<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Ashley Hromatko on Medium]]></title>
        <description><![CDATA[Stories by Ashley Hromatko on Medium]]></description>
        <link>https://medium.com/@ashley.hromatko?source=rss-c8f57eb6d3ac------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*ISipC0NV_E33bkeus_CFkA.jpeg</url>
            <title>Stories by Ashley Hromatko on Medium</title>
            <link>https://medium.com/@ashley.hromatko?source=rss-c8f57eb6d3ac------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 22:33:38 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@ashley.hromatko/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[The Unexpected Lessons of a Corporate Split: A FinOps / ITFM Perspective]]></title>
            <link>https://medium.com/@ashley.hromatko/the-unexpected-lessons-of-a-corporate-split-a-finops-itfm-perspective-da6d24ea32bd?source=rss-c8f57eb6d3ac------2</link>
            <guid isPermaLink="false">https://medium.com/p/da6d24ea32bd</guid>
            <category><![CDATA[itfm]]></category>
            <category><![CDATA[finops]]></category>
            <dc:creator><![CDATA[Ashley Hromatko]]></dc:creator>
            <pubDate>Tue, 20 Oct 2026 16:58:59 GMT</pubDate>
            <atom:updated>2026-10-20T16:58:59.594Z</atom:updated>
            <content:encoded><![CDATA[<h3>Lessons from FinOps During a Major Corporate Split</h3><p><em>What a year of corporate separation taught me about technology, financial flexibility, and the power of people working together.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wVkmQ7HltBrm6JS3nIm5dg.png" /></figure><p>When a company announces a major separation, the headlines tend to focus on business strategy, shareholder value, and the opportunities ahead for two independent companies.</p><p>What doesn’t make the headlines is everything that has to happen behind the scenes in IT to make that separation possible.</p><p>Over the past year, I’ve had a front row seat to the complexities of navigating a major corporate split from a FinOps and IT Financial Management (ITFM) perspective.</p><p>And if there’s one thing I’ve learned, it’s that separating a company is about much more than moving applications, dividing contracts, and standing up new infrastructure.</p><p>It’s about understanding how technology actually supports the business and discovering just how much decisions made years earlier influence your ability to change.</p><p>I remember sitting in a room early in the process, trying to understand the magnitude of what was ahead of us.</p><p>One of the questions was simple: <strong>How many applications do we actually have?</strong></p><p>Estimates around the room ranged from 600 to 1,400.</p><p>That’s quite a range when you’re trying to build a migration schedule, estimate costs, identify dependencies, and determine which applications belong to which future company.</p><p>Once we had our inventory and understood the challenge ahead, it was time to get moving.</p><p>Looking back, that moment captures one of my biggest takeaways:</p><p><strong>A major company split doesn’t necessarily create your technology management challenges. It exposes the ones you already have.</strong></p><p>Here are some of the lessons I’d share with other FinOps practitioners, ITFM leaders, and technology executives.</p><p><strong>1. Look at your quarterly earnings report. That’s how you should think about tagging.</strong></p><p>I’ll start with something we actually got right.</p><p>Shortly before our separation, we had updated our cloud tagging strategy to align applications and resources with the company’s revenue reporting segments.</p><p>At the time, the objective was better financial visibility and accountability. When the separation was announced, that decision became incredibly valuable.</p><p>We already had a foundation for identifying which cloud resources supported each business.</p><p>It wasn’t perfect. Shared applications and corporate functions such as Finance, IT, Marketing, and Legal introduced complexity. Not every application fit neatly into one segment.</p><p>As the separation progressed, other tagging attributes became critical. Application ownership helped us identify who was accountable for decisions. Disposition tags helped track whether workloads were moving, staying, or being cloned. Those same attributes later became important for Transition Service Agreement (TSA) reporting and cross company charges.</p><p>Now that the companies are establishing their future operating models, updating ownership information is one of our most important cleanup activities.</p><p><strong>My advice:</strong> Open your company’s latest quarterly earnings report. Look at how your organization reports revenue and performance to investors. Then ask yourself whether you can report technology costs using those same business dimensions.</p><p>If the answer is no, that’s a gap worth closing.</p><p>And don’t stop at cloud. Those relationships should extend to on premises infrastructure, SaaS, software licensing, and hardware.</p><p><strong>2. Know your inventory. And I mean really know it.</strong></p><p>Remember those guesses of 600 to 1,400 applications?</p><p>That uncertainty was the result of years of applications being deployed, acquired, changed, and managed across different teams and systems.</p><p>We had a CMDB, application records, and information embedded in cloud resource tags. What we didn’t have was one consistently maintained picture that answered all the questions needed for a separation.</p><p>So teams got to work.</p><p>We pulled information together from multiple sources and built a tool to support rapid application and migration reviews.</p><p>For each application, we needed to understand where it lived, which company it belonged to, where it needed to move, its dependencies, technical complexity, and final disposition.</p><p>What started as an inventory exercise became the foundation of our migration plan and one of the most valuable artifacts both companies were left with.