LLMs and Generative AI in the enterprise.
Inspire, develop, and guide a winning organization.
Understand the unique values and behaviors of a successful organization.
Create visible workflows to achieve well-architected software.
Understand and use meaningful data to measure success.
Integrate and automate quality, security, and compliance into daily work.
An on-demand learning experience from the people who brought you The Phoenix Project, Team Topologies, Accelerate, and more.
Learn how to enhance collaboration and performance in large-scale organizations through Flow Engineering
Learn how making work visible, value stream management, and flow metrics can affect change in your organization.
Clarify team interactions for fast flow using simple sense-making approaches and tools.
Multiple award-winning CTO, researcher, and bestselling author Gene Kim hosts enterprise technology and business leaders.
In the first part of this two-part episode of The Idealcast, Gene Kim speaks with Dr. Ron Westrum, Emeritus Professor of Sociology at Eastern Michigan University.
In the first episode of Season 2 of The Idealcast, Gene Kim speaks with Admiral John Richardson, who served as Chief of Naval Operations for four years.
Exploring the impact of GenAI in our organizations & creating business impact through technology leadership.
DevOps best practices, case studies, organizational change, ways of working, and the latest thinking affecting business and technology leadership.
The debate over in-office versus remote work misses a fundamental truth: high-performing teams succeed based on how they’re organized, not where they sit.
Leaders can help their organizations move from the danger zone to the winning zone by changing how they wire their organization’s social circuitry.
The values and philosophies that frame the processes, procedures, and practices of DevOps.
This post presents the four key metrics to measure software delivery performance.
April 28, 2025
As enterprises build and scale their GenAI implementations, forward-thinking leaders are already anticipating the next wave of innovations. This article explores emerging trends in enterprise AI and provides practical guidance for preparing your organization for what’s coming next.
The GenAI landscape is evolving at breakneck speed. As John Willis observes in “Dear CIO: Navigating the Shadows,” there’s often tension between AI innovation and traditional IT governance. CEOs are hiring Chief AI Officers and instructing them to move quickly, sometimes bypassing CIOs and established IT processes. This rapid evolution creates both opportunities and challenges for enterprise leaders who must balance innovation with responsible implementation.
Three key trends are reshaping what’s possible with enterprise GenAI:
Early GenAI applications focused primarily on text, but the technology is rapidly expanding to encompass multiple forms of content simultaneously. Paige Bailey from Google describes how their latest models can process and generate text, images, and audio in a single integrated workflow.
This multimodal capability is transforming applications across industries:
In media and entertainment, Spotify has implemented AI to translate podcasts into different languages. Rather than simply generating subtitles, their system transcribes the original audio, translates the content, and then recreates the podcast host’s voice speaking the new language. Joe Beutler from OpenAI explains that this approach makes content accessible to international audiences at a scale that would be impractical with traditional translation methods.
In healthcare, Color Health has deployed vision-based AI to analyze medical records for cancer patients. Their system can review complex medical documentation in just five minutes, dramatically improving care in underserved areas where specialists are scarce. This addresses a critical gap for patients in regions like California’s Central