By Imtranur Rahman, Senior Solutions Architect — AWS
Jaya Padma Mutta, Manager, Solutions Architects – AWS
Channa Basavaraja, Senior Solutions Architect — AWS
Manju Arakere, Senior Solutions Architect — AWS

Generative AI (Gen AI) is a game-changer, augmenting human creativity, driving innovation, and boosting productivity. It offers a competitive edge across industries, necessitating new strategies, structures, and talent management for future success. This disruptive technology has captivated the world with its potential to transform industries across the board. Goldman Sachs estimates it could increase global GDP by $7 trillion and productivity growth by 1.5% over a decade.

Organizations are eager to harness gen AI for real-world innovations and productivity gains. However, they face challenges like resource constraints, complexity, and lack of execution and governance models. As a result, they’re increasingly turning to AWS Partners to bridge the gap between gen AI’s potential and its practical implementation, unlocking its full value.

AWS recognizes that having a robust gen AI practice is crucial for our partners to stand out, leverage AWS’s suites of services effectively, and deliver an exceptional customer experience. This blog post guides our Software and Services Path partners in developing a successful gen AI strategy.

As you implement gen AI offerings, take a responsible approach that balances its potential with safe, secure, transparent development. At AWS, we define responsible AI as being made up of eight key dimensions: fairness, explainability, veracity and robustness, privacy and security, safety, controllability, governance, and transparency. Partners should adopt responsible AI practices to maximize gen AI’s benefits for society, while actively identifying and mitigating potential risks and negative consequences.

Here is the suggested workflow with 5-phases our partners can adapt to build and activate a robust gen AI practice.

Figure 1. Phases for establishing Gen AI Practice

Executive Sponsorship

Successful gen AI practice implementation requires executive sponsorship aligned with business goals and vision. AWS Partners work closely with stakeholders to tailor gen AI solutions to their unique challenges and strategic objectives. An executive sponsor is vital for effective gen AI adoption as they champion the vision, address concerns, implement risk mitigation strategies, and drive continuous improvement, to keep the organization agile. They also ensure strategic alignment, ethical governance, cross-functional collaboration, and resource allocation.

Following strategic alignment, organizations should conduct a SWOT analysis to map their core capabilities and identify their fit within the AWS gen AI stack, as shown in Figure 2. This analysis helps organizations assess the impact of new gen AI practice offerings on operations and customers, enabling informed decisions to maximize benefits and mitigate risks.

Figure 2. Generative AI stack

Phases for establishing Gen AI Practice

Phase 1 — Technical Activation

Partners can build a successful gen AI practice by investing in employee training, streamlining processes with AWS guidance, and hands-on enablement with AWS technology. This three-pronged approach – focusing on people, processes, and technology – allows partners to showcase gen AI’s practical applications to customers through real-world examples, driving innovation and transformation. This approach also helps identify gaps, such as one for technical skills, allowing to create a roadmap to upskill people through training.

People: Successful gen AI adoption demands strong leadership, clear organizational vision, and inclusive employee involvement. Executives must communicate gen AI’s role as an augmentative tool, not a replacement, enhancing experiences for all stakeholders. Democratizing AI through persona-based training from AWS Skillbuilder, hands-on workshops, regular partner enablement sessions, and change management enables effective human-AI collaboration, streamlining processes, automating repetitive tasks, and fostering strategic innovation.

Process: AWS offers comprehensive gen AI resources and programs to support partners in effectively leveraging gen AI to improve customer experiences and achieve business objectives. AWS Cloud Adoption Framework for AI (CAF-AI) serves as a guide for a partner’s AI strategy and adoption journey. Additionally, AWS provides the Generative AI Center of Excellence and Partner Transformation Program, Generative AI Impact initiative as well as other funding mechanisms to accelerate partners’ gen AI expertise.

Technology: Partners must adopt a comprehensive strategy, integrating AWS AI/ML services like Amazon Bedrock, Amazon SageMaker, Amazon Comprehend, and Amazon Transcribe for natural language processing and generation. Utilize AWS Serverless offerings for models deployments and managements, use AWS Trainium and AWS Inferentia for training and inference, implement robust data foundation for large-scale data storage, transformation, security, and governance. Finally, they can experiment rapidly using custom solution from AWS solution libraries.

