Department of State Generative AI Playbook A plan to develop and scale secure, responsible, useful, and adaptable enterprise generative artificial intelligence solutions for government missions. Published by the U.S. Department of State in July 2026, the Generative AI Playbook is a practical guide for U.S. government agencies developing and scaling enterprise AI solutions, with a focus on generative AI technology. It documents the Department's experience developing, testing, deploying, expanding, and assessing StateChat, its first Sensitive But Unclassified-enabled enterprise generative AI chatbot. ^^ The playbook provides a repeatable approach, real-world examples, actionable steps to accelerate AI delivery, and sample deliverables. It balances strategic direction with operational detail to help agencies progress from experimentation to implementation. ^^ The guidance is not one-size-fits-all. Agencies are encouraged to adapt the suggested processes and methodologies to their missions, needs, authorities, technologies, risks, and organizational circumstances and to consult their own legal, security, privacy, records-management, accessibility, acquisition, and compliance offices before implementation. ^^ References to commercial products or vendors do not constitute endorsement by the Department of State. The playbook describes StateChat as a case study rather than prescribing one mandatory technical architecture or delivery method. ^^ Submitter's Note: This StratML rendition has been compiled from the source by ChatGPT and reviewed in the form at https://stratml.us/forms/Claude/Part1.html U.S. Department of State DOS 3e3b6667-82ba-42e1-85c3-a1feac5d166b Advances the foreign policy and national-security interests of the United States and developed the enterprise generative AI experience documented in this plan. Dr. Kelly Fletcher Serves as the Department's Chief Information Officer and co-sponsored the playbook as guidance for transforming operations, improving decision-making, empowering the workforce, and strengthening mission impact through generative AI. Amy Ritualo Serves as Acting Chief Data and Artificial Intelligence Officer and co-sponsored the playbook to help agencies anticipate challenges, seize opportunities, and deliver meaningful AI-enabled results. Bureau of Diplomatic Technology Provides the enterprise technology capabilities and organizational support underlying StateChat and the Department's broader generative AI platform. Center for Analytics Led the design, development, testing, deployment, adoption, and continuing evolution of StateChat and the broader enterprise generative AI platform. DT Cyber Operations Supports cybersecurity requirements, threat testing, safeguards, authorization, and secure operation of enterprise AI capabilities. DT Enterprise Infrastructure Supports the infrastructure, networking, availability, and global operation required for enterprise AI services. DT Enterprise Chief Information Security Office Supports information-security governance, risk management, authorization, and secure administration of the enterprise platform. Bureau of Diplomatic Security Blue Team Supports adversarial, cybersecurity, and risk testing of StateChat and its underlying systems. Office of Accessibility and Accommodations, Accessibility Division Provides accessibility expertise and supports conformance with Section 508 and related requirements. Independent Testing, Evaluation, Verification, and Validation Staff Independently evaluate AI functionality, risks, safeguards, reliability, accessibility, and performance against defined requirements. Records Office Supports records-management requirements and practices for AI-generated and AI-assisted content. Privacy Office Supports privacy analysis, impact assessment, safeguards, and responsible handling of information within enterprise AI systems. General Services Administration Technology Modernization Fund Provided investment supporting StateChat, secure AI infrastructure, application programming interfaces, controlled development environments, staffing, and enterprise deployment. Department of State Executive Leaders Set direction, authorize action, sponsor AI initiatives, communicate approval, align priorities, and help overcome organizational barriers. Department of State Bureaus, Missions, Embassies, and Consulates Adapt, test, use, and provide feedback on generative AI capabilities according to their missions, operational needs, and governing requirements. StateChat Delivery Team Combines product ownership, application development, platform administration, governance, testing, customer success, communications, training, change management, innovation, and research expertise. StateChat Users Apply StateChat in mission and administrative workflows, identify valuable use cases, test capabilities, report problems, and guide iterative improvement. AI Testers and Early Adopters Test security, accessibility, functionality, performance, usability, safeguards, and practical value during successive development phases. U.S. Government Agencies Adapt the playbook's methods, tools, deliverables, examples, and lessons to develop and scale enterprise AI solutions for their own missions. Legal, Security, Privacy, Records, Accessibility, Acquisition, and Compliance Offices Review proposed AI implementations and establish the requirements, controls, authorizations, contractual provisions, and policies appropriate to each agency. External Technology and Research Partners Provide specialized expertise, pretrained models, technical capabilities, implementation support, research, lessons learned, and development resources. American Public Benefits from improved government operations, decision-making, productivity, mission performance, and responsible delivery of public services. A future in which secure, responsible, trustworthy, and effective artificial intelligence strengthens American prosperity and government mission performance. d882f366-e15c-40a0-a64a-a89f5b16722e To develop, deploy, scale, and continuously improve enterprise generative AI capabilities that empower personnel, streamline workflows, improve decision-making, and deliver meaningful mission results. 