</p><p>The lesson goes beyond applications.</p><p><strong>Knowing where infrastructure physically resides is not the same as knowing which business it supports.</strong></p><p>Imagine a data center assigned to Company A that hosts applications belonging to Company B.</p><p>The physical infrastructure might transfer to one company, but the workloads still need to move elsewhere.</p><p>When application to server relationships are incomplete, the burden falls on application teams to figure out where their workloads are hosted and what needs to move.</p><p>My recommendation is to treat inventory as an ongoing financial management capability, not just an IT asset management exercise.</p><p>Know what you own, where it lives, who is accountable, what it costs, and what it depends on.</p><p>Don’t wait until someone hands you a migration deadline to figure it out.</p><p><strong>3. Your cloud architecture can determine your separation costs years later.</strong></p><p>Some of our cloud architectural decisions had been made more than five years before the separation.</p><p>Multiple applications were deployed into shared Azure subscriptions.</p><p>That architecture served the organization for years. But when the business needed to separate, we couldn’t simply move an entire subscription to a new environment because the applications didn’t all belong to the same future company.</p><p>Instead, we faced significantly more involved migrations at the application and resource level.</p><p>More dependency mapping, resource recreation, configuration changes, testing, and coordination.</p><p>Would I recommend putting every application in its own subscription?</p><p>Not necessarily. There are legitimate security, operational, and platform considerations.</p><p>But I would absolutely encourage organizations to consider <strong>workload separability and portability as architectural design principles.</strong></p><p>Organizational structures change. Businesses acquire companies, sell divisions, and reorganize.</p><p>Design your cloud environments with enough flexibility to support change without requiring massive migrations every time the business evolves.</p><p><strong>4. The best contract today might become tomorrow’s biggest headache.</strong></p><p>Of everything we navigated, vendor contract complexity was probably my biggest surprise.</p><p>In the months leading up to the separation announcement, we had achieved some significant commercial restructuring wins.</p><p>We had negotiated agreements that made financial sense for the company as it existed at that time.</p><p>Then the business strategy changed.</p><p>One example involved a three year digital consumption wallet. At the anticipated separation date, we would be approximately halfway through the agreement.</p><p>The problem? The wallet couldn’t be divided between the two future companies.</p><p>We eventually worked through an arrangement that transferred the agreement to the larger consumer. But that meant the other company needed a new commercial arrangement, resulting in additional dissynergy costs.</p><p>The original financial decision wasn’t necessarily wrong.</p><p>But the experience reinforced something I now believe strongly:</p><p><strong>A negotiated discount is only one component of a good technology agreement. Flexibility has financial value, too.</strong></p><p>As the separation progressed, we became cautious about long term renewals.</p><p>Sometimes that meant sacrificing discounts.</p><p>And as a FinOps practitioner, willingly passing up savings isn’t exactly second nature!</p><p>But the newly formed companies would have new leaders, new strategies, and different priorities. Locking them into long term decisions before they could establish their own direction wasn’t necessarily the right business decision.</p><p>If I could go back, I would advocate for stronger cross functional review of major multiyear commitments when significant business changes are being considered.</p><p>Can the agreement be transferred? Can consumption commitments be divided? What happens to remaining obligations in a divestiture?</p><p>These questions deserve attention before contracts are signed.</p><p>And vendor relationships matter just as much.</p><p>Some partners understood the urgency and worked creatively with us. Others seemed more interested in using the separation as an opportunity to introduce new products or modernization plans.</p><p>There’s a time for modernization. But during a separation, the focus needs to be on establishing two functioning businesses.</p><p>Sometimes success is knowing when to negotiate, when to compromise, and when to move on.</p><p><strong>5. Shared applications create dependencies that no migration schedule can ignore.</strong></p><p>Not every application belongs neatly to one company.</p><p>Shared and corporate platforms often support multiple business segments, departments, and applications simultaneously.</p><p>Sometimes the foundational services must be duplicated and operational before dependent applications can even begin migrating.</p><p>We experienced this with platforms such as Power Platform and Microsoft Fabric. Foundational capacity and services needed to be established before applications relying on them could move.</p><p>And once those services were available, application teams still needed to know how to connect.</p><p>What are the new directory paths? Which URLs changed? What security groups are required?</p><p>That sounds simple until dozens of teams need the information and changes are happening rapidly.</p><p>Communication became just as important as technical execution.</p><p>Cloning shared applications also introduces some of the longest lasting separation costs. Duplicate infrastructure, licensing, storage, and support can continue well beyond Day 1.</p><p>There may also be years of shared data that need to be separated, cleansed, or removed.</p><p><strong>Don’t just ask when an application can move. Ask what must be ready before it can move, and what costs remain after it does.