Phase 2 — Solutions and offering Development

This phase is pivotal for AWS partners in establishing their gen AI practice. Using AWS services, partners can craft innovative accelerators tailored to client needs. Partners can consult the three main use case categories for strategic planning, as shown in figure 3.

Figure 3. Generative AI use case categories

When selecting a Generative AI use case, identify opportunities to innovate, differentiate, and enhance current offerings. For new solutions, prioritize based on the 3E’s (Ease to Sell, Ease to Build, Ease to Replicate) and assess them against eight key areas using the specific questions in the radar chart below.

Figure 4. Radar Chart for assessing Generative AI use case

Once the use cases are identified, follow this workflow:

• Defining the Problem Statement:
  Collaborate with clients to clearly define the problem and target use case through a working backwards approach. Understand their challenges, requirements, and desired outcomes. Develop a gen AI solution statement that addresses their needs and provides tangible value, with identified KPIs and metrics to measure success.
• Solution Architecture and Integration:
  Define solution architecture aligning with AWS Well-Architected Pillars, ensuring scalability, availability, and security. Use AWS services like Amazon Bedrock, Amazon S3, Amazon SageMaker, AWS Lambda, Amazon API Gateway, and Amazon Elastic Kubernetes Service for integration and deployment. Prioritize data quality, model performance, security, and ethics. Follow a 6-step process to select the best architecture option, engaging AWS Architects if needed.

  Figure 5. 6-Step Process to Choose right Architecture

• Data Preparation and Curation:
  Data is critical for gen AI solutions, while 93% of CDOs recognize data strategy’s importance for gen AI value extraction, 57% lack a robust strategy, per a recent study. Implementing a robust data strategy based on AWS Data Services for data preparation, ingestion, storage, transformation is pivotal. Upholding data integrity and regulatory adherence is key for developing trustworthy, ethical gen AI solutions. You can implement a robust data strategy by leveraging the principles outlined in this blog.
• Model Selection and Training:
  After data preparation, select the appropriate gen AI model and training approach using AWS machine learning services. When selecting a service, consider your specific needs. Amazon offers diverse options Amazon Q for out-of-the-box solutions, Amazon Bedrock for managed services, Amazon SageMaker for custom ML development, and AWS Trainium for proprietary models. Each caters to different levels of customization and expertise, ensuring you find the perfect fit for your application.
  

  Figure 6. Model selection framework

• Deployment and Testing:
  For application deployment, you can select patterns like blue-green, canary, and in-place, along with services like AWS CloudFormation, AWS Elastic Beanstalk, AWS Lambda to simplify and automate the process. Rigorous testing, including unit, integration, and user acceptance tests, are crucial to validate performance, adherence to standards, and organizational AI tenets. Services like AWS CodeBuild, AWS CodePipeline, and Amazon DevOps Guru streamline testing and deployment, while Amazon CloudWatch and AWS CloudTrail ensure continuous monitoring and logging.

This phase is iterative, necessitating a feedback loop for continuous improvement. Regularly gather client feedback, analyze usage patterns, and monitor performance metrics using services like Amazon QuickSight, Amazon CloudWatch, and Amazon OpenSearch. Unlike traditional projects, adopting an AI-enabled culture is an ongoing process, requiring periodic monitoring and adjustments to achieve optimal performance.

Phase 3 – Delivery Activation

Start small, avoid hype, and focus on tangible value. Adapt to data challenges, manage business changes, and ensure production readiness, while considering Responsible AI from the start. The key practices for a successful delivery organization structure are highlighted below:

• • AI Center of Excellence
    Establishing an AI Center of Excellence (AICoE) is crucial for navigating the complexities of AI implementation. The CoE must be a cross-functional team that ensures responsible, cost-effective, trustworthy, and secure AI implementation. It aligns solutions with strategic priorities, addresses talent gaps, and harmonizes AI execution across the organization in a well-governed and inclusive manner, comprising experts from business, security, data science, and other domains.
  • Scalable delivery team
    Based on our experience from multiple customer engagements, there is no one-size-fits-all operating model for gen AI. It will depends on factors such as organization’s AI maturity, business size, scope, and industry-specific requirements. An adaptable operating model is key as AI capabilities mature – evolving from decentralized to centralized as demand scales, ensuring continued success.