1238f0c0-829d-4d6c-9a5d-3f22db485342 Mission Impact Focus AI development and use on meaningful improvements in mission performance, decision-making, productivity, and public service. Responsibility Develop and use AI in ways that protect civil rights, civil liberties, privacy, institutional responsibilities, and public trust. Trustworthiness Ensure AI capabilities are safe, secure, resilient, explainable, privacy-enhanced, contextually accurate, accountable, transparent, valid, and reliable. Security Protect sensitive information, systems, users, and operations through rigorous cybersecurity, governance, testing, authorization, and access controls. User-Centeredness Engage users throughout development, respond to their needs, and prioritize capabilities that produce practical value in real workflows. Accessibility Design and test AI capabilities so they are usable by people with disabilities and conform to applicable accessibility requirements. Collaboration Combine expertise across organizational, professional, governmental, academic, and private-sector boundaries. Experimentation Start with focused pilots, test assumptions early, learn from successes and failures, and improve through rapid iteration. Adaptability Tailor processes, technologies, policies, and applications to changing missions, risks, user needs, and technical capabilities. Transparency Communicate priorities, capabilities, limitations, safeguards, accomplishments, decisions, and lessons learned clearly. Groundwork Establish the strategic, organizational, technical, human, governance, and financial foundations required for secure and scalable enterprise AI development and adoption. G1 1 Center for Analytics Establishes and applies iterative, multidisciplinary approaches for delivering data, analytics, and AI capabilities. General Services Administration Technology Modernization Fund Provides external resource support for strategic hiring, infrastructure investment, and movement from pilots to enterprise deployment. Executive Leaders Set strategic direction, align priorities, sponsor initiatives, support governance, and secure organizational commitment and resources. Multidisciplinary AI Specialists Contribute engineering, data science, security, governance, accessibility, user-experience, communications, training, policy, research, and change-management expertise. The playbook warns that AI initiatives can derail before development begins because of misaligned goals, skeptical stakeholders, missing expertise, insufficient organizational support, and resource shortfalls. The Department established an Enterprise Data Strategy in 2021, an AI Strategy in 2023, and a combined Data and AI Strategy in 2025. These strategies aligned personnel around shared priorities, increased transparency about accomplishments and direction, articulated a public commitment to responsible and trustworthy AI, and left bureaus and missions room to tailor implementation to their differing needs. The Department had also established the Center for Analytics in 2019 to embed data and analytics into decision-making, capture efficiencies, and advance foreign policy. Led by the Chief Data and AI Officer and housed in the Bureau of Diplomatic Technology, the Center developed an iterative and agile implementation method called the campaign model. Matrixed campaign teams integrated users, technical development, and policymaking; learned through hands-on experimentation; identified process gaps and operational needs early; and adapted governance while projects were underway. Before StateChat, the Center used this model in twelve six-month campaigns. The Department also designed its enterprise generative AI platform from the outset as a managed shared service rather than merely as the system hosting one chatbot. Security, contracts, access processes, and technical configurations were structured broadly enough to support production data, new applications, secure experiments, application programming interfaces, and more than 120 builder projects. The playbook advises agencies to secure funding early, develop hiring pipelines through fellowships and partnerships, make specialist positions attractive, and invest in upskilling existing personnel so institutional knowledge and emerging technical expertise develop together. Data and AI Strategies Establish clear enterprise data and AI strategies that align priorities and principles, articulate responsible and trustworthy intentions, increase transparency, and enable mission-specific implementation. O1_1 1.1 A shared strategic direction enables coordinated and safe adoption across a large organization without requiring uniform implementation in every component. The Department's successive strategies established continuity while accommodating the rapid development of AI technology and the differing operational needs of bureaus, missions, and posts. Organizational Structures Establish dedicated innovation capacity and a cross-functional implementation method that integrate users, technology development, governance, and policymaking. O1_2 1.2 The Center for Analytics and its campaign model enabled technology and policy to develop in parallel. Multidisciplinary teams gathered requirements from leaders and working-level personnel, examined current processes, defined solutions, set strategic goals, identified business needs and governance