Design observability for AI apps and agents selling through Microsoft Marketplace
In the last post, API resilience and reliability patterns for AI apps and agents, we focused on what happens when AI systems encounter failure—and how resilient execution paths keep that failure contained. Timeouts fire with intent. Retries stay bounded. Circuit breakers provide overload protection. When resilience is designed well, your system continues to function even as conditions change, forming the foundation of AI reliability engineering. You can always get curated step-by-step guidance through building, publishing and selling apps for Marketplace through App Advisor. This post is part of a series on building and publishing well-architected AI apps and agents in Microsoft Marketplace. The series focuses on AI apps and agents that are architected, hosted, and operated on Azure, with guidance aligned to building and selling solutions through Microsoft Marketplace. Observability for AI systems AI apps and agents are shifting traditional observability, which was designed for systems based on simple assumptions, where requests followed linear paths and workloads behaved predictably. Execution in AI systems consumes tokens at a highly variable rate rather than fixed compute units. Requests unfold across multiple reasoning steps. Agents perform work that spans APIs, models, retrieval layers, and applications. A single interaction may pause, branch, retry, or exit early depending on inferred intent, context, and constraints. Instead of asking whether services are running, observability for AI systems asks: what is the system doing right now—and why? Is an agent spending its time reasoning, waiting on dependencies, retrying tool calls, or exiting early due to enforced limits? Is cost increasing because value is increasing, or because execution paths are expanding without progress? AI observability requirements shift the focus in the following subtle, but critical ways: From resource availability to workflow state From performance metrics to signals From incidents to patterns Core observability dimensions for AI apps and agents Once observability shifts toward understanding behavior, clarity comes from tracking state across the agents in the workflow. For AI apps and agents, observable indicators, such as those detailed below, show how work unfolds and changes during real usage—especially in trials and early adoption: Execution flow shows how a request moves through agents, tools, and workflows. This highlights where execution progresses smoothly, where it slows, and where it concludes early. This makes agent outcomes explainable and keeps behavior consistent across tenants. Cost and token behavior reveals how execution translates into consumption. Token usage per request, per agent step, and per retry shows where value is being delivered and where execution paths expand without proportional benefit. This insight connects runtime behavior directly to Marketplace billing expectations and evaluations. Latency and wait states distinguish active processing from time spent waiting on dependencies. Seeing where time is consumed helps explain slow experiences and guides decisions about optimization, caching, or resilience improvements. Failure classification provides structure when systems degrade and supports effective AI incident management. Separating tool failures from planning failures, and transient issues from terminal exits, keeps investigations focused and prevents protective behavior from being misread as instability. Tenant‑level patterns surface how behavior repeats at scale. Uneven load, and recurring degradation often appear first during trials and shape the customer's perception. Together, these dimensions turn telemetry into understanding—supporting clearer conversations, faster triage, and predictable execution as usage grows. Why observability matters By this point in the journey, your AI app or agent has implemented bounded execution paths, cost controls, and quality of service safeguards. As a result, failure degrades gracefully instead of spreading. These resilience techniques determine how your solution behaves under pressure. The data gathered from observability platforms like Application Insights and Azure Monitor explains why it behaves that way. For AI and agentic systems, infrastructure health alone rarely answers the questions that matter. Services can be up, CPUs can be idle, and queues can look healthy while agents loop inefficiently, retries quietly expand cost, or workflows exit early without delivering value. From the customer’s perspective, the experience feels inconsistent even though the platform appears stable. AI app observability closes this gap by revealing system behavior rather than system status. It shows how requests move, where work concentrates, and how constraints shape outcomes. At Marketplace scale, these patterns repeat across tenants and trials. What appears once during an evaluation often appears again as adoption grows. Observability connects runtime behavior back to the design choices introduced in earlier posts: Usage‑based billing introduced variability in consumption Performance optimization introduced tradeoffs among latency, quality, and cost Resilience patterns introduced controlled failure and bounded execution Observability allows you to explain outcomes during trials, validate assumptions as usage grows, and support post-launch AI operations confidence across customers and environments. Without this visibility, teams react to symptoms. With it, they recognize patterns. From execution paths to behavioral signals Observability begins at the same place resilience begins—API boundaries. These boundaries define where responsibility