</strong></p><p><strong>6. Financial forecasting becomes an exercise in managing uncertainty.</strong></p><p>I’ve spent years building technology forecasts, and forecasting during a major company split is a different experience.</p><p>We reforecasted our cloud costs approximately ten times over twelve months.</p><p>Not because previous forecasts were poorly constructed, but because the assumptions kept changing.</p><p>Which applications were going to each company? Which needed to be cloned? Would workloads remain on premises or migrate to the cloud? When would they actually move?</p><p>At one point, our anticipated timeline accelerated from roughly 18 to 20 months down to nine months.</p><p>That changes everything.</p><p>Migration schedules influence how long duplicate environments operate, when infrastructure is provisioned, when services can be retired, and how much transitional support is required.</p><p>One thing I would do differently is invest more heavily in dynamic scenario planning.</p><p>I see an opportunity to use automation and AI assisted modeling to evaluate changing migration scenarios instead of repeatedly rebuilding forecasts from scratch.</p><p>But that only works if the underlying application inventory, financial data, and assumptions are reliable.</p><p>We also learned how important it was to have financial processes capable of keeping pace with the migration factory.</p><p>When a team needs infrastructure, licensing, or cloud capacity to complete a migration, the speed at which funding can be approved becomes a delivery dependency.</p><p><strong>The goal isn’t to create a perfect forecast in an environment that changes daily. It’s to make uncertainty visible and help leaders understand the financial consequences of their decisions.</strong></p><p><strong>7. Sometimes the right FinOps decision is to stop optimizing. <em>Temporarily</em>.</strong></p><p>This might have been the hardest lesson for my little FinOps heart.</p><p>During normal operations, we’re constantly looking for opportunities to eliminate waste, rightsize workloads, and improve commitment coverage.</p><p>But during a major separation, those priorities compete with something bigger.</p><p>Getting the business separated successfully.</p><p>Our application teams faced a genuine tradeoff. Optimize the environment they just migrated, or move on to the next application in an already aggressive schedule?</p><p>Sometimes the answer was to keep moving.</p><p>That meant some environments were overprovisioned. Some optimization activities were put on hold. And in certain cases, there simply wasn’t time to procure replacement hardware, making TSAs necessary.</p><p>Those decisions weren’t necessarily failures of cost management.</p><p>They were business tradeoffs.</p><p>I had to get comfortable with the idea that <strong>the lowest cost decision wasn’t always the best decision for the business.</strong></p><p>The trick was picking the right battles.</p><p>Would optimization delay a critical migration? Would moving too quickly create an outage? Could we improve efficiency without slowing another team down?</p><p>I think of the deferred work as <em>optimization debt.</em></p><p>Like technical debt, it needs to be visible, understood, and eventually addressed.</p><p>Otherwise, temporary separation costs quietly become the new baseline.</p><p><strong>8. FinOps can be the conductor, not just the accountant.</strong></p><p>One of the most rewarding parts of the experience was stepping beyond traditional FinOps responsibilities.</p><p>Our leadership team established an application migration factory, creating a repeatable process to assess, plan, migrate, and support applications.</p><p>Because of my experience with large scale migrations and existing cross functional relationships, I stepped into a role helping connect application teams with the technical groups they depended on.</p><p>Security. Cloud engineering. Database administrators. End user computing. Windows and Linux teams.</p><p>My focus was making sure those dependencies didn’t become barriers.</p><p>We improved foundational processes, built playbooks, increased automation, provided better visibility, and reduced request turnaround times.</p><p>Processes that previously took days could, in some cases, be completed in hours.</p><p>I also became an escalation point when teams encountered issues that threatened to slow migrations.</p><p>It reinforced something I’ve believed for a long time about FinOps professionals.</p><p>We often know a little bit about almost everything.</p><p>We know applications because we help explain their costs. We know platform teams because we partner on optimization. We know Procurement and Finance through budgets and contracts. And we understand the processes connecting all of them.</p><p>That gives FinOps a valuable perspective during major transformations.</p><p>We can do more than explain the cloud bill.</p><p>We can help connect people, surface dependencies, quantify tradeoffs, and keep decisions moving.</p><p><strong>FinOps can be the conductor that helps the orchestra play together.</strong></p><p><strong>9. Day 1 isn’t the finish line.</strong></p><p>There is enormous focus during a separation on achieving Day 1 readiness.</p><p>But from a FinOps and ITFM perspective, the financial work doesn’t end when the companies become independent.</p><p>Now comes the work of understanding the true ongoing cost of each company.</p><ul><li>Which applications are still running under TSAs?</li><li>Where are we carrying duplicate infrastructure?</li><li>Which workloads were overprovisioned to meet migration deadlines?</li><li>What commitments make sense for the new consumption patterns?</li><li>Have application ownership and resource tags been updated?</li><li>Can Finance and technology teams return to reliable budget versus actual reporting?</li></ul><p>For us, the next chapter includes working with our highest spending application teams to rightsize environments, evaluate commitments, improve financial allocation, and address technical debt.