gaps early, and adapted through experimentation rather than waiting for every issue to be resolved in advance. Focused Pilots Conduct small-scale, time-bounded pilots to demonstrate feasibility, identify valuable use cases, build expertise, generate momentum, and establish repeatable delivery processes. O1_3 1.3 The Department used focused campaigns to streamline processes, break down data silos, test emerging technologies, and deliver visible results. Teams tested continuously with small groups of targeted users, collected real-time feedback, refined their solutions, and documented successes and failures for reuse. One early generative AI pilot, the AI for Reports Modernization campaign, digitized, translated, and extracted key data from more than 115,000 documents; improved financial-data accuracy by 28 percent; enabled tracking of more than $800 million in lease liability; saved 76,500 hours; and avoided $6 million in costs while supporting compliance with new accounting requirements. The initiative later received a Best-in-Class award. Expertise Assemble multidisciplinary teams with the technical, operational, governance, testing, customer-success, research, and policy expertise required for enterprise AI delivery. O1_4 1.4 The StateChat delivery team combined product owners, application developers, platform engineers, data scientists, governance experts, security analysts, testing specialists, communications and training personnel, change-management practitioners, and innovation and research staff. It also relied on support from cybersecurity, infrastructure, information-security, records, privacy, accessibility, and other offices. The playbook emphasizes that AI implementation requires emerging specialties as well as the institutional and mission knowledge of existing personnel. Workforce Capacity Develop hiring, fellowship, partnership, and staff-development channels that expand AI expertise while preserving and applying institutional knowledge. O1_5 1.5 The Department benefited from federal fellowship programs, universities, AI laboratories, vendors, consultancies, and internal specialists. Agencies are advised to secure hiring authorities and funding early, establish pipelines for scarce skills, design compelling roles, and provide upskilling opportunities so current personnel can participate in emerging AI work rather than treating outside expertise as a substitute for the workforce. Resource Support Secure sufficient funding, staffing, infrastructure, acquisition support, and external assistance to advance successful AI initiatives from experimentation to enterprise deployment. O1_6 1.6 The General Services Administration Technology Modernization Fund supported StateChat and the broader platform by funding infrastructure, strategic hiring, secure application programming interfaces, controlled development environments, and enterprise deployment. The playbook identifies external funding mechanisms as particularly valuable when internal resources are insufficient to move a promising pilot into sustained operational use. Enterprise AI Platform Establish a flexible, managed enterprise AI platform that supports sensitive information, analytics, application development, data engineering, secure experimentation, reuse, and evolving mission needs. O1_7 1.7 The platform was designed as a shared enterprise service rather than as a single-purpose StateChat system. Security controls and contracts were structured broadly enough that applications and sandbox experiments could work with production data from the outset when authorized. Business processes provided enterprise visibility and enabled builders to obtain needed tools and data while following defined rules of behavior. The configuration developed for StateChat subsequently supported major enterprise applications and more than 120 builder projects, including crisis-management tools addressing urgent and life-threatening situations. Enterprise Governance Establish policies, review processes, safeguards, visibility, and rules of behavior that enable responsible access to AI tools and data. O1_8 1.8 The Department applied established security, privacy, records, accessibility, infrastructure, acquisition, and governance disciplines to novel generative AI issues. These included authorizing a FISMA High system for sensitive AI workloads, conducting a Privacy Impact Assessment for a generative AI platform, establishing records-management practices for AI-generated content, and determining how builders could experiment securely. Although no preexisting end-to-end generative AI implementation method was available, the Department found that established professional disciplines, when applied collectively, could address the emerging requirements. Organizational Learning Document and share successes, failures, reusable practices, and operational lessons to reduce repeated mistakes and accelerate subsequent AI delivery. O1_9 1.9 Campaign teams recorded what worked and what failed and shared their findings broadly. Starting small and moving quickly helped establish safe experimentation practices, build a culture of innovation, inform responsible-use policies, generate confidence, and create momentum for enterprise-wide expansion. The playbook treats documented learning as a reusable organizational asset rather than as incidental project history. Feasibility Demonstrate that a secure, responsible, accessible, reliable, and globally usable enterprise generative AI solution can be developed and improved. G2 2 Product Owners Set product direction, translate organizational needs into prioritized capabilities, define performance requirements, and secure executive sponsorship. Application Developers Build, integrate, test, and iteratively improve the minimum viable product and