# AI for ECC to S/4HANA Migration: Where AI Actually Reduces Delivery Effort > How AI for ECC to S/4HANA migration cuts delivery effort in requirements, ABAP, testing, and governance, plus the parts it does not touch. Canonical: https://www.aifalabs.com/hub/ai-for-ecc-to-s4hana-migration Publisher: AiFA Labs (https://www.aifalabs.com) AI reduces the delivery effort in an ECC to S/4HANA migration by automating the software lifecycle work around it. It generates requirements, assists custom ABAP adaptation, generates unit tests, and drafts documentation, while the surrounding SAP delivery workflow keeps clean core checks and audit evidence connected to the build. It does not run the underlying system or data migration, and it does not replace the human decisions a migration turns on. At AiFA Labs, we use SAP's documented AI tooling alongside SASA (SAP AI SDLC Assist), our platform for automating requirements, specifications, development, testing, and governance across the SAP delivery lifecycle. This page shows where that effort reduction is genuine and where it is not. The deadline is part of the story: SAP mainstream maintenance for the Business Suite 7 applications that include ECC ends on 31 December 2027 (SAP maintenance strategy), so for most programs the binding constraint is delivery capacity, not budget. What follows is a stage by stage map of where AI helps, the tools that do the work, the guardrails that keep an accelerated migration auditable, and the parts of the job AI does not touch. ## Key Takeaways - AI reduces ECC to S/4HANA migration effort where the work sits, in requirements, ABAP development, testing, documentation, and governance, not in the coding keystroke. - SAP mainstream maintenance for Business Suite 7, which includes ECC (SAP ERP 6.0), ends 31 December 2027, with optional extended maintenance to 2030 (SAP maintenance strategy). - SAP's ABAP-trained Joule for Developers generates code, unit tests, and code explanations, and helps migrate custom ABAP to S/4HANA (SAP, Joule for Developers). - AI-generated ABAP must pass the same ABAP Test Cockpit and clean core checks as human-written code, so acceleration does not lower the compliance bar (SAP, clean core enforcement). - Map AI to each delivery stage before you commit a wave plan. Request an S/4HANA SDLC assessment to see where your own effort and risk concentrate. ## Why the ECC to S/4HANA Migration Clock Makes Delivery Effort the Real Constraint The reason to plan an ECC to S/4HANA migration around delivery effort, rather than licenses, is that SAP has fixed the calendar and committed the target platform for the long term. Mainstream maintenance for the Business Suite 7 core applications, which include SAP ERP 6.0, the release most teams call ECC, runs until the end of 2027. After that, SAP offers optional extended maintenance to the end of 2030 at, in its own words, “a premium of two percent points on the maintenance basis for all support offerings for the scope of SAP Business Suite 7” (SAP maintenance strategy, support.sap.com). Teams that buy nothing extra move to customer specific maintenance instead. The destination, meanwhile, is committed: SAP states that “until 2040, there will always be at least one release of SAP S/4HANA in maintenance”. So the pressure sits on the source system and the calendar, which makes the speed and quality of delivery the thing worth optimizing. The ECC to S/4HANA maintenance clock. Dates from SAP’s maintenance strategy. ### What Are the SAP Maintenance Deadlines for ECC and S/4HANA? Here are the dates SAP publishes and what each one means for planning a program. | Date | What SAP commits | What it means for your program | | --- | --- | --- | | 31 December 2027 | Mainstream maintenance ends for the Business Suite 7 core applications, which include SAP ERP 6.0 (ECC) | The real planning horizon | | 2028 to 2030 | Optional extended maintenance at a 2% premium, for the scope of SAP Business Suite 7 | A paid extension, not a default | | After extended maintenance | Customer specific maintenance for those who do not opt in or who run past 2030 | A fallback, not a plan | | Through 2040 | SAP S/4HANA innovation commitment, with at least one release always in maintenance | The target is committed long term | Dates and quotations from the SAP maintenance strategy for SAP Business Suite 7 and SAP S/4HANA. ## Greenfield, Brownfield, or Selective: Where the Effort Lands in Each S/4HANA Approach SAP defines three ways to reach S/4HANA, and the one you choose decides where your delivery effort, and therefore where AI helps most, will concentrate. Read the three approaches through that lens, because the same AI capability pays back at a different point in each. The three SAP transition approaches, and where delivery effort concentrates in each. ### New Implementation (Greenfield) A new implementation, which SAP also calls the greenfield approach, is a “new implementation of SAP S/4HANA, e.g. for customers migrating a legacy system”, with the benefit of “reengineering and process simplification based on pre-configured business processes” (SAP transition paths). You rebuild on standard and carry forward only what earns its place, so the effort concentrates early, in requirements and specification, where the work is deciding how far to adopt the standard and what to keep as a deliberate exception. That is where AI-assisted requirements and design work pays back first. ### System Conversion (Brownfield) A system conversion, the brownfield approach, is “a complete technical in-place conversion of an existing SAP Business Suite ERP system to SAP S/4HANA” that lets you “keep your investment in custom code” (SAP transition paths). Because you bring the existing estate with you, the effort concentrates in custom-code adaptation, testing, and documentation. This is the approach where AI has the most to remove, since the work is large, repetitive in shape, and well defined by SAP's own readiness checks. ### Selective Data Transition A selective data transition is, in SAP's description, “an alternative to a System Conversion and New Implementation” that “combines the advantages of both approaches without their limitations, using innovations in S/4HANA while selectively leveraging and re-using your existing investments” (SAP selective data transition). The effort profile is mixed, so the AI payoff is mixed too: some greenfield-style design work, and some brownfield-style code and test adaptation. ## Where AI Reduces Delivery Effort Across the S/4HANA Program An ECC to S/4HANA program runs through SAP's own delivery method, SAP Activate, whose six phases are Discover, Prepare, Explore, Realize, Deploy, and Run (SAP Activate methodology). The effort that overruns is rarely the typing of code. It is the requirements, the specifications, the custom-code adaptation, the test build, the documentation, and the evidence that ties them together. AI removes measurable effort in each of those, and much of it comes from SAP's own tooling. Here is where, stage by stage. Where AI reduces delivery effort across the SAP delivery lifecycle, by stage. ### Requirements, from Workshop Transcripts and System Findings AI can turn the raw material of a requirements phase into structured requirements automatically. Inside SAP Cloud ALM, a generally available feature generates business requirements directly from fit-to-standard workshop transcripts, which SAP describes as analyzing the discussion, populating a predefined template, and “shifting the consultant's focus from manual transcription to strategic review and refinement” (SAP Business AI release highlights, Q4 2025). SAP puts the saving at up to 50% on requirement creation and up to 20% on the follow-on user stories, figures it marks as estimated benefits that will vary. The same tool also creates requirements straight from a finding in SAP Readiness Check (SAP Cloud ALM requirement management), so the output of the system analysis becomes a tracked requirement with no one re-keying it. Readiness Check itself is deterministic analysis, not AI. It is the input that the AI-assisted step builds on. ### Specifications, Functional and Technical Design For the specification work that follows, SAP's developer AI drafts the artifacts a specification depends on rather than the prose around them. Joule for Developers can, in SAP's words, “generate code, UI, data models, and sample data across SAP programming models for Java, JavaScript, and ABAP” and automate repetitive tasks like documentation and sample data generation (Introducing Joule for Developers). In practice, the data models, sample data, and first-cut objects a design would otherwise describe from scratch are generated and explained, leaving the design decisions to people. SAP does not publish a feature that writes a full technical specification document, so we describe this as accelerating the build artifacts, not replacing the design. ### Development and Custom Code, the Biggest Effort Sink This is where an AI trained on the actual language does the most. SAP's ABAP model powers Joule for Developers, which SAP describes as “trained on millions of lines of ABAP code” (Joule for developers) and whose top-used features it names as unit test generation, code generation, and autocompletion (SAP, July 2025). For a migration specifically, SAP positions it to “migrate custom ABAP code faster with documentation chat, code explanation, and issue diagnostics”. Before any of that, the Custom Code Migration app “evaluates development objects to be adopted to SAP S/4HANA for compatibility and identifies unused custom code” and runs the S/4HANA checks to analyze which code needs adapting (SAP, analyzing customizations), and it is integrated with ABAP Test Cockpit behind the scenes with a Quick Fix that corrects code to conform with S/4HANA changes (SAP, adapting custom code). SAP is now extending this into agentic AI. Its S/4HANA Custom Code Migration Agent automates ATC analysis, applies deterministic quick fixes, and uses AI-powered remediation