Navigating the Realities of AI Implementation: POV from Google Cloud Next 25

This week, I officially joined Pythian as Senior Director of AI Services. Having spent a good chunk of my career in the Google ecosystem, initially in the intense world of scientific and supercomputing before focusing my academic background on artificial intelligence (AI), I'm now deeply involved in building out Pythian's AI professional services portfolio.

My unofficial first week at Pythian was an invigorating one at Google Cloud Next. 

I spoke to hundreds of people—from Googlers, to my new teammates at Pythian, to the team at Nvidia, and many many more! Talking to numerous customers, as I do in my role, gives me a front-row seat to the practical hurdles they encounter when trying to implement AI. What's been particularly interesting here at Google Cloud Next is seeing how these challenges align with the broader industry trends. 

One particularly insightful conversation was with the RavIT Show. In my conversation with RavIT, I shared some of my key takeaways, particularly around the real-world challenges that businesses are facing as they dive into the exciting world of AI. It's been a fantastic experience being here, connecting with so many bright minds and seeing the incredible advancements on display.

Agentic AI was the buzzword at GCN 25 

One thing really stood out to me at this year's event—the buzz around agents and agentic workflows. Google has clearly been investing heavily in their AI differentiators for a long time, and their approach to embedding these agentic capabilities directly into their cloud platform is something I believe will be truly impactful. Instead of just providing the building blocks, they're making these intelligent workflows more readily available.

Focus on agents and agentic workflows: Agents are a top-of-mind topic, and Google's approach of embedding these workflows into their cloud platform is seen as significant.  

Varying adoption rates across industries: Some traditionally technology-averse industries are adopting AI quickly, while others like manufacturing and supply chain have been slower with generative AI.

Navigating the AI triad: Incubation, adoption, and production

From my conversations with enterprise leaders, it's clear we're transitioning from what I call the "era of proof of concepts (PoC)" for AI, which predominantly took place in 2023 and 2024. The key challenge? Moving beyond these initial experiments to production-level AI. I see the enterprise AI journey falling into three segments—incubation, adoption, and production. Transitioning between these three segments is where most struggle.

Incubation: The PoC playground

Incubation is where organizations "play in the open field," experimenting with various AI models and use cases. This is the PoC phase, where different technologies are explored, often without a clear roadmap. While vital for initial learning, many get stuck here, failing to plan for scaling these experiments. Technically, this involves trying out different frameworks like TensorFlow or PyTorch, and diverse data sets.

Adoption: Building the AI infrastructure

Adoption is about moving to a "ready-made playground." It's about building the infrastructure—the platform, frameworks, and the AI layer itself. Without maturity in the platform and frameworks, driving value in the AI layer becomes difficult. This involves selecting the right data management systems, machine learning frameworks, and deployment pipelines. It's about establishing a cohesive AI development environment.

Production: Embedding AI in the real world

Production is where AI solutions are embedded into existing stacks and workflows, driving real business value. This requires a robust support system and a clear roadmap for maintenance and optimization. It's about taking AI from the playground to the real world, ensuring scalability and reliability. Technically, this involves deploying models at scale, monitoring performance, and ensuring robustness to data and operational changes. The key here is "embed AI," where AI services are seamlessly integrated. We also deal with "choice fatigue," helping customers choose the right models and build upon them effectively.

Mind the AI gap

The critical gap is the lack of a clear strategy to transition through these phases. Many excel at incubation but struggle to scale and integrate AI into core processes. Bridging this gap is essential for realizing AI's full potential.

Pythian's approach to AI implementation

We've built a strong foundation on our deep expertise in databases and data management, and my passion lies in helping our customers layer powerful AI capabilities on top of that solid base. 

• Building on data expertise: Pythian leverages its deep expertise in databases and data management to build a solid foundation to support AI services.

• Providing a "playground" for development: Pythian offers a structured environment for businesses to experiment with, develop, and refine their AI PoCs before taking them to production.

• Establishing robust success criteria: Pythian works with executives to define clear success criteria and metrics from the outset to ensure ROI is measurable and achievable.  

• Emphasis on foundational layers: Pythian emphasizes the importance of a mature platform and frameworks as prerequisites for successful AI adoption. We categorize implementation into platform, frameworks, and the AI layer itself.

• Guiding through choice: Pythian helps customers navigate the complex AI landscape by assisting them in choosing the right models and building effectively on top of them.  

• Enablement-focused lifecycle: Pythian offers a comprehensive lifecycle for customers, including understanding their business problems, translating them into AI strategies, providing implementation roadmaps, and offering post-implementation support.  

• Field CTO office: Experienced Field CTOs help businesses understand their needs and translate them into high-value, high-priority AI adoption strategies.

Pythian’s deep expertise, strategic platform partnerships, and expert-led suite of services ensure our customers are granted market-leading guidance on how to avoid the potential pitfalls when transitioning from one phase of the enterprise AI journey to another, and how to arrive at ROI in the most expeditious fashion possible. Uncover how Pythian can support your AI journey by taking part in our AI Workshop.