its supporting application capabilities. Platform and Governance Specialists Authorize, administer, secure, and govern the enterprise platform and its access to systems and information. AI Testing Specialists Test new functionality and model performance against requirements for safety, security, reliability, accessibility, usability, and contextual accuracy. Customer Success Specialists Surface user needs and establish communications, training, support, feedback, and change-management capabilities from the beginning of product development. Innovation and Research Specialists Explore emerging technologies, prototype capabilities, investigate risks and opportunities, and inform product development and AI strategy. Cybersecurity Teams Define system requirements, test cyber defenses, validate controls, identify vulnerabilities, and support authorization to operate. Accessibility Specialists Evaluate the prototype and its releases for accessibility and support compliance with Section 508 requirements. Foreign Policy Advisors Help ensure that technical design, testing, safeguards, and potential use cases reflect diplomatic missions and operational contexts. Alpha Testers Test early capabilities, security, access, identity, authentication, usability, performance, and practical value before broader deployment. Private-Sector Vendors Provide technical capabilities, implementation expertise, products, services, and access to pretrained models needed to accelerate development. Research Institutions Contribute research expertise, emerging practices, specialized knowledge, and independent perspectives on enterprise AI development. Federal Fellowship Programs Provide specialized personnel and help expand access to emerging technical and product-development expertise. By late 2023, the Department had completed several successful AI campaigns and had begun building the expertise and organizational confidence required for a larger generative AI initiative. Technology Modernization Fund investment began in May 2024, enabling StateChat to enter a short, targeted alpha phase. Alpha was designed to expose fundamental failures quickly rather than postpone testing until the product appeared complete. The central feasibility question was whether the Department could provide a secure chatbot that avoided unacceptable risks, performed adequately, improved over time, and functioned across a global operating environment extending from Washington headquarters to embassies and consulates. The alpha effort moved StateChat from concept to testable prototype by assembling the delivery team, establishing internal and external partnerships, building a minimum viable product, and rigorously testing core assumptions. The Department advises agencies to build a multidisciplinary team early and scale it strategically; engage cybersecurity, legal, privacy, governance, accessibility, and operational stakeholders from the beginning; obtain the highest appropriate authorization to operate as a foundational activity; release a working but limited prototype; and use layered testing to determine whether the product is ready for broader use. Strategic Partnerships Form internal and external partnerships early to provide the expertise, authority, capabilities, resources, and organizational support required for enterprise AI development. O2_1 2.1 Building an enterprise AI solution is as much an organizational challenge as a technical one. Early engagement helps identify blind spots, clarify requirements, reduce downstream resistance, and establish shared responsibility for adoption. Internal partners included personnel from Diplomatic Technology, Diplomatic Security, DT Enterprise Infrastructure, the Office of Accessibility and Accommodations, Independent Testing, Evaluation, Verification, and Validation, cybersecurity offices, user-experience functions, and foreign-policy programs. Cyber teams helped identify system requirements and supported the approximately year-long authorization process. External relationships included the Technology Modernization Fund, three federal fellowship programs, two universities, four major AI laboratories, technology consultancies, private-sector vendors, research institutions, and other federal agencies. These partners provided implementation expertise, pretrained models, technical capabilities, best practices, and lessons learned that accelerated progress. Delivery Team Assemble and progressively mature a multidisciplinary delivery team with clear product, development, platform, governance, testing, customer-success, and research responsibilities. O2_2 2.2 The StateChat team began as a small and informal group and expanded as the product and user population matured. The playbook identifies six principal functions for a mature generative AI delivery team. The Product Owner is the business leader who sets the vision, translates organizational needs into prioritized features, and secures executive sponsorship to overcome bureaucratic obstacles. Application Development consists of platform engineers and data scientists who build the product. Platform and Governance includes technologists and security analysts who authorize and securely administer the enterprise platform. Testing includes AI testing specialists who evaluate new functionality against expected performance requirements defined by the Product Owner. Customer Success includes communications, training, and change-management specialists who drive adoption and fluency. Innovation and Research includes AI and analytics specialists who explore emerging technologies, prototype new features, and inform development and strategy. The dedicated product owner served as the bridge between technical capabilities and business needs, helping ensure that StateChat addressed priority