How Philip Morris International’s Outcome‑Driven EA Practice Won The 2025 Forrester EA Award In EMEA
For CIOs and architecture leaders, PMI’s example is less about a successful AI rollout than about the operating model that made scaling possible. The management decision is to treat enterprise architecture as a governance and value-creation function, not a late-stage review gate. That shifts who owns risk, reuse and prioritisation: legal, compliance, security and strategy are part of the design process, not downstream approvers. The trade-off is speed versus control, but PMI’s approach suggests that early guardrails can reduce later rework and avoid fragmented local AI practices that are hard to govern.
The AI Factory model also changes the portfolio conversation. Reusable components, shared standards and a catalog of approved building blocks create measurable leverage, but only if leaders are disciplined about where bespoke development is still justified. The practical question for governance boards is not “Can we do this with AI?” but “Should this use case consume enterprise platforms, shared components or a one-off build?” That requires explicit criteria for business value, risk, and reuse, plus metrics that track time-to-value, avoided duplication and technical debt avoided—not just pilots launched.
Adoption is the third management issue. PMI appears to have paired access to tools with repositories, communities and workflow integration, which matters because broad licences alone do not change work. For transformation leaders, the next questions are operational: Who curates approved use cases? How are exceptions handled? What evidence is needed before a component enters the enterprise catalog? And how will leaders know whether AI is improving decision quality, employee productivity and customer experience rather than simply increasing experimentation?
Philip Morris International (PMI) launched an AI Factory, deployed generative AI tools to 35,000 employees, and built more than 450 AI use cases across the enterprise. That impressive track record is one of the reason why PMI won Forrester’s 2025 Enterprise Architecture Award in the EMEA region.
PMI’s Enterprise Architecture And Technology Transformation (EA&TT) team redesigned its operating model to make AI scalable, governed, and reusable from day one. This is Forrester’s Outcome-Driven Architecture (ODA) Model in practice: architecture that proves its value through risk reduction, speed, reuse, experience, and business value.
AI Governance First, Scale Second
AI engagements with our client typically start with AI governance. In most cases, shadow AI experiments multiply, risk controls vary by team, and no one owns the enterprisewide picture. To avoid this potential crisis, PMI’s EA&TT team worked alongside legal, compliance, security, and strategy leaders to codevelop an AI governance framework with embedded guardrails and defined risk management processes. They strengthened PMI’s Trustworthy AI Framework, introduced in April 2025, to enforce transparency and continuous monitoring for every AI system in production. This groundwork enabled PMI to scale AI adoption without regulatory surprises and brand trust incidents.
The AI Factory: Composability At Enterprise Scale
The next challenge we see across our client base is industrializing AI beyond initial pilots. Teams build one-off solutions, duplicate effort, and never achieve compounding returns. PMI’s EA&TT team cofounded the AI Factory to solve exactly this problem. This AI Factory produces reusable, compliant AI components that development teams assemble rather than rebuild from scratch. PMI reports 40% faster time to value and 30% less development effort compared to bespoke builds. The EA&TT team ensured that every component met enterprise security and compliance standards before it entered the catalog. This is ODA’s composability principle at work: turning one-off innovation into enterprise leverage that compounds over time.
Thirty-Five Thousand Employees With AI In Their Hands
Adoption is where many AI programs lose momentum. Tools get deployed, but workflows don’t change. PMI’s EA&TT team took a different approach, embedding AI directly into day-to-day work across demand forecasting, scientific research, and personalized consumer engagement.
More than 35,000 employees now have access to generative AI tools including Microsoft Copilot, and 65% of office-based staff use them regularly. The EA&TT team built transparent repositories and community mechanisms that let teams discover, share, and adopt proven solutions across the organization. The results show up in faster delivery cycles, better employee experience, and more adaptive customer engagement across every business unit.
What CIOs Should Take From PMI’s Playbook
We advise CIOs to think of enterprise architecture as a strategic capability, not a review function. PMI’s EA&TT team operated exactly this way: cofounding the AI Factory, codesigning governance with the C-suite, and embedding reusable architecture into every AI initiative.
Forrester’s ODA Model defines six attributes of high-performing EA: valuable, accountable, influential, collaborative, agile, and innovative. PMI demonstrated all six as observable operating behaviors. It reduced technical debt, accelerated modularization, governed AI responsibly, trained a global architecture community, and embedded architecture into strategic value streams.
Want The Full Story?
This blog highlights only a fraction of PMI’s architecture transformation. To dive deeper into the data, examples and operating model detail behind PMI’s award‑winning practice, read the full case study.
Enjoyed this article? Sign up for our newsletter to receive regular insights and stay connected.