shifts and where behavior becomes visible. Observability focuses on signals that explain decisions made by the system as it executes instead of relying on raw logs that describe isolated events. Every resilience mechanism emits behavioral signals. Viewed together, these signals provide far more value than logs alone. Logs answer whether something happened. Behavioral signals explain why it happened and how the system responded. Circuit breakers change state as load builds and recedes. Retry loops show whether failures resolve quickly or exhaust their limits. Timeout enforcement reveals where dependencies slow execution. Fallback paths and early terminations show how the system protects itself while preserving outcomes for customers. This perspective matters most for agents. Agent execution unfolds as a series of choices—plan, call a tool, retry, exit early—rather than a single request‑response cycle, which requires monitoring AI agent behavior to remain understandable and consistent at scale. Observability that tracks these decisions makes agent behavior understandable, consistent, and defensible as usage grows across customer tenants. Observability at the agent layer As AI systems become more agent‑driven, observability needs to move closer to where decisions are made. Agents introduce variability by design. They plan, adapt, and choose workflow paths dynamically. Without first‑class visibility into that behavior, execution can appear unpredictable even when the underlying system is healthy. Observability at the agent layer acts as the feedback loop that keeps execution safely bounded. It shows how agents use the freedom you give them—and where that freedom begins to stretch into inefficiency. Observability follows how the agent did its job instead of treating the agent’s interaction as a single outcome. Several indicators help make agent behavior understandable. Step count per request reveals how much reasoning effort a prompt requires. Planning iterations show whether an agent converges quickly or cycles through alternatives. Tool invocation frequency highlights when agents rely heavily on external systems. Early exits compared to full completion explain whether limits and fallbacks activate as designed. Taken together, these indicators help distinguish healthy exploration from inefficient reasoning and degraded execution. An agent exploring briefly before converging adds value. An agent looping through tools without progress signals pressure, uncertainty, or dependency issues. This distinction reinforces a core principle of agentic systems: models reason probabilistically, adapting to context as it changes. Your system observes deterministically—measuring execution, enforcing boundaries, and clarifying outcomes. When those roles stay separate and well‑instrumented, agent behavior becomes transparent, predictable, and ready for Marketplace scale. Observability across environments The type of Marketplace offer you choose shapes what observability customers expect and how responsibility is shared. For SaaS offers, publishers typically own end‑to‑end execution. Observability centers on agent behavior, workflow completion, token usage, latency, and dependency impact across tenants. Publishers rely on consistent signals—often surfaced through tools like Azure Monitor, Application Insights, and Microsoft AI Foundry—to explain how requests behave as scale and load increase. For container‑based offers and Azure Managed Applications, observability expectations are more distributed. Publishers expose clear execution outcomes, limits, and failure signals at application boundaries. Customers, in turn, observe infrastructure health, scaling behavior, and downstream systems within their own environments. This separation ensures each party has visibility into what they control without creating ambiguity. Learn more about Choosing your marketplace offer type for AI Apps and agents. Execution behavior differs across environments for predictable reasons. Scale increases, tenant mix broadens, and external dependencies behave differently under real load. What must stay consistent is how behavior is interpreted. Signal definitions, thresholds, and failure classification should mean the same thing in Dev, Stage, and Prod. Learn more about designing a reliable environment strategy for Microsoft Marketplace AI apps and agents. Staging environments are where this consistency is validated. Observing retries, timeouts, and graceful degradation before production prepares you for Marketplace evaluations, which often resemble production conditions. Observability gaps tend to appear first during customer evaluation—when clarity matters most. Publisher and customer visibility boundaries Purpose: Parallel Post #13 responsibility clarity, now for observability As observability matures across environments, clarity around responsibility becomes essential. For Marketplace solutions, trust grows when publishers and customers each see what they own—and understand where that visibility ends. Publishers are responsible for instrumenting execution paths end to end. That means making workflows traceable, limits visible, and failure modes explainable. Observability should surface behavior—how requests progressed, where execution concluded, and why—rather than exposing raw internal errors that require insider knowledge to interpret. Customers focus their observability on what they control. This includes monitoring downstream systems, infrastructure behavior, and environment‑level alerts within their own estate. When visibility aligns with ownership, teams can act quickly and decisively. Exposing too much internal detail can overwhelm customers and blur accountability. Observing too little behavior creates