</p><p>It also means taking what we’ve learned from FinOps and applying it more broadly to ITFM.</p><p>On premises infrastructure, SaaS, licensing, hardware, and the associated operational costs deserve the same financial visibility and accountability we’ve worked to build in cloud.</p><p>But there is another important consideration.</p><p>After a transformation of this magnitude, <strong><em>teams need breathing room</em></strong>.</p><p>They need time to reduce toil, simplify processes, address the work that was intentionally deferred, and get back to innovating.</p><p>And from what I’ve seen, people are ready to innovate again.</p><p>The challenge for leadership is creating the space to address yesterday’s debt while enabling tomorrow’s opportunities.</p><p><strong>The Real Superpower Was How People Showed Up for Each Other</strong></p><p>I’ve shared a lot about architecture, contracts, tagging, forecasting, and the financial realities of separating a company.</p><p>But when I look back on this experience, those aren’t the things I’ll remember most.</p><p>I’ll remember the people.</p><p>This could have easily turned into a blame game. Teams protecting their own priorities, pointing fingers when something went wrong, or getting stuck in processes that weren’t designed for the pace we needed.</p><p>Instead, I watched people genuinely come together around a shared goal.</p><p>Some of my favorite moments were when one team solved a problem and immediately created a playbook so the next team wouldn’t have to solve it all over again.</p><p>When people shared information instead of holding onto it.</p><p>When teams shifted to around the clock support because that’s what was needed.</p><p>When we stepped away from endless meetings and tracking boards and focused on simply getting the next thing done.</p><p>It wasn’t always pretty.</p><p>It wasn’t always perfect.</p><p>But people showed up for each other.</p><p>And I genuinely believe that’s what made this possible.</p><p>We talk a lot about resilience in technology. Usually, we’re talking about redundancy, disaster recovery, architecture, and system availability.</p><p>But this experience reminded me that <strong>resilience is just as much about people, relationships, and processes as it is about technology.</strong></p><p><em>It’s the trust you’ve built with a vendor long before you need a favor.</em></p><p><em>It’s the relationships across IT that allow someone to pick up the phone and get an issue resolved.</em></p><p><em>It’s the willingness to document a solution so someone else can move faster.</em></p><p><em>It’s knowing when to let go of perfection and focus on what’s best for the business.</em></p><p><em>And it’s leaders creating an environment where people are empowered to solve problems together.</em></p><p>If I could leave other FinOps and IT leaders with three lessons, they would be simple:</p><ol><li><strong>Build for change.</strong> Your architecture, contracts, governance, and processes should be resilient enough to adapt when the business changes.</li><li><strong>Know what you own.</strong> Understand your applications, infrastructure, dependencies, ownership, and costs before someone asks you to separate them.</li><li><strong>Invest in relationships.</strong> The people and partnerships you develop during normal operations may become your greatest asset during extraordinary circumstances.</li></ol><p>FinOps absolutely has an important role in major business transformations. Our financial insights, cross functional relationships, and ability to connect technology decisions to business value can make a meaningful difference.</p><p>But the biggest lesson I walked away with wasn’t about cost optimization, forecasting, or even cloud architecture.</p><p>It was about what happens when people rally around a shared purpose.</p><p>Because at the end of the day, you can have the best financial models, the most detailed migration plans, and all the right technology.</p><p><strong>It’s the people who make it happen.</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=da6d24ea32bd" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Transitioning from EA to MCA and impacts on Azure:]]></title>
            <link>https://medium.com/@ashley.hromatko/transitioning-from-ea-to-mca-and-impacts-on-azure-0c0f25879fb1?source=rss-c8f57eb6d3ac------2</link>
            <guid isPermaLink="false">https://medium.com/p/0c0f25879fb1</guid>
            <category><![CDATA[finops]]></category>
            <category><![CDATA[macc]]></category>
            <category><![CDATA[mca]]></category>
            <category><![CDATA[ea]]></category>
            <category><![CDATA[azure]]></category>
            <dc:creator><![CDATA[Ashley Hromatko]]></dc:creator>
            <pubDate>Mon, 02 Feb 2026 18:22:28 GMT</pubDate>
            <atom:updated>2026-02-02T20:21:38.774Z</atom:updated>
            <content:encoded><![CDATA[<p>While many articles explain the basics of EA, MCA, or SCE agreements, this guide takes a deeper, practical approach. It focuses on <strong>clarifying the Azure billing hierarchy, highlighting the new features available to practitioners, outlining critical implementation considerations, and sharing real-world experience</strong> encountered when migrating from an Enterprise Agreement (EA) to a Microsoft Customer Agreement (MCA).</p><p><em>Please note this reflects my personal experience and perspective and does not represent the views of my employer.</em></p><h3>Understanding MCA Billing <strong>Hierarchy</strong> &amp; Terms</h3><p>It’s essential to grasp the fundamentals of the billing hierarchy introduced with MCA in Azure. Unlike the EA structure, which centered around enrollments, departments, and accounts, MCA organizes billing around <strong>billing accounts, billing profiles, invoice sections, and subscriptions</strong>. This new hierarchy not only determines how costs are aggregated and reported, but also defines who can access, manage, and analyze costs at each level.