GenAI Won't Fix A Broken Cloud

Forbes

18-06-2025

Hrushikesh Deshmukh, Senior Consultant, Fannie Mae.
Generative AI (GenAI) has rapidly become the backbone of current digital transformation. This is driving chatbots, automating code and providing predictive insights at scale.
If your cloud infrastructure is fragmented, your teams are disjointed or your workflows are patchwork-based, AI will only amplify the dysfunction that exists. As AI is a driver of advancement, organizations first need to address the cultural and architectural flaws that constrain them. Absent this foundation, AI is noise rather than value.
Most corporate executives see GenAI as a magic key to increased productivity, creativity and efficiency in operations. There is a common assumption that having AI deployed will yield revolutionary results. But without the right foundations, AI solutions often end up being superficial and unreliable:
• Cloud infrastructures are too complex and insufficiently documented to integrate and scale easily.
• DevSecOps pipelines are unreliable and lack good observability, which lowers reliability and response time.
• Information is still siloed, replicated and stale, compromising AI efficiency and precision.
• FinOps processes are reactive instead of proactive, generating runaway expenses and inefficiencies.
To truly unlock GenAI's potential, organizations must first address these foundational weaknesses.
GenAI can be revolutionary, but it is not a cure-all solution on its own. It is an additional layer on top of all the things that architecture already accomplishes. Only 15% of organizations that deployed AI at scale experienced sustained gains in their operating models, one of the reasons being poor architectural readiness.
Businesses need to concentrate on the following architectural pillars:
• Modular, API-Driven Microservices: These provide flexibility and interoperability and make it easy for AI components to integrate into current systems.
• Event-Driven, Scalable Infrastructure: AI workloads can be unpredictable, spiking without warning.
• Unified Observability And AIOps: Unified observability solutions plus AIOps provide real-time monitoring, speed of issue identification and root-cause determination.
• Secure Data Governance And Policy-As-Code Enforcement: Data is the main source for GenAI, and its compliance, quality and accessibility all impact model performance.
Cloud architecture needs to be viewed as a living, dynamic system—not a fixed deployment. It must be optimized continually to accommodate shifting workloads, governance requirements and new AI tools emerging.
While cloud architecture is the body that enables AI, organizational culture is the nervous system that decides how well it works. More than 70% of AI projects fail, most often due to cultural and organizational factors rather than technological issues.
Businesses need to go through a cultural shift. This entails changing both mindset and operational behavior:
• From Ticket-Based Support To Self-Service And Platform Thinking: Enabling teams to resolve issues on their own and get access to tools without waiting enables faster innovation and less friction.
• From Siloed Responsibilities To Cross-Functional Accountability: AI projects need to involve IT, data science, business operations and compliance teams collaborating around common goals.
• From "Build Fast, Fix Later" To "Design Smart, Scale Responsibly": Careful planning is required with AI. Short-term thinking can result in badly tested models, breaches and scalability problems.
Culture is key to how GenAI is adopted, regulated and scaled. Organizations that value collaboration, exploration and responsibility will be better positioned to develop AI practices that are both successful and sustainable.
To provide true value with GenAI, companies can start with a serious evaluation of their cloud infrastructure. AI can flourish only when developed on agile, transparent and scalable infrastructure. Enterprises should pose important questions before implementing GenAI:
• Is the cloud architecture optimized for speed, cost management and visibility?
• Are CI/CD and MLOps pipelines dynamic enough to accommodate changing AI workloads?
• Can teams make AI insights operational fast or are they slowed down by manual work?
• Have strong policies for the ethical, explainable and responsible use of AI been put in place?
Cloud transformation is not a choice; it is a necessity. Only by restructuring their infrastructure, optimizing operations and building in responsible practices can companies realize the full, sustainable potential of AI.
While cultural and organizational factors are a common reason AI projects fail, data-related issues are another major obstacle to success.
GenAI systems are only as reliable as the data they're built on. To ensure trustworthy and accurate outputs, organizations must take a disciplined approach to data management, including:
• Ensuring data consistency and integrity across sources and systems
• Implementing metadata control and maintaining data lineage for traceability
• Centralizing data catalogs to support discoverability and reuse
• Enforcing access controls and policy-based governance to protect sensitive information
• Regularly auditing data for freshness, accuracy and relevance
To unlock real value from GenAI, businesses must treat data as a managed, strategic asset—not an afterthought.
If your systems are fragmented and your culture lacks cohesion, AI will only accelerate the dysfunction. Organizations must resist the urge to chase AI blindly without first addressing foundational flaws. No level of intelligence, however advanced, can compensate for poor design or a misaligned vision. Build it right first and then let AI amplify your success from there.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?