enterprise problems. Alpha Prototype Build and iteratively improve a minimum viable product capable of testing the solution's core utility, security, performance, accessibility, and global operability. O2_3 2.3 Rapid prototyping and iterative development were central to alpha. Early versions of StateChat were intentionally sparse but sufficient to test whether the product could be useful to Department personnel. Prototyping helped assess feasibility, uncover challenges before major resource commitments, and investigate data security, compliance, and ethical concerns in a controlled environment. The team tested new ideas quickly, gathered user and tester feedback, and adjusted functionality and safeguards through repeated iterations. The lesson was to start before the team felt fully ready. The early prototype permitted only one chat thread per user, deleted the thread after 24 hours, required hours to process uploaded documents, and had no connections to internal Department systems. Despite these limitations, real-user testing immediately revealed the importance of document upload capabilities and connections to Department datasets, enabling the team to prioritize demonstrated needs rather than assumed preferences. Minimum Requirements Define the functional, security, accessibility, risk, performance, and usability conditions the prototype must satisfy before proceeding to broader testing. O2_4 2.4 Alpha development sought to establish that StateChat could meet cybersecurity standards, avoid unacceptable risks such as toxic outputs, perform well enough for beta users, improve over time, and operate throughout the Department's global footprint. Minimum requirements were intended to answer core feasibility questions rather than to require a feature-complete product. The team distinguished between conditions that had to be satisfied before broader exposure and capabilities that could be added through subsequent learning and iteration. Alpha Testing Conduct targeted testing with knowledgeable specialists and willing early adopters to identify foundational safety, security, functionality, access, and usability issues. O2_5 2.5 Alpha testing involved approximately 150 participants over several weeks. The cohort included people with AI expertise, personnel in cybersecurity and responsible-AI functions, representatives of offices with governance and security responsibilities, users capable of validating core functionality, and known generative AI power users. Testers provided feedback on access, identity, authentication, cyber-threat resistance, functionality, and prototype viability. Multiple cybersecurity teams examined whether the system could safely support sensitive information in realistic Department workflows. Testing supported adjustments to controls, functionality, and user experience and helped the system obtain a Conditional Authority to Operate, allowing development to continue into beta. Layered Testing Apply complementary functional, AI-performance, and user-acceptance testing to each major prototype release. O2_6 2.6 The Department used a three-pronged testing approach. Internal functional testing validated technical infrastructure and functionality through stress testing, security-control validation, Section 508 accessibility review, and regression testing. AI performance testing assessed functionality, contextual accuracy, misinformation, cybersecurity, user interface and experience, and reliable platform performance, with reference to the National Institute of Standards and Technology AI Risk Management Framework and government-wide requirements. User acceptance testing brought in subsets of prospective users to determine whether early iterations met real-world needs and to provide usability insights. This layered approach reduced the likelihood that satisfactory performance in one dimension would obscure serious shortcomings in another. Independent Evaluation Use independent testing, evaluation, verification, and validation expertise to assess AI risks and performance against recognized trustworthy-AI characteristics. O2_7 2.7 The Department relied heavily on Independent Testing, Evaluation, Verification, and Validation specialists before launch. Their approach drew from Executive Order 13960 and the National Institute of Standards and Technology AI Risk Management Framework. StateChat was evaluated against seven core characteristics identified in the playbook: Safe; Secure and Resilient; Explainable and Interpretable; Privacy-Enhanced; Fair with Contextual Accuracy Maintained; Accountable and Transparent; and Valid and Reliable. Testing covered ten risk categories tailored to diplomatic AI systems. Independent examination helped establish confidence in the platform's risk posture and supported informed decisions about progression to later development phases. Security Authorization Obtain an authorization to operate at a level that permits the AI solution to use the sensitive information required for meaningful mission utility. O2_8 2.8 Security was treated as a prerequisite for value rather than as a review to be deferred until launch. Securing a FISMA High Authority to Operate required approximately one year and deep cooperation with cybersecurity and governance teams from the outset. Although aiming for a High authorization increased early complexity, it enabled StateChat to integrate sensitive data sources such as cables, SharePoint content, and internal reports; support multiple Sensitive But Unclassified data categories; operate in a secure closed environment; establish understandable user expectations; and provide practical value to personnel working with actual Department information. Without the High authorization, the system would have been limited to unclassified, nonsensitive material, constraining