friction, especially when issues cross boundaries and lack context. Clear visibility enables faster triage, sharper ownership boundaries, and fewer escalations rooted in ambiguity. Observability as an enabler for scale, billing, and trust From a customer’s perspective, observability answers two fundamental questions: Can I understand what happened? and Can I trust this at scale? When the answer to both is clear, observability becomes part of the value your Marketplace offering delivers. When system behavior is visible and explainable, customers gain confidence that adoption and growth will remain predictable. Observability directly supports usage‑based billing by tying execution behavior to measured consumption. Clear visibility into token usage, retries, and execution paths helps validate how usage is calculated and supports transparent billing conversations. It also enables ongoing performance tuning and caching strategies by showing where latency accumulates, where work repeats, and where optimization delivers measurable impact. Observability reinforces confidence in resilience mechanisms, confirming that limits, fallbacks, and degradation paths activate as designed under real‑world conditions. Beyond validation, observability creates a continuous feedback loop. Execution data informs pricing adjustments, guides changes to limits, and helps refine default configurations as customer behavior evolves. What’s next in the journey With execution behavior observable and explainable, the focus shifts to how AI systems are operated safely as change accelerates. The upcoming posts will discuss deployment strategies, CI/CD pipelines for agents, and progressive rollouts build on this foundation—ensuring AI apps evolve confidently as usage and expectations grow. Key Resources See curated, step-by-step guidance to help you build, publish, or sell your app or agent (no matter where you start) in App Advisor Quick-Start Development Toolkit can connect you with code templates for AI solution patterns Microsoft AI Envisioning Day Events How to build and publish AI apps and agents for Microsoft Marketplace Get over $126K USD in benefits and technical consultations to help you replicate and publish your app with ISV SuccessDiscover new Microsoft Marketplace innovations announced at Microsoft Build
At Microsoft Build, Microsoft shared new opportunities for software development companies and partners to build, scale, and monetize AI apps and agents through Microsoft Marketplace. Explore how Microsoft Marketplace is helping software companies accelerate go-to-market strategies, expand customer reach, simplify procurement, and unlock new revenue opportunities across the Microsoft ecosystem. Learn how organizations can take advantage of Azure and Marketplace capabilities to support AI innovation and deliver enterprise-ready solutions faster. Whether you’re building intelligent applications, growing your commercial marketplace presence, or exploring new ways to monetize AI-powered solutions, this is a valuable resource for understanding the latest Microsoft Marketplace announcements and opportunities coming out of Build. 👉 Read more: Build, scale, and monetize apps and agents with Microsoft MarketplaceAccelerate your AI or agent build to sell on Marketplace with Quick-Start Development Toolkit
Want to skip right to coding in minutes? Start with the interactive wizard in App Advisor Building AI products quickly is becoming table stakes. Building them in a way that supports scalability, repeatability, and a path to commercialization is where software companies create advantage. The challenge now is reducing the time between identifying an opportunity and getting developers working inside a proven structure that supports real deployment outcomes. That’s where the AI, agentic, and Copilot branch of the Quick-Start Development Toolkit helps. Embedded directly within App Advisor, Quick-Start Development Toolkit helps software companies move from concept to implementation faster using guided development patterns, trusted architectures, deployable reference code, and practical resources designed to reduce friction across the development process. Build AI & agentic products faster without starting from scratch Development teams often know the customer scenario they want to solve. What slows momentum is deciding where to begin, selecting architecture patterns, and aligning implementation decisions across teams. The Quick-Start Development Toolkit helps remove that uncertainty. By answering a few focused questions about what you want to build, who it serves, and the products you’re building with, you’re matched with a development pattern designed to accelerate execution. Each development pattern includes: Self-serve, click-to-deploy reference code aligned to your scenario, Sample solution architecture to help visualize products and reduce guesswork, and Practical how-to resources and implementation guidance to overcome friction points, Everything is structured to support faster decision making and help teams move confidently into development. Accelerate development with purpose-built AI accelerators The AI and agent branch of Quick-Start Development Toolkit includes development accelerators designed around high-value scenarios, so your team can spend less time assembling foundations and more time building differentiated experiences. Each of these accelerators is built and fully maintained by Microsoft experts, so you can be confident your code template isn’t stale. Our most popular accelerators include: Multi-Agent Custom Automation Engine Accelerator: Delegate complex, repetitive tasks to AI agents that act on your behalf—executing work efficiently, reducing manual effort, and ensuring