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kbnt7SmnzLQA8a4N8h8nGg.png" /></figure><p><strong>Billing Account <br></strong>A billing account as the <strong>top-level container for your commercial relationship with Microsoft</strong>. It represents the contract under which all Azure usage is billed (EA or MCA/MACC) and defines how invoicing and payment are handled across the organization.</p><ul><li>Account-level properties include agreement type, payment terms, currency, tax profile, billing contacts, and invoicing cadence.</li></ul><p><strong>Billing Scope<br></strong>A Billing scope as a <strong>lens or slice of cost visibility within the billing account</strong>. It defines <em>which portion of usage and costs you are viewing, managing, or exporting</em>.</p><ul><li>Billing Scope is how you will toggle between EA &amp; MCA in the Azure Cost Management + Billing Portal</li></ul><p><strong>Billing Profile </strong><br>A Billing Profile represents an <strong>invoice and its associated billing info</strong>. It’s where you manage invoices, payment methods, and PO numbers, and can split costs into <strong>Invoice Sections</strong>.</p><ul><li><strong>Policies</strong> for Azure Marketplace purchases and Reservations can be enforced here.</li><li>Ideal for <strong>siloed business units</strong> — for example, an R&amp;D team that has its own PO to track spend independently.</li></ul><p><strong>Invoice Sections </strong><br>Invoice Sections allow you to <strong>group costs within a Billing Profile</strong>, such as Marketplace purchases or costs tied to specific projects or departments.</p><ul><li>Subscriptions and purchases can be mapped to an Invoice Section, providing <strong>more granular reporting</strong>.</li></ul><p><em>While I like the idea of invoice sections, they don’t always work for organization because of landing zone constraints. If multiple business units can create resource groups within the same shared subscriptions, mapping those subscriptions across multiple invoice sections will not work because invoice sections are 1:1 with subscriptions.</em></p><p>Read more on Microsoft Customer Agreement administrative roles in Azure: <a href="https://learn.microsoft.com/en-us/azure/cost-management-billing/manage/understand-mca-roles?source=recommendations">https://learn.microsoft.com/en-us/azure/cost-management-billing/manage/understand-mca-roles?source=recommendations</a></p><h3>MCA Introduces New Features</h3><p>Beyond the contract itself, MCA brings a host of <strong>new features and capabilities</strong> designed to give teams more control, flexibility, and visibility over Azure costs.</p><p><strong>Multiple Invoices</strong></p><ul><li>New Billing Hierarchy allows for multiple invoices per billing account via billing profiles</li></ul><p><strong><em>New</em> Invoice “generate” Feature</strong></p><ul><li>A helpful new feature in the Azure portal allows you to <strong>reproduce invoices if you need to change the Bill to or PO #. </strong>I’ve oddly had to use this a few times now with a PO change or wanting to change the bill to address and it is nice to not have to submit a ticket. You are limited to doing this once per invoice.</li></ul><p><strong>Commitment Consumption Section</strong></p><ul><li>If you have a MACC agreement, MCA introduces a <strong>Commitment Consumption widget </strong>if you have a MACC agreement as well. Personally, this is “just okay” features. I’d still recommend creating your own commitment burn-down so you can track where you forecasted you would be vs where you are. Include pivoting data by usage vs Marketplace.</li></ul><p><strong>Multi-Tenancy Support</strong></p><ul><li>A single MCA can support <strong>multiple tenants</strong> associated with the primary billing account. Users on linked tenants can access MCA and Azure Cost Management data without guesting accounts into the primary tenant, streamlining access and reporting. This is useful for a Dev Tenant and if you have additional tenants via Mergers and Acquisitions.</li></ul><p><strong>API Update</strong></p><ul><li>Cost Management upgraded to Azure Resource Manager (ARM) APIs. Read more on changes : <a href="https://learn.microsoft.com/en-us/azure/cost-management-billing/costs/migrate-cost-management-api">https://learn.microsoft.com/en-us/azure/cost-management-billing/costs/migrate-cost-management-api</a></li></ul><h3>Tactical Lessons</h3><p>As a FinOps practitioner, I’ve often found myself scouring blogs and guides, searching for the “hidden gems” — the small but critical details that can make or break a migration. The next section dives into those tactical considerations, highlighting the things you don’t want to overlook when moving from EA to MCA, from setup nuances to operational steps that ensure a smoother transition.</p><p><strong>Banking and Remittance</strong></p><ul><li>MCA uses <strong>different payment routing</strong> than EA; remit info must be updated.</li><li>Your finance may request proof of banking, such as: Non-editable PDF on company letterhead with banking info and voided check or invoice copy. Microsoft was quick to provide this.</li></ul><p><strong>Tax Exemption Letters</strong></p><ul><li>Any existing EA tax exemption letters <strong>must be resubmitted</strong> under MCA. Plan to open a support ticket for this before you get your next bill.</li></ul><p><strong>Exports and Reporting</strong></p><ul><li>You will need to create new export schedules for the Cost and usage details exports in the MCA Billing Scope.</li><li>Microsoft cannot be appended EA → MCA exports together. Your billing process will need to be updated to handle both export types. Be prepared for new columns &amp; different values between the two export files.</li><li>This is where FOCUS has shined by providing consistent cost &amp; usage specification. As part of this transition, we prepared for this by switching our EA exports to FOCUS and back generating the past year.