adoption and weakening its core value proposition. Customer Success Establish early user-support, communication, feedback, and community capabilities that surface user needs and prepare the organization for broader testing. O2_9 2.9 Within weeks of beginning alpha, the Department established basic customer-success operations consisting of an office hour, an in-platform feedback button, a weekly newsletter, and a community of practice hosted on Microsoft Teams. These modest mechanisms provided direct contact with users and helped reveal practical needs before beta. The playbook's lesson is that customer success should begin at the ground floor rather than being added only after technical development is complete. Beta Readiness Confirm that testing results, safeguards, core functionality, and user experience justify progression from a limited alpha prototype to broader beta development. O2_10 2.10 The feedback loop between testers and developers enabled early improvements to core functionality, security controls, access, and user experience. Conditional authorization permitted development to continue, while layered testing provided evidence that the system could safely be offered to a larger group. Alpha completion did not mean the product was complete; it meant the team had sufficient confidence that the central feasibility assumptions had survived disciplined testing and that broader user participation could produce useful additional learning. Usefulness Demonstrate that users adopt the enterprise AI solution, derive practical value from it, and help guide its development toward enterprise readiness. G3 3 Beta Users Test emerging StateChat capabilities in practical workflows, report issues, propose enhancements, share successes, and demonstrate whether the product delivers useful value. Power Users Provide informed and detailed feedback, participate in interviews and prioritization, demonstrate advanced uses, and help identify high-value capabilities. User Support Personnel Create guidance, answer questions, operate support channels, resolve problems, and help users build confidence and fluency. Communications Personnel Explain beta limitations, share development progress and use cases, manage expectations, and maintain transparent relationships with users and leaders. Change-Management Specialists Design support and engagement approaches around user personas, needs, barriers, and differing levels of readiness. Department Leaders Support controlled expansion, communicate organizational backing, consider user evidence, and help establish readiness for enterprise launch. Product Development Team Collects and interprets usage evidence, prioritizes requests, communicates decisions, and iteratively improves the product. After alpha demonstrated that a functional generative AI chatbot could be offered securely to a limited global user group, beta sought to determine whether personnel would actually adopt StateChat and find it useful. During five months of beta development, the user population increased from approximately 150 to 8,000. The Department used this controlled expansion to showcase emerging capabilities, propagate successful uses identified by users, establish scalable development and support processes, and build a backlog of high-demand features. The playbook advises agencies to expand gradually, create a comprehensive change-management strategy, understand user personas and pain points, provide multiple communication and support channels, listen to users, prioritize transparently, communicate development decisions openly, and establish explicit conditions for ending beta. Users were treated as development partners rather than passive recipients of a finished system. Controlled Expansion Expand access in manageable stages so the team can learn from users, refine processes, and develop scalable support without overwhelming personnel or systems. O3_1 3.1 StateChat expanded from 150 to 8,000 users over five months. Phased growth enabled the Department to repeat and refine development and testing processes throughout the product lifecycle, learn from early users, prepare introductory resources, and deliver them to successive waves without creating excessive user or team burnout. Volunteers requested access through an automated form and could participate when they authenticated with a Department credential. Controlled expansion created an implicit agreement with users: remain engaged and the team would continue building with them. Expectation Management Communicate the beta product's limitations, development status, and evolving capabilities honestly so users understand the purpose and conditions of participation. O3_2 3.2 The Department deliberately emphasized that the product was unfinished. The recurring message, “StateChat is in beta, and it shows,” conveyed humility and set realistic expectations with both users and leaders. This openness reduced the risk that temporary limitations would be interpreted as permanent failures and encouraged users to view themselves as partners in improvement. Communications also distinguished current capabilities, planned work, and matters outside the team's scope. User Support Provide coordinated support resources and communication channels that help users understand, access, test, and productively apply the beta system. O3_3 3.3 The developing support system included office hours, basic guidance, resource hubs, feedback mechanisms, communities of practice, and direct communications. These channels helped users overcome access and usage barriers while giving the delivery team continuous insight into confusion, unmet needs, and promising applications. The playbook recommends designing support around distinct user personas and their pain points rather than assuming that