for more complex ECC custom-code adaptations, while keeping developers responsible for review and activation (SAP, S/4HANA Custom Code Migration Agent). Where the target architecture supports it, new and modernized code moves toward ABAP Cloud, SAP's recommended development model for “clean core compliant business apps, services or extensions” (ABAP Cloud FAQ). So the estate is triaged, the throwaway is found, the adaptations are drafted and explained, and the surviving code is modernized toward the clean core model. On cost, SAP shows an illustrative 30% figure that it marks as assumed for a hypothetical company, so we treat it as an SAP assumption rather than a result. ### Testing, Unit, Functional, and Regression Testing is the second large workstream, and SAP Cloud ALM is built to carry it. Its test management lets teams “prepare and execute manual and automated tests” and links every test to processes, requirements, and user stories “to support an E2E implementation process with full traceability” (SAP Cloud ALM test management). Automation runs through a public Test Automation API, so SAP and partner tools plug in and, in SAP's words, “reduce your manual testing efforts” while keeping orchestration and reporting in one place (SAP, integrating test automation providers). AI generates the tests themselves, since Joule can create unit tests and SAP Build Code can “produce rapid unit tests with AI for existing code” (SAP developer tools). One boundary matters here: SAP ships ready test content for standard processes, drawn from SAP Signavio Process Navigator, but custom or customized processes are not covered and need their own cases (SAP, explaining the test strategy). Those custom cases are separate work, and generating them from the specification is part of what SASA does, which we come to below. ### Documentation, Specs, Process Docs, and Summaries Documentation is usually the first thing to fall behind, and AI keeps it moving. In SAP Cloud ALM, a generally available feature lets users trigger an AI-generated summary of a long document, edit it, and keep it as a persistent section (SAP Business AI release highlights, Q2 2026). Joule generates documentation and sample data as part of the developer flow (SAP, Joule for developers), and SAP Signavio Process Manager produces process documentation from the model itself, including the ability to “customize process documentation with templates” (SAP Signavio Process Manager). The point for a migration is that the documentation is generated from the same artifacts being built and tested, so it stays closer to the truth than a separately written document would. ### Manual Handoffs, Removing the Re-keying Between Stages Much of the delay in a program is not inside any one stage but in the handoff between them, where a requirement is retyped as a user story or a specification is re-authored as a test. SAP Cloud ALM is designed to remove that by keeping one connected record. It ensures “traceability from processes, requirements, tasks, tests, down to the deployment to production” and lets teams consume methodology tasks with accelerators from SAP Activate roadmaps (SAP Cloud ALM implementation). Because every item stays linked, work moves from stage to stage as connected records rather than as fresh documents someone has to write again. ### Governance, Compliance and Evidence Kept Intact For governance, the effort that usually balloons is assembling the compliance evidence at the very end, under audit pressure. Reducing it is a matter of making the checks and the record part of the build rather than a phase after it. That is important enough, and specific enough, that we cover it on its own below. ## What AI Tools Support S/4HANA Transformation? The tools fall into two groups: AI that generates or assists, and deterministic analysis that feeds it. Naming them honestly matters, because a Readiness Check is not an AI, and treating it as one sets the wrong expectation. The table below sorts the main SAP-documented tools by stage and by which kind they are. Each tool name links to SAP's own documentation. | Tool | What it does | Delivery stage | AI or analysis | | --- | --- | --- | --- | | Joule for Developers | Generates and explains ABAP, generates unit tests, assists custom-code migration | Development, testing, docs | AI | | SAP Build Code | Generative AI code environment for Java and JavaScript, clean-core S/4HANA extensions | Development, testing | AI | | SAP Cloud ALM AI features | Generates requirements from workshop transcripts, summarizes documents | Requirements, documentation | AI | | SAP Signavio AI features | Process mining with embedded AI to see where processes actually run | Discovery, requirements | AI | | SAP Readiness Check | Analyzes simplification items, custom code, and add-on compatibility for a conversion | Discovery | Analysis | | Custom Code Migration app | Identifies unused code and what needs adapting, with a Quick Fix | Development | Analysis and fix | | S/4HANA Custom Code Migration