results align with your organization's standards. Conversation Knowledge Mining Accelerator: Improve contact center performance with AI-powered conversation intelligence—analyzing audio and text data on a large scale to show insights, improve service, and drive smarter decisions. Accelerate agentic applications for Unified Data Foundations (with Microsoft Fabric): Accelerate decision making at scale with secure, agentic AI built on a unified data foundation with two use cases for sales performance and customer insights. Each pattern includes common use cases, related resources, and pathways to adjacent scenarios so teams can continue progressing without losing momentum. The goal is to help your team move from experimentation to a product that can be packaged, deployed, and prepared for customers. You can see more of our accelerators here Coming this week: The Microsoft IQ solution accelerator leverages a shared intelligence layer to unify data, knowledge, and workflows, enabling AI-powered insights and coordinated actions for measurable business outcomes. Build with Microsoft Marketplace outcomes in mind Development choices shape commercial outcomes. Starting with trusted architecture and structured implementation guidance can help reduce redesign cycles later when preparing to package, publish, and scale. Quick-Start Development Toolkit helps software companies: Shorten time from idea to deployable AI product, Improve alignment across implementation decisions, Reduce development overhead through reusable foundations, and Create repeatable pathways toward publishing and selling. When development starts with clarity, commercialization becomes easier. Keep moving forward with App Advisor Quick-Start Development Toolkit is embedded within App Advisor because building is only one stage of the journey. App Advisor helps connect decisions across design, development, publishing, and growth so teams can continue moving forward with less context switching and more confidence. As your solution evolves, App Advisor provides curated, step-by-step guidance to help you prepare for Marketplace readiness and make the next decision faster. Ready to start? Explore Quick-Start Development Toolkit Start where you need help with App Advisor
Publishing readiness for AI apps and agents in Microsoft Marketplace
Discover how to prepare AI apps and agents for publishing in Microsoft Marketplace. This Marketplace Community article explains why readiness starts before Partner Center, focusing on the operational, technical, and organizational foundations required to ensure solutions can be evaluated, purchased, and operated reliably. As AI systems manage identity, data, runtime behavior, and subscription lifecycles, gaps in readiness can create friction during certification and customer adoption. Clearly defined identity boundaries, consistent data handling practices, and predictable responses to subscription events help ensure solutions behave as expected across environments and tenants. Learn how to establish publishing readiness that supports smooth certification, reliable operations, and confident customer adoption at Marketplace scale. Read more: Publishing readiness for AI apps and agents on Microsoft MarketplaceDesign CI/CD pipelines for AI apps and agents in Microsoft Marketplace
Discover how to design CI/CD pipelines for AI apps and agents selling through Microsoft Marketplace. This Marketplace Community article explains why controlling how changes reach production is essential for maintaining predictable behavior, reliability, and customer trust. As AI systems evolve through updates to code, models, prompts, and agent logic, behavior can change in ways that impact cost, performance, and outcomes. Structured pipelines that isolate change, validate behavior, and enable safe promotion and rollback help ensure updates are introduced deliberately—without unexpected impact across environments or tenants. Learn how to design CI/CD strategies that support safe iteration, controlled releases, and consistent behavior as AI solutions scale in Marketplace environments. Read more: Design CI/CD for AI apps and agents selling through Microsoft MarketplaceDesign reliable environment strategies for AI apps and agents in Microsoft Marketplace
Discover how to design a reliable environment strategy for AI apps and agents selling through Microsoft Marketplace. This Marketplace Community article explains why structured Dev, Stage, and Production environments are essential for safe updates, predictable behavior, and long‑term customer trust. As AI systems evolve through prompt updates, model changes, and shifting data contexts, behavior can vary across environments. Clear environment separation, controlled promotion paths, and consistent configuration boundaries help prevent regressions, support validation, and ensure changes can be introduced safely without impacting production workloads. Learn how to design environment strategies that enable confident iteration, support Marketplace readiness, and help customers operate solutions predictably at scale. Read more: Designing a reliable environment strategy for Microsoft Marketplace AI apps and agentsEnforce AI entitlements using Marketplace commerce signals
Discover how to enforce entitlements in AI apps and agents using Microsoft Marketplace commerce signals. This Marketplace Community article explains why purchase and subscription data must be integrated at runtime to ensure customers only access what they’ve paid for. As AI apps and agents dynamically invoke tools, expose capabilities, and operate without constant user input, static enforcement approaches fall short. Translating Marketplace signals into deterministic runtime behavior—across SaaS, containers, virtual machines, and managed applications—ensures access is controlled, auditable, and aligned with subscription state. Learn how to design entitlement enforcement that remains consistent through plan changes, scaling workloads, and real‑time agent decisions. Read more: Integrate Marketplace commerce signals to enforce entitlements in AI appsDiscover how AI-powered agents on Microsoft Fabric are accelerating retail merchandising decisions