Figure 7. Scalable AI project delivery team

• Make customer self-sufficient post delivery
  In the evolving gen AI landscape, delivering a successful project is just the start. Partners must equip customers with knowledge, tools, and support to manage and enhance AI solutions long-term.
  1. Knowledge Transfer and Enablement: Implement structured training programs, hands-on workshops, and comprehensive documentation to educate customer teams on implementation, model and application management, and maintenance.
  2. Flexibility and Adaptability: Design adaptable AI solutions and operating models, enabling customers to evolve with changing needs. Provide guidance on expanding AI use cases and scaling solutions.
  3. Ongoing Coaching and Mentorship: Assign dedicated coaches to work with a customer’s AI CoE, providing regular guidance and support to navigate challenges and evolve their AI strategy.
  4. Ecosystem Engagement: Connect customers to the AI ecosystem of providers, experts, and peers. Encourage community participation to build a self-sustaining support network and stay informed.

Partners empower customers to seamlessly manage and evolve Generative AI solutions by incorporating dedicated support, ecosystem connections, and adaptive models into the delivery plan, ensuring long-term success through collaborative partnership.

Phase 4 — Go-To-Market (GTM) Planning

During the GTM phase, partners define strategies for target buyer personas, develop value propositions, identify target industries and segments, create an overall plan including co-sell/co-delivery models, and establish a demand generation strategy.

An effective distribution strategy is vital for a partner’s gen AI success. Leverage AWS’s infrastructure and ecosystem to establish efficient channels like sales, resellers, or marketplaces, to reach target customers. Optimize channel mix, train and support partners, and monitor performance continuously. Prioritize distribution to boost customer satisfaction, drive revenue, and cement status as a trusted gen AI provider.

• Value Proposition and Messaging: Clearly articulate your gen AI offering’s unique value proposition, highlighting its efficiency, cost-effectiveness, and competitive advantages. Quantify potential benefits.
• Pricing Models: Develop a pricing strategy with different tiers or bundles to cater to varying needs and budgets, aligning with organizational procurement processes.
• Pilot Programs and PoC: Offer pilot programs or proof-of-concept initiatives to demonstrate value and build trust within customer organizations.
• Metrics and Refinement: Establish clear metrics to track success, including lead generation, conversion rates, customer satisfaction, and ROI. Regularly review and adjust the strategy.
• Content Marketing: Create informative content like blog posts, white papers, webinars, or case studies focused on gen AI applications within your chosen industry.
• Partner Network: Collaborate with partners on marketing initiatives, exploring co-marketing opportunities or joint solution offerings for broader reach.
• Sales Strategies: Develop targeted sales strategies, quantifying ROI for clients and tailoring your approach to address their specific challenges and needs.

Phase 5 — Refine and Scale

Partners in the final refine & scale phase are actively selling and delivering gen AI services on AWS, and refining their GTM strategy based on customer feedback. Through an AWS flywheel approach, they accelerate identifying high-value opportunities, creating products/experiences, and scaling rapidly. It’s a “think big, start small, scale fast” approach, continuously identifying new use cases to deliver more offerings, speeding up transformation and creating incremental value.

Figure 8. Generative AI Flywheel

For continued success with AI initiatives, organizations must tailor their operating models and methodologies to their specific context, fostering collaboration between business and technical teams. They should focus on delivering tangible business value through pragmatic AI use cases, addressing challenges proactively. Actively seek feedback from stakeholders and end-users, and leverage partner ecosystems to complement skills, scale capacity, and access broader expertise. By adopting a collaborative, value-driven, and adaptable approach, organizations can position themselves for successful AI deployments at scale.

Conclusion

As outlined in this blog post, together with executive sponsorship and the 5-phase workflow implementation, AWS Partners can establish a robust gen AI practice. This practice will aim to harness the potential of gen AI for the betterment of humanity while proactively addressing risks and mitigating adverse consequences.