Agent | Agentic AI that explains custom code, makes changes, and orchestrates the ATC checks and quick fixes to adapt ECC code for S/4HANA, with human review | Development | Agentic AI + analysis | | ABAP Test Cockpit | Static checks for quality, security, and clean-core adherence | Development, governance | Analysis | The distinction to carry into a plan is simple: the analysis tools tell you the size and shape of the work, and the AI tools reduce the effort of doing it. A program that runs the analysis first and then points AI at the result gets both. ## How AI Keeps an Accelerated Migration Clean-Core Compliant and Auditable The faster we generate code and tests, the more two questions matter. Is it still clean core, and can we prove what happened? SAP's tooling answers both, and there is a separate, public frame for governing the AI itself. The four governance layers that keep an accelerated migration auditable. ### Keeping Custom Code Clean-Core Compliant as AI Generates It AI-generated ABAP must clear the same checks as anything a person writes, because the checks run on the code, not the author. SAP describes clean-core compliance as “enforced by syntax checks and ABAP test cockpit checks in ABAP development tools for Eclipse as part of the ABAP Cloud development model” (SAP, exploring released APIs). ABAP Test Cockpit offers “a large variety of checks ranging from performance checks and security checks, to checks for adherence, to guidelines and programming conventions” (SAP, ABAP Test Cockpit). SAP also defines four clean core compliance levels, A through D, from fully compliant extensions that use only released, stable interfaces down to those it does not consider clean (SAP, extending S/4HANA Cloud the right way), and SAP Cloud ALM's System view dashboard “gives you an overview of your system's clean core compliance” (SAP Cloud ALM clean core adoption). Generating code faster does not move that gate. It just means more code arriving at it. ### The Evidence Trail, Traceability and Audit Logs An accelerated program still has to show its work, and the record is built in. The same traceability that removes handoffs also produces the audit trail, linking requirements, tasks, and tests down to deployment (SAP Cloud ALM implementation). For activity-level evidence, the SAP Audit Log Viewer service records end-user activity in SAP Cloud ALM, which SAP calls “security-relevant chronological records that provide documentary evidence for an event or activity” (SAP, Audit Log Viewer service). When someone asks how a change was justified, tested, and approved, the answer is a query rather than an excavation. ### Governing the AI Itself, EU AI Act, NIST AI RMF, and GDPR Accountability There is also the matter of governing the AI itself, where the useful references are public. The EU Artificial Intelligence Act, Regulation (EU) 2024/1689, points the right way for delivery: it calls for “the automatic recording of events (logs) over the lifetime of the system” (Article 12), transparency that lets a person interpret the output (Article 13), and effective human oversight (Article 14). The NIST AI Risk Management Framework frames the same discipline as Govern, Map, Measure, and Manage, and the GDPR’s accountability principle, Article 5(2), requires a controller to “be able to demonstrate compliance” (Regulation (EU) 2016/679), which is the legal version of the evidence trail. None of this makes SAP’s tooling or ours a regulated high-risk system; that turns on specific product and use-case rules, not on the fact that AI wrote code. We treat it as the backdrop for using AI-assisted delivery responsibly. ## What AI Does Not Remove From an S/4HANA Migration AI takes SDLC effort out of a migration. It does not take out the migration. Being clear about the edges is what makes the rest of this credible, so here is what stays in human hands. - Data migration is separate work. Everything above is about building and testing software. Moving and reconciling data is a parallel discipline with its own tools and its own risk, and none of the AI here removes it. - Fit-to-standard is a human decision. Generating a requirement from a workshop still leaves the real choice to people, which is whether to adopt the standard process or keep a deliberate exception. SAP's own feature hands that judgment back to a reviewer by design. - Custom processes still need custom tests. SAP's ready test content covers standard processes. Anything you have customized needs its own cases, generated or written. - Oversight stays. The governance frame above exists because a person has to remain accountable for what the AI produced. Faster output raises the value of review, it does not remove it. - The numbers are stated estimates. The percentages SAP publishes are estimates, or in one case an assumed model for a hypothetical company. We quote them as SAP's figures and make no claim that AI cuts migration time by a fixed percentage, because no official source states one. ## How to Put AI to Work Across