company, uses its modern AI-powered platform to collect and analyze millions of daily transactions to detect suspicious activity. Kinectify has been...www.microsoft.com
- Related coverage: kinectify.com
Kinectify Unveils New Service on Microsoft Cloud for Sovereignty for Tribal Nations
Kinectify and Microsoft collaboration introduces risk management services for tribal gaming allowing sovereign control over data while utilizing cloud services.www.kinectify.com - Related coverage: microsoft.com
- Related coverage: techcommunity.microsoft.com
- Related coverage: kinectify.com
Joseph Martin, CEO Kinectify: "We are building a muscular technology team and delivering capabilities previously unseen in gaming compliance"
CEO of Kinectify Joseph Martin discusses current AML technology and Kinectify technology capabilities with Yogonet.www.kinectify.com - Related coverage: ausleisure.com.au
- Related coverage: marketplace.microsoft.com
Kinectify AML Solution
Kinectify is a fully configurable AML and RG platform for the casino and online gaming industrymarketplace.microsoft.com
- Related coverage: info.microsoft.com
Azure_Hero_Illustration_Wave
ProEXR File Description =Attributes= channels (chlist) compression (compression): Zip16 dataWindow (box2i): [0, 0, 3499, 1968] displayWindow (box2i): [0, 0, 3499, 1968] lineOrder (lineOrder): Increasing Y pixelAspectRatio (float): 1 screenWindowCenter (v2f): [0, 0] screenWindowWidth (float): 1...info.microsoft.com
- Related coverage: learn.microsoft.com
Code of Conduct for Microsoft AI Services | Microsoft Learn
Code of Conduct for Microsoft AI Serviceslearn.microsoft.com - Related coverage: austrac.gov.au
Anti-money laundering and counter terrorism laws cover thousands more businesses | AUSTRAC
Criminals are exploiting Australia’s real estate market and complex company structures to hide and move illicit money - it’s something financial crimes regulator AUSTRAC says it sees regularly. From today, tens of thousands more businesses are now covered by Australia’s anti-money laundering and...www.austrac.gov.au
- Related coverage: yogonet.com
Aristocrat leads fresh funding in Kinectify; Eastern Band of Cherokee Indians, Acies, others participate | Yogonet International
Aristocrat's Jason Walbridge joins Kinectify board Joseph Martin, Kinectify's CEOwww.yogonet.com - Related coverage: owler.com
Kinectify Funding - Owler
Kinectify has historically raised $0 in funding. Learn more about Kinectify’s funding sources, their 2 investment rounds and investors on Owler, the world’s largest community-based business insights platform.
www.owler.com
- Related coverage: yogonet.com
סטארט-אפ טכנולוגי AML Kinectify סוגר סבב גיוס ראשוני בהובלת OpenBet, Eastern Band of Cherokee Indians ו-Acies Investments | Yogonet International
סת' שור, איילר וקרייציק, וואגר בין המשקיעים ג'וזף מרטין, מנכ"ל קינקטיפיי.www.yogonet.com - Related coverage: fincen.gov
FinCEN.gov
Certain money services business (MSBs) are required to report suspicious activity. The following MSBs are subject to the suspicious activity reporting (SAR) requirement:Money transmittersMoney order - issuers, sellers and redeemersTraveler’s check - issuers, sellers and redeemersUS Postal...www.fincen.gov