Retail organizations are under increasing pressure to move faster and make smarter, data-driven decisions at scale. In this latest Marketplace Partner Spotlight, Microsoft highlights how AI agents built on Microsoft Fabric are helping merchandising teams transform complex operational data into actionable insights—without leaving the security of their existing data environment. By leveraging Microsoft Fabric and OneLake as a unified data foundation, partners like Lucid Data Hub are enabling retailers to automate time-intensive reporting processes and shift toward continuous, insight-driven workflows. These business-ready AI agents can analyze large volumes of sales and operational data, surface meaningful trends, and deliver clear recommendations—empowering buyers and store leaders to act faster and with greater confidence. The impact is tangible: merchandising teams can reduce hours of manual analysis into minutes, uncover item-level performance insights, and identify opportunities across store clusters to optimize outcomes. If you’re exploring how AI agents, Microsoft Fabric, and the Microsoft Marketplace ecosystem can drive intelligent automation in retail, this article offers practical insights and real-world examples to help you get started. 👉 Read the full article AI agents on Microsoft Fabric for faster retail merchandising decisionsAI agents on Microsoft Fabric for faster retail merchandising decisions
For our latest in the Partner Spotlight series, we’re highlighting a partner building business-ready AI agents on Microsoft Fabric, so organizations can turn governed enterprise data into faster decisions and automated workflows. I connected with the team at Lucid Data Hub to learn how Lucid Agents Hub brings agentic experiences directly to customers’ data in OneLake, helping retail teams move beyond manual reporting and into repeatable, insight-driven action. About Venu Amancha, Founder & CEO, Lucid Data Hub builds business-ready AI agents that run directly on enterprise data within Microsoft Fabric. Our platform, Lucid Agents Hub, enables organizations to move beyond reporting and into automated, insight-driven workflows without moving data outside their existing security boundaries. _______________________________________________________________________________________________________________________________________________________________ [JR] Who is your solution designed for, and what does it help them do? [VA] Lucid Agents Hub is designed for teams who need to make frequent, high-impact decisions from large volumes of operational data especially merchandising teams, buyers, and store operations leaders in retail. Instead of spending hours assembling recaps and interpreting dashboards, they can receive agent-generated insights and clear, actionable recommendations on a predictable cadence. The AI agent eliminated that manual cycle entirely. It now surfaces those insights automatically, every week, in minutes. One example is our Retail Sales Performance AI Agent, which automates the weekly sales insights cycle for merchandising teams and buyers by analyzing millions of rows of weekly sales, item, and store data across banners and store clusters. [JR] Can you give an example of how the Retail Sales Performance AI Agent solved a customer’s problem? [VA] At Heritage Grocers Group, merchandising teams spent 5+ hours every week manually building sales recaps. They could see what happened—but not why. Buyers lacked a clear view of category trends, item-level performance, quantity shifts, and store-cluster patterns. The Retail Sales Performance AI Agent eliminated that manual cycle. It now surfaces those insights automatically every week in minutes detecting item-level declines, identifying fast-moving margin-positive SKUs, flagging underperforming items by store cluster, and delivering recommendations directly to buyers and store managers. [JR] Which Microsoft technologies or services are foundational to what you’re building? [VA] The solution runs natively on Microsoft Fabric, using OneLake as the unified data layer. Our agents operate directly on enterprise data and inherit existing governance and access controls without additional configuration. Outputs flow into the dashboards, collaboration platforms, and reporting workflows customers already use, so insights show up where decisions get made. Microsoft Fabric was a deliberate choice, not just a default. Our enterprise customers especially in retail already have their critical data living in the Microsoft ecosystem. OneLake means there’s a single, governed copy of that data. No duplication, no movement, no additional risk surface. Our agents read directly from that layer, which means the security and compliance boundaries that customers have already invested in carry over automatically. The value of building on Fabric goes beyond the technical architecture. It fundamentally changes how enterprise buyers evaluate and procure a solution like ours. When IT and security teams see that agents operate entirely within their existing Fabric environment with role-based access controls, workspace permissions, and audit logs they already control. It removes the largest barrier to enterprise adoption: trust. Procurement conversations that used to require months of security