Your S/4HANA Delivery Lifecycle If you are planning a program now, the practical move is to apply AI in the order the work actually happens, so each stage feeds the next. This is the sequence we run. - Baseline the estate. Run SAP Readiness Check and the Custom Code Migration app to see the simplification items, which custom code is used, and what needs adapting. - Generate requirements from the room. Capture the fit-to-standard workshops and turn the transcripts, and the Readiness Check findings, into tracked requirements in SAP Cloud ALM. - Draft and adapt the code under the checks. Use the ABAP-trained developer AI to generate, explain, and adapt custom code for S/4HANA, moving appropriate new and modernized development toward ABAP Cloud while ABAP Test Cockpit enforces clean core requirements. - Generate the tests and run them with traceability. Generate unit tests with the developer AI, build the functional cases the custom processes need, automate through the Test Automation API, and link every test back to its requirement. - Let the documentation come from the work. Summarize and generate documentation from the same artifacts, so it stays current instead of being rebuilt at the end. - Keep the evidence. Rely on traceability and the audit log so the compliance record is a by-product, not a scramble before go-live. This is the lifecycle we work in every day, and it is the reason we built our platform around the whole of it rather than any single stage. ## How We Accelerate SAP Delivery With SASA (SAP AI SDLC Assist) We build SASA (SAP AI SDLC Assist) because an ECC to S/4HANA migration is, underneath the labels, an SAP delivery lifecycle problem on a deadline. The work that gates it is the specification, the ABAP, the tests, the documentation, and the evidence, which is the same work we automate across the lifecycle. SASA is SAP-first and clean core-driven, and it is built to cover the whole lifecycle rather than a single stage. On our projects it converts business requirements into WRICEF-aligned functional design specifications, turns those into technical designs, generates ABAP aligned to SAP clean core principles, and auto-generates unit and functional test cases from the specifications, all inside the existing SAP environment. Where a code assistant helps write a method and a test tool runs a case, we orchestrate the stages so work does not fall into the gaps between tools. On one SAP ABAP development program in oil and gas, we saw application delivery run about twice as fast, with the functional and technical design specification work fully automated. We report that as our own client result, and the comparison baseline and the exact lifecycle scope measured are for AiFA to confirm before publication. SASA does not replace the data migration or SAP's readiness analysis, and it is not a substitute for the decisions your people make. It removes the lifecycle effort the deadline does not leave time for. ## Request an S/4HANA SDLC Assessment If you are weighing an ECC to S/4HANA move, the first useful step is to see where your own effort and risk actually sit. We run a short S/4HANA SDLC assessment that maps your requirements, custom code, testing, documentation, and governance against where AI can take effort out, so your wave plan is built on your estate rather than a generic timeline. Request an S/4HANA SDLC assessment and we will show you where the acceleration is real for you. ## About the Author Sagar Chakraborty is Director of Artificial Intelligence Innovations and Strategy at AiFA Labs, where he leads the team building SASA (SAP AI SDLC Assist), the company's AI platform for the SAP delivery lifecycle, and works most closely with Fortune 500 companies in life sciences and healthcare. He holds a PhD and a Master of Technology in Artificial Intelligence from IIT Jodhpur, completed a management program at IIM Calcutta, holds two patents in document-processing neural architectures, and has spent over ten years building AI products, starting as a software engineer in 2014 and previously shipping at Amazon Robotics, Wipro and Allied Media. TradeFlock named him one of India's Top 10 AI Leaders in 2025, and his research is connected to AI labs at IIT Jodhpur and IIT Kharagpur. His SAP expertise is in the delivery lifecycle, which is the lens this article applies to how system integrators accelerate delivery with AI; the architectural, tooling and governance claims here rest on the linked primary sources rather than on personal opinion. Read Sagar Chakraborty's full profile. Connect with Sagar on LinkedIn. ## Frequently Asked Questions ### Can AI Accelerate an ECC to S/4HANA Migration? Yes, but mainly by reducing the software lifecycle effort around it. AI generates requirements, assists custom ABAP adaptation, generates unit tests, and drafts documentation, while the surrounding SAP workflow keeps clean core checks and audit evidence connected to the build. It does not perform the underlying system or data migration, which stays specialist