Accelerating hybrid cloud with generative AI

Insights

• In practice, hybrid cloud environments are often patchy and incoherent.
• Transforming a patching hybrid cloud is time-consuming and difficult due to unclear strategy, lack of skills, and organizational complexity.
• Generative AI can deliver solutions to these persistent problems by quickly analyzing data, rapidly drafting potential designs, automating testing, and optimizing migration plans.
• Proper use of generative AI in hybrid cloud transformation depends on good data, the right AI platform, strategic model selection, and cross-functional governance decisions.
• Guardrails for AI usage should account for input quality, data sensitivity, model bias, and proper systems integration.

Hybrid cloud environments provide businesses an agile IT infrastructure to support innovation and accelerate time to market. Or at least, that’s the theory. In practice, hybrid cloud deployments are often patchy and lack strategic coherence. This holds back the ability to provide a flexible, secure, reliable, and scalable technology foundation.

But moving to a strategic and coherent hybrid cloud landscape can be incredibly challenging and time-consuming. Hybrid cloud transformations often take 18 to 24 months or beyond, far longer than anticipated.

These delays can be due to a lack of strategic clarity, challenges in understanding the existing landscape, and the requirement for more complex security controls. A common roadblock documentation and single source of truth on ownership and dependencies hold strategic hybrid cloud deployments back.

But skills paucity is also a significant challenge. According to IDC, 87% of IT leaders in North America reported digital transformation delays due to a lack of sufficient IT skills.

All these issues add up to lost time — especially in large, complex organizations. The discovery phase alone can take two to three months, involving data collection, assessing the existing landscape, and defining target-state needs.

The design phase may take six to eight months, which involves defining the target-state architecture, validating solution components on meeting service-level agreements (SLAs), making architectural changes to meet those SLAs, developing low-level designs, and validating and coordinating with original equipment manufacturers (OEMs) to align the overall design with both OEM and customer standards.

Once discovery and design are complete, the migration phase can take nine to 15 months. It involves determining right cloud placement for workloads, targeting state application deployment design, planning workload migration schedule, considering dependencies, and addressing volume and velocity considerations. Finally, operations take over the transformed platform.

In total, these phases can delay a large business’s ability to implement a strategic hybrid cloud strategy by one to three years.

Generative AI’s role in hybrid cloud

Generative AI is poised to revolutionize hybrid cloud transformations, offering solutions to long-standing challenges. Purpose-built AI agents can streamline each phase of the process.

In the discovery phase, AI can analyze vast datasets to identify existing infrastructure gaps and generate comprehensive reports with actionable insights, significantly reducing the time required for manual analysis. During the architecture and design phases, generative AI can rapidly generate high-level and low-level designs, considering enterprise architecture principles as well as specific environment constraints and technical standards. Generative AI agents can further accelerate the build and validation phase by automating configuration generation and verification, test case preparation, and test result analysis.

Finally, AI can optimize the migration process by generating migration plans that account for all dependencies, automatically creating runbooks, and providing real-time schedule tracking and reporting.

Generative AI solutions in the market are fast evolving. For public cloud-focused enterprises, consider Azure OpenAI, AWS Bedrock hosted models/SageMaker, or Google Gemini. Private cloud adopters can explore Meta Llama or Mistral large language models (LLMs), complemented by orchestration and agentic frameworks like LangChain, LangGraph, or AutoGen, on a GPU/CPU infrastructure.

Extend the existing Cloud Center of Excellence (CCoE) to focus on generative AI solutions, or establish a dedicated AI CCoE, depending on the organization’s maturity, According to AWS, CCoE needs to analyze how generative AI solutions can integrate into existing cloud architectures and migration roadmaps. This includes identifying dependencies between generative AI and target cloud platforms, as well as policy and compliance factors. With rapid cadences of updates to generative AI services, the CCoE must stay on top of product changes and adjust migration plans accordingly.

These AI-driven advancements promise to significantly accelerate hybrid cloud transformations and improve the overall efficiency of the process.

Key strategies for successful AI adoption

Generative AI can dramatically accelerate hybrid cloud transformations, but its power must be harnessed strategically. Enterprises — regardless of their target cloud architecture — should prioritize four key strategies.

1. Data strategy: AI's performance is entirely dependent on the data it receives. A well-defined data strategy should identify and collect relevant internal data (e.g., security policies, infrastructure standards); source valuable external data (e.g., best practices, templates) from service providers and partners; and establish data governance processes to ensure data quality and accuracy.
2. AI platform strategy: The choice between a private or public AI platform requires careful analysis. Enterprises must assess their security requirements and data sensitivity, evaluate the cost-effectiveness of each option, and determine the level of agility needed.
3. Model continuum strategy: A successful model strategy should define clear criteria for model selection, implement robust training methodologies, establishing a continuous retraining process to adapt to changes, and implementing rigorous validation procedures to ensure model accuracy and reliability.
4. Governance strategy: CCoE should have a cross-functional team comprising the head of cloud transformation, AI strategist, data scientist, compliance and ethics officer, and transformation functional lead. It should define a clear charter and measurable goals such as migration velocity and talent productivity.

Manage risks in AI adoption

Generative AI offers significant potential for accelerating hybrid cloud transformation. But it also brings inherent risks that can affect planning and execution. Following are some key risks and how to deal with them:

1. Input quality and human oversight: Generative AI outputs are only as good as the inputs into it — and these systems do not always interpret these inputs effectively. Generative AI can have difficulty interpreting complex technical documentation and diagrams, which are often critical for understanding the existing architecture and its nuances. It may also struggle to understand a business’s budget and time constraints. It’s important to ensure human review and validation of all AI-generated outputs. Generative AI must be provided with comprehensive, accurate input data, including detailed documentation and contextual information. Compare AI-generated outputs with existing documentation and expert knowledge, and create a process that allows for corrections, with feedback reintegrated into the AI system.
2. Data sensitivity and privacy: Providing generative AI with sensitive information about an organization's infrastructure, applications, and data can raise privacy and security concerns — potentially becoming a roadblock to AI adoption. The data leveraged for training or stored in the AI could potentially be exposed or misused. First and foremost, anonymize or mask sensitive data before providing it to the AI. Further, strengthen AI platforms with appropriate data encryption and access controls, in line with established enterprise policies.
3. Bias and assumptions: Generative AI models can inherit biases from their training data, which may lead to skewed or inaccurate outputs. This could happen due to incorrect assumptions about the organization's environment or requirements. Train AI models on diverse and representative datasets throughout the hybrid cloud transformation journey. Test the AI's outputs against a variety of use cases and scenarios and incorporate human expertise to validate and challenge the AI's assumptions.
4. Integration and compatibility: Generative AI tools may not seamlessly integrate with existing enterprise systems or workflows. It could be due to different data formats, application programming interfaces (APIs). Select AI tools that are compatible with existing systems and workflows, use standardized APIs and data formats, and conduct thorough testing and validation of AI tool integrations.

Enterprises involved in medium to larger hybrid cloud transformation programs should carefully look into generative AI capabilities and their potential to reduce transformation timelines and optimize IT budgets, enabling investments in new initiatives. Consider human in the loop for all generative AI outcomes — with humans verifying correctness and bringing further innovation. Hybrid cloud provides the foundation for enterprise digital transformation, but applications/workloads should be modernized to fully reap its benefits.