review cycles are now dramatically faster. What we didn’t fully anticipate was how much Fabric’s native integration capabilities would simplify end-to-end delivery. Going in, we expected to spend significant engineering time on data pipeline infrastructure. What we found instead was that Fabric’s data ingestion, lakehouse, and compute layers fit together in a way that let our team focus almost entirely on agent logic and business outcomes, not infrastructure plumbing. That shift in where we spend our effort has meaningfully accelerated how quickly we can deploy for new customers and extend to new use cases. [JR] How are you using AI today in Lucid Agents Hub, and what business outcomes have customers seen? [VA] We use Microsoft Azure AI Foundry for core AI and language model capabilities, and Microsoft Fabric Copilot (Fabric IQ) as the data and compute backbone. Together, they power agents that analyze weekly sales data across banners and store clusters, generate narrative-quality insights at the category and SKU level, and deliver clear recommendations without human intervention in the analysis cycle. 5+ hours of weekly manual effort eliminated Item-level sales declines and fast-moving margin-positive SKUs surfaced automatically Top-growth categories and underperforming items identified by store cluster Recommendations delivered directly to buyers and store managers weekly This all leads to faster decisions, stronger merchandising actions, and measurable improvements in product mix, availability, and overall sales performance. [JR] Any architectural decisions or best practices you’d recommend to other partners building agents? How did you approach building securely? [VA] We designed the solution as a coordinated set of specialized agents one for data ingestion, one for validation, and one for insight generation and delivery. Each agent owns a focused task, and together they run as a connected, end-to-end workflow. This makes the system easier to maintain, consistent in its logic, and straightforward to extend to new banners, categories, or use cases. Agents run entirely within the customer’s Microsoft Fabric environment data never leaves the customer’s security perimeter. All access controls, role-based permissions, and governance policies are inherited directly from Fabric. [JR] What motivated you to publish on Microsoft Marketplace? And did you use any Microsoft tools or benefits to support your publishing process? [VA] Publishing on Microsoft Marketplace was a straightforward decision. It gives enterprise customers immediate confidence that they’re procuring from a trusted, Microsoft-validated source instead of navigating a separate vendor relationship. It also simplifies procurement transactions run through an established Microsoft channel; so, customers can move faster than in traditional sales cycles. And it expands our reach to buyers already operating in the Microsoft ecosystem who actively look to Marketplace for solutions. We actively use Marketplace Rewards, which has been valuable for amplifying go-to-market efforts and accessing Microsoft co-marketing resources. We also leverage AI-enabled Marketplace Listing Optimization and related Marketplace content guidance provided through Marketplace Rewards. We used this support primarily to improve our marketplace messaging, positioning, and listing content so it would better resonate with enterprise buyers evaluating solutions within the Microsoft ecosystem. [JR] What key takeaways would you share with other partners building and publishing agents? Any unexpected wins or challenges along the way? [VA] Building and publishing agents can be a complicated endeavor. To other partners, we’d say, start with workflows that are repetitive and directly tied to decisions weekly merchandising recaps are a perfect example. Think end-to-end, not task by task. And build on governed enterprise data from the start, because that’s what drives trust and adoption. An unexpected win was how quickly merchandising teams adapted. Receiving plain-language summaries broken down by banner, store cluster, category, and SKU was more accessible than navigating dashboards. Teams made faster, more confident decisions without needing to interpret raw data themselves. _______________________________________________________________________________________________________________________________________________________________ Closing reflection Lucid Data Hub shows how agents built on Microsoft Fabric can turn governed enterprise data into repeatable, decision-ready insight helping teams act faster while keeping security boundaries and access controls intact.How governed agentic AI is transforming professional services workflows
Discover how partners are building industry-specific AI solutions in Microsoft Marketplace using governed, agentic workflows tailored for professional services. In the latest Partner Spotlight, Richard Baskerville, Senior Director of Strategic Global Alliances at Intapp, shares how “Firm AI” delivers purpose-built capabilities aligned to the unique needs of professional and financial services firms—combining deep workflow expertise with the scale, security, and compliance of Microsoft Azure. This article explores how agentic AI enables end-to-end workflow automation while maintaining human accountability, helping firms modernize operations across client engagement, risk management, and business processes. See how Microsoft Marketplace empowers partners to bring trusted, enterprise-ready AI solutions to customers, accelerating adoption and unlocking scalable growth opportunities. Read the full article: Firm AI for professional services: governed, agentic workflows built on Microsoft Azure | Microsoft Community Hub