work. ### Which Parts of an ECC to S/4HANA Migration Can AI Automate? The build side of the lifecycle: requirements from workshop transcripts and Readiness Check findings, ABAP generation and custom-code adaptation, unit-test generation, documentation and summaries, and the handoffs between stages. Process discovery and clean-core checks are automated analysis rather than generative AI, and every output is reviewed by a person. ### Can AI Reduce the Time Required for an S/4HANA Migration? It reduces effort in specific stages, but there is no official SAP figure for cutting total migration time, and we do not claim one. SAP publishes stage-level estimates, for example up to 50% less effort on requirement creation, and marks them as estimates. Plan around the effort you can remove, not a single headline percentage. ### How Long Does an ECC to S/4HANA Migration Take? SAP does not publish a single duration, and any figure that promises one is guessing. It depends on the approach and the size of your custom estate, so the useful number is not a duration at all. It is the 31 December 2027 end of mainstream maintenance (SAP maintenance strategy), with a paid extension to the end of 2030. Plan backward from that date, and size the work against your own estate rather than a benchmark from someone else’s migration. ### What Skills Does an AI-Accelerated S/4HANA Team Need? The mix shifts from producing artifacts to reviewing them. When AI drafts requirements, generates ABAP, and writes test cases, the scarce work becomes senior review and judgment: experienced ABAP engineers to sign off on generated code, process owners to make the fit-to-standard calls, and test leads to confirm the generated cases actually prove the process. Data migration specialists still sit outside all of this. In headcount terms, plan for fewer hours of hands-on authoring and more hours of senior review, because that is where the risk now concentrates. ### Can We Use These AI and Analysis Tools Before the Conversion Begins? Yes, and that is the ideal time. SAP Readiness Check, the Custom Code Migration app, and SAP Signavio process mining run against your current ECC system before any conversion starts, so you can size the custom-code work and see where your processes actually run up front. Starting there turns a vague migration into a scoped one, which is what makes the later AI-assisted build predictable rather than hopeful. ## References - SAP. Maintenance strategy for SAP Business Suite 7 and SAP S/4HANA. support.sap.com. - SAP. Discover the transition paths to SAP S/4HANA. SAP S/4HANA Regional Implementation Group. - SAP. Selective Data Transition engagement. support.sap.com. - SAP. SAP Activate methodology. sap.com. - SAP. SAP Business AI release highlights, Q4 2025 (AI-assisted requirement generation). news.sap.com, January 2026. - SAP. Requirement Management, SAP Cloud ALM implementation expert portal. support.sap.com. - SAP. SAP Readiness Check. pages.community.sap.com. - SAP. Introducing Joule for Developers: AI-powered capabilities across SAP. news.sap.com, March 2025. - SAP. Joule for developers. sap.com. - SAP. How Joule for Developers and ABAP AI capabilities transform the developer experience. news.sap.com, July 2025. - SAP. SAP developer tools (SAP Build Code). sap.com. - SAP. ABAP Cloud FAQ. pages.community.sap.com. - SAP. Analyzing customizations after system conversion (Custom Code Migration app). learning.sap.com. - SAP. Adapting custom code. learning.sap.com. - SAP. SAP S/4HANA Custom Code Migration Agent (Custom Code Migration Guide for SAP S/4HANA). SAP Help Portal, 2026. - SAP. Exploring released APIs (clean core enforcement). learning.sap.com. - SAP. Improving code quality using ABAP Test Cockpit. learning.sap.com. - SAP. How to extend SAP S/4HANA Cloud the right way (clean core compliance levels). news.sap.com, August 2025. - SAP. Why SAP Cloud ALM supports the adoption of clean core. support.sap.com. - SAP. Test Management, SAP Cloud ALM. support.sap.com. - SAP. Integrating test automation providers, SAP Cloud ALM. help.sap.com. - SAP. Explaining the test strategy. learning.sap.com. - SAP. SAP Business AI release highlights, Q2 2026 (AI-assisted document summary). news.sap.com, July 2026. - SAP. SAP Signavio Process Manager. sap.com. - SAP. SAP Signavio Process Intelligence. sap.com. - SAP. SAP Cloud ALM implementation (traceability). support.sap.com. - SAP. Activating the SAP Audit Log Viewer service. help.sap.com. - European Union. Regulation (EU) 2024/1689 (Artificial Intelligence Act). EUR-Lex, 2024. - European Commission. AI Act, Article 12 (record-keeping). ai-act-service-desk.ec.europa.eu. - European Commission. AI Act, Article 13 (transparency). ai-act-service-desk.ec.europa.eu. - European Commission. AI Act, Article 14 (human oversight). ai-act-service-desk.ec.europa.eu. - NIST. Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1. nvlpubs.nist.gov, 2023. - European Union. Regulation (EU) 2016/679 (General Data Protection Regulation), Article 5(2). EUR-Lex.