</li></ul><p><strong><em>Timing matters!!! </em></strong><em>A single month may contain both EA and MCA costs, so carefully plan and track these exports. If you are also using a MACC with the MCA for the first time, be aware timing of executing this transition could leave cost on the EA when your agreement for a period that is technically included in your MACC burn-down commitment period.</em></p><h3>Dashboards and Monitoring</h3><h3>Cost Management UI Experience — Misses the Mark in Key Areas</h3><p>The Cost Management experience under MCA introduces several challenges that teams should plan for in advance.</p><ul><li>First, <strong>any alerts or budget monitoring configured under EA do not carry over</strong>. These must be <strong>recreated manually</strong> in the MCA environment to maintain continuity in cost oversight and anomaly detection.</li><li>Similarly, <strong>custom views saved in Cost Analysis do not migrate</strong>. Any tailored reports or filtered views your teams rely on will need to be rebuilt after the transition.</li><li>The most painful UI limitation, however, is <strong>loss of historical visibility in the Cost Analysis view at cutover</strong>. Once you move to MCA, the default Cost Analysis experience shows <strong>only post-migration data</strong>. To view historical costs, you must manually switch the <strong>billing scope</strong> and navigate back to historical EA cost analysis reports.</li></ul><p>This user experience is especially disruptive when the transition occurs mid-year. Ideally, organizations should be able to view full-year trends within a single UI, but instead, this split forces teams to export two separate datasets (EA and MCA) and rely heavily on Excel-based analysis just to understand spend trends.</p><p>This is also where <strong>third-party FinOps tools can provide significant value</strong>. External platforms often offer smoother historical continuity and unified reporting across EA and MCA. <em>That said, this is a critical reminder to notify your FinOps vendor ahead of time, understand how they handle EA-to-MCA transitions, and assess the impact on dashboards, reports, and trend analysis before cutover.</em></p><h3>Azure Calculator</h3><ul><li>Currently, <strong>discount visibility</strong> in the Azure Calculator depends on account access:</li><li>EA discounts are visible only to EA Administrators or delegated Department/Account Owners.</li><li>MCA discounts are visible only to users with subscription access (Owner/Contributor/Reader).</li></ul><p><strong><em>Note to Microsoft:</em></strong><em> Consider a specialized role for product owners or engineers who regularly perform cost estimates in the calculator but don’t need full MCA subscription access. This would improve forecasting and planning without over-provisioning access.</em></p><h3>Real-World Story: Timing Challenges</h3><p>In 2025, we became early adopters of the EA → MCA transition — and like all early adopters, we learned some lessons the hard way. This is the story of how <em>timing </em>determined where real dollars ultimately landed.</p><p>We meticulously planned our transition, locking in a date and explicitly requesting execution as close to midnight on the 30th as possible. The assumption was simple: Microsoft would automatically complete the EA → MCA cutover at the end of the billing period (Saturday night). But when I logged in Monday morning, the transition hadn’t occurred.</p><p>What we learned next mattered. There were customer-side steps that had to be completed by FinOps — steps we didn’t realize were gating the cutover. That gap created nearly two full days of spend that never hit MACC, including high-value Reserved Instance and Savings Plan renewals that trigger on the first of the month.</p><p>We ultimately partnered with Microsoft to resolve the issue through a manual adjustment workaround — but the experience surfaced a clear gap.</p><p>My recommendation to Microsoft is straightforward: this process needs automation. Customers should be able to complete all prerequisites in advance and schedule a guaranteed midnight transition to eliminate partial-month exposure and financial ambiguity.</p><h3>Real-World Story: Reservations &amp; Savings Plans — Holding Your Breath</h3><p>Early in our EA → MCA journey, I came across a <a href="https://www.linkedin.com/pulse/moving-from-ea-mca-what-you-should-know-dina-fatkulbayanova-jqbye/?trackingId=jyX00oslW%2BdiEcCT9kbEUg%3D%3D">blog</a> that included the following warning:</p><blockquote>“You should also check with your Microsoft account team on commitment transfer details (reservations and savings plans). Make sure you’ve acknowledged that if reservations or savings plans need to be cancelled and repurchased during the transition, a new end term will be added — as if it were a brand-new purchase.”</blockquote><p>Every FinOps practitioner with a long-standing love-hate relationship with rate commitments probably had the same visceral reaction reading that: <em>“NOOOO… please tell me these commitments transition flawlessly.”</em></p><p>We may have gotten lucky, as our Reservations and Savings Plans moved cleanly from the EA billing scope to the MCA billing scope with no disruption.</p><p>But I’ve since read multiple customer stories where that wasn’t the case, and the financial operational preparation is a heavy life. According to Brad Payne’s <a href="https://www.linkedin.com/pulse/hours-matter-timing-precise-brad-payne-euj1c/?trackingId=OiNbu3VIQdGq%2BNXrTCSVdw%3D%3D">blog</a> on timing, “<em>Reserved Instances may transfer to the new billing profile and be refunded on the next billing cycle (but cannot be exchanged), while Savings Plans may not transfer at all and instead be refunded</em>.” Those details matter — especially when you’re managing many individual long-term commitments.</p><p>My strong recommendation: make this a <strong>deliberate, explicit conversation</strong> with your Microsoft account team before you transition.