one resource or communication method will serve everyone. Multichannel Communications Use multiple communication channels to build awareness, explain development, share useful applications, and reach users in differing organizational contexts. O3_4 3.4 The Department used weekly email updates, intranet announcements, leadership presentations, and other channels to reach personnel where they already obtained information. This approach provided multiple entry points for learning about StateChat, increased visibility, generated enthusiasm, and encouraged innovation by circulating use cases and success stories. Transparent communication included what the team was building, why particular features were prioritized, and which requests were not being pursued. Feedback Channels Provide accessible and varied mechanisms through which users can report problems, suggest enhancements, describe successes, and evaluate their experience. O3_5 3.5 Dedicated channels included integrated feedback forms, customer-satisfaction surveys, office-hour questions, interviews, observations, and other user interactions. Multiple avenues accommodated differences in user preferences, available time, expertise, and willingness to provide feedback. Quantitative evidence on usage patterns, prompting success rates, and satisfaction supplemented qualitative reports, providing a more complete picture of performance before enterprise launch. The playbook warns that useful feedback will arrive immediately and in volume; without a collection and organization system, teams may drown in requests or overlook the needs that matter most. Feedback Prioritization Collect, synthesize, evaluate, and prioritize user feedback through a structured and transparent process that aligns development with practical needs and organizational priorities. O3_6 3.6 The Department developed a structured approach called Thunderdome. First, the team collected input from multiple sources, including integrated forms, satisfaction surveys, office hours, and observations, so that prioritization would reflect a broad audience rather than only the loudest voices. Second, the team reviewed the input, identified recurring themes, and surfaced potential features and improvements. Third, participants conducted a focused prioritization session to consider technical feasibility, organizational alignment, and potential impact. Fourth, users voted on the narrowed list of enhancements. Fifth, the team used the results to create and publish a high-level roadmap, generally organized in six-month increments. This converted scattered feedback into actionable development choices and created visible accountability. User-Informed Roadmap Create and communicate a development roadmap that reflects user priorities, technical feasibility, organizational needs, and transparent product decisions. O3_7 3.7 The first Thunderdome synthesized feedback from six channels, approximately 800 survey respondents, and 22 interviews with StateChat power users. It produced a six-month roadmap reflecting the ten most-requested user features. By publishing the outcome, the Department enabled users to see how their participation affected product decisions, strengthened trust, and demonstrated that feedback was not merely collected but acted upon. The roadmap did not imply that every request would be fulfilled; it documented which items were prioritized and provided reasons grounded in impact, feasibility, and organizational direction. Continuous Improvement Maintain repeated cycles of feedback, prioritization, development, testing, communication, and refinement in partnership with users. O3_8 3.8 The Department sought to solicit feedback, acknowledge its receipt, explain how it was being considered, and show when user input influenced new features or applications. The playbook recommends continuous improvement rather than one-time consultation. Ongoing prioritization and communication establish a durable relationship in which users contribute experience, the team makes and explains development choices, and the product evolves in response to observed needs and performance. Usage Evidence Use quantitative and qualitative evidence to determine whether users engage with the beta system, succeed in applying it, and perceive meaningful value. O3_9 3.9 The Department combined user comments, interviews, survey responses, office-hour observations, usage patterns, prompting success rates, satisfaction data, and sustained engagement. This mixed evidence helped distinguish isolated enthusiasm from repeatable utility and allowed the team to identify functionality requiring improvement. User validation was treated as one of several conditions for launch rather than as a substitute for technical stability, security, and operational readiness. Enterprise Readiness Determine that technical stability, user validation, support capacity, and organizational commitment are sufficiently aligned to conclude beta and proceed to enterprise launch. O3_10 3.10 Beta was not intended to continue indefinitely. The Department prepared for enterprise deployment when core functionality performed reliably under load, critical defects had been resolved, users provided consistently positive feedback and sustained engagement, support systems could accommodate scale, and organizational backing was sufficient. The team distinguished mission-critical functions requiring near-perfect reliability from lower-risk capabilities for which controlled imperfections could support faster deployment and learning. Launch was treated as the next stage of an evolving product, not as the end of maintenance, improvement, or user engagement. 2026-07-01 https://www.state.gov/wp-content/uploads/2026/07/DOS-Generative-AI-Playbook_July-2026.pdf Owen Ambur Owen.Ambur@verizon.net