</p><h3>Getting Started and Recommendations</h3><p><strong>1. Start with Microsoft Documentation</strong><br>Begin by grounding your team in the official guidance. A solid starting point is Microsoft’s Tech Community guide: <em>Moving to MCA from EA</em>. Understanding the recommended process up front can save time and prevent costly missteps.</p><p><strong>2. Engage Your Account Managers Early</strong><br>Coordinate with Microsoft as early as possible to validate expectations. Your account managers will connect you to the Azure Commerce Engineering Operations team, who can provide hands-on support for setup, guidance, and post-transition troubleshooting.</p><p><strong>3. Plan Billing Scopes and Invoice Sections Carefully</strong><br>Design your billing hierarchy thoughtfully to avoid conflicts with your landing zone architecture. Consider how subscriptions, invoice sections, and profiles will map to business units, cost centers, and reporting requirements.</p><p><strong>4. Ensure Operational Readiness Across Teams</strong><br> A successful transition touches multiple functions:</p><ul><li><strong>Procurement</strong>: Sign the new MCA contract.</li><li><strong>Finance</strong>: Create new purchase orders (in needed) and update remit information.</li><li><strong>Cloud Engineering</strong>: Create and configure subscriptions, one is required for the transition.</li><li><strong>Engineering Teams</strong>: Communicate portal navigation to users.</li><li><strong>FinOps</strong>: Be ready to merge and normalize exports to maintain consistent cost and usage reporting.</li></ul><h3>Conclusion</h3><p>Migrating from EA → MCA is more than a <strong>billing pointer change</strong>; it affects <strong>processes, roles, dashboards, exports, and cost visibility</strong>.</p><p>With thoughtful planning, clear communication, and the right tooling in place, organizations can transition to MCA smoothly while maintaining strong financial governance. That said, it’s important to recognize that a successful transition does require deliberate FinOps investment to ensure reporting continuity, operational stability, and sustained cost transparency.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0c0f25879fb1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Lessons from Allocating Azure OpenAI Costs at Corteva]]></title>
            <link>https://medium.com/@ashley.hromatko/lessons-from-allocating-azure-openai-costs-at-corteva-7beb806e3eca?source=rss-c8f57eb6d3ac------2</link>
            <guid isPermaLink="false">https://medium.com/p/7beb806e3eca</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[openai]]></category>
            <category><![CDATA[cloud-computing]]></category>
            <category><![CDATA[finops]]></category>
            <category><![CDATA[azure]]></category>
            <dc:creator><![CDATA[Ashley Hromatko]]></dc:creator>
            <pubDate>Tue, 20 May 2025 17:10:36 GMT</pubDate>
            <atom:updated>2025-05-20T17:10:36.726Z</atom:updated>
            <content:encoded><![CDATA[<p>Imagine a sales rep standing on the edge of a cornfield, chatting with a farmer about this season’s seed options. The farmer brings up a new product they heard about that is not yet widely adopted yet and asks the question….</p><p><strong>“What’s the drought tolerance on the new hybrid XYZ2025 product?”</strong></p><p>In the past, the sales rep would need to jot down the question, do some research, dig through product sheets, maybe play some phone tag… before finally circling back days with the answer to the farmer.</p><p>Imagine how AI will bring business value to our customers when that same rep can pull out a phone, open an AI-powered chat tool, and type the question in real time.</p><p><strong>So, the question on every business executive’s mind right now is: Are we truly getting value from AI, and what’s the cost trade-off?</strong></p><p>The improved customer experience is a clear win — but we also need to understand the infrastructure cost behind it. Did 1 helpful response cost us <strong>$3</strong> or .<strong>60 cents</strong>? The answer matters when you’re scaling AI across a global enterprise.</p><p>As an FinOps leader, I’ve spent the past few months diving into <strong>Azure OpenAI cost allocation</strong>, exploring everything from infrastructure configuration, process governance to token-based showbacks. This post walks through what I learned, what surprised me, and what I’d tell any FinOps practitioner looking to get ahead of the AI cost allocation curve.</p><p><strong>🧭 Why Azure OpenAI, and Why Now?</strong></p><p>Our OpenAI spend is still relatively small compared to other Azure services, which makes it the perfect time to understand it deeply and get guardrails in place <strong>before</strong> we scale. My initial goal was simple: take our largest OpenAI Resource Group Costs — our dev workspace — and figure out:</p><ul><li>Who’s using it?</li><li>How are they using it?</li><li>What trends can we explain?</li><li>And most importantly — <strong><em>How do we tell a value story to leadership?</em></strong></li></ul><p><strong>🛠️ AI Infrastructure: Think “Shared Platform”</strong></p><p>We’ve opted for a <strong>shared OpenAI platform</strong> rather than giving teams dedicated workspaces. Think of it like a shared managed database — with different apps using separate schema or in the case of Open AI workspaces.</p><p>Key components include:</p><ul><li>Azure Resource Groups</li><li>Azure OpenAI workspaces</li><li>Azure API Management</li><li>Coming soon…. Redis*</li></ul><p>*Redis can be used to cache API responses, which helps in reducing latency and improving the performance of API calls. For example, when an API request is made, Redis can check if a semantically similar question has been asked before. If it finds a match in the cache, it returns the cached response instead of querying OpenAI again (Saving $$).</p><p><strong>🧩 Applying FinOps Principles to OpenAI</strong></p><p>We applied several <strong>FinOps capabilities</strong>:</p><ul><li><strong>Data ingestion</strong>: Pulling logs and billing data</li><li><strong>Benchmarking</strong>: analyzing usage patterns</li><li><strong>Unit economics</strong>: Determining cost per key</li><li><strong>Showback</strong>: Token usage reports for end-user visibility</li><li><strong>Onboarding</strong>: View our process through the lens of how to we automate and make it more secure.</li></ul><p><strong>🚧 Initial Gaps found</strong></p><ul><li><strong>Manual API key management</strong>: No trackable intake process or capturing key business data used for reporting</li><li><strong>No standard naming</strong>: API key display names were inconsistent (even duplicates!)</li><li><strong>Lack of cataloging API key purpose: </strong>I’ll explain this more below</li></ul><p>Another hurdle: <strong>API keys don’t support Azure tagging</strong>, (Side note: Microsoft we ALWAYS need tags on services). We had to maintain a reference table to track key ownership, business units, and purpose.</p><p><strong>🔄 Token Allocation Model</strong></p><p>Here’s how we tackled allocation in a shared environment:</p><ol><li>Count successful token requests per API key</li><li>Calculate the percentage of successful token requests for each key by dividing its success token requests by the total successful token requests for the month.</li><li>Times the <strong>% </strong>per key by the workspace’s monthly spend</li></ol><p>This gave us an estimated monthly $ cost per API key — enough to get started on meaningful showback.</p><p><strong>🔄 Asking the Questions to Understand Business Value</strong></p><p>Month-over-month we will need to repeat this formula as both consumption and total workspace spend will change.</p><p>We needed to dive further into the purpose of these keys to know who or what are these keys associated with?</p><p>As the FinOps lead, I personally reached out to all API owners and asked them these questions and started capturing this as business context per key.</p><p><strong>Are these a user key (personal playground) </strong>or <strong>Application Key </strong>(<strong>associated with an</strong> <strong>application)?</strong></p><p>If user key</p><ul><li>Who is the user</li><li>What business org do they report to</li><li>What are the using the key for? POC, Training, or a sandbox for a figure Application</li><li>Do they know to request a new key per Application</li><li>Do they know how much they are consuming</li></ul><p>If application key</p><ul><li>What’s the name of the application</li><li>Where is the application in the life cycle — preprod or operational</li><li>Is this an internal or external application</li><li>Is the workspace in the correct region needed</li><li>What is expected growth/ seasonality</li></ul><p><strong>📊 Data ingestion &amp; Visualizing</strong></p><p>We are still in the <strong>crawl phase</strong> of gathering data manually. For these initial analysts we used data from these three sources.</p><ul><li>Azure portal &gt; Monitoring Analytics &gt; Subscriptions &amp; User tabs</li><li>Third-party FinOps platform for monthly Resource Group spend.</li><li>Excel for holding Business Metadata associated with each unique API Key</li></ul><p>For reporting, we focused on the following four dashboards. <br><strong><em>Note:</em></strong><em> All data presented here has been anonymized and numerical values are for illustrative purposes only.</em></p><p><strong>Month-over-month usage</strong> by API key (individual vs app)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/721/1*sEOJ1dinSggUPsYKeLyvTQ.png" /></figure><p><strong>Org &amp; function mapping</strong> of individual users</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/570/1*jluXhA0_nKY7IhvrC3ob0g.png" /></figure><p><strong>App-level costs</strong> by name</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/975/1*f242RhvqVN0I0w0A5yTkHw.png" /></figure><p><strong>Scenario planning — </strong>What would be the optimized cost if we had a Redis cache handling frequently asked prompts.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FxyaeG-jYGD4XL3-qY8ysg.png" /></figure><p><strong>📊 Communicating the Story</strong></p><p>When we shared this data with AI Platform lead &amp; IT leadership, it sparked conversations around.</p><ul><li>Standing up an AI CoE</li><li>Adding onboarding requirement that any new AI applications should go through a pre- deployment EA review</li><li>Any existing application, we put a $ threshold to trigger a post deployment EA review</li><li>Automating this data to Power BI</li><li>Backlogged process improvement tasks (including areas to automate)</li></ul><p><strong>⚠️ Challenges and Surprises</strong></p><p>A few things caught me off guard:</p><ul><li><strong>The speed of change</strong>: Mid-project, Microsoft released changes to reservation for OpenAI. We did a quick ROI and decided to stay on PAYGO for Dev.</li><li><strong>Azure Portal Reports Needs Improvements</strong>: Unable to create an csv file from monitoring analysts&#39; section in the portal and no ability to tag keys.</li><li><strong>Forecasting without context</strong>: Without historical baselines, we were guessing at future forecasts. Now we know at least the users &amp; applications driving spend and can consult them for our next budget cycle.</li><li><strong>Lack of playbooks</strong>: FinOps practitioners don’t have much guidance yet for OpenAI-specific cost models.</li></ul><p><strong>📚 Lessons and Advice for Other FinOps Teams</strong></p><p>Applying familiar FinOps patterns to a bleeding-edge tech like OpenAI reminded me: it’s still just another unit cost. It’s all about understanding consumption and aligning it with value.</p><p>My advice for others:</p><ul><li><strong>Start small</strong>, while spend is still manageable</li><li><strong>Establish metadata standards early</strong></li><li><strong>Get curious about the value side</strong>: Who’s using AI and why?</li></ul><p>And if you’re starting your own FinOps journey with Azure OpenAI feel free to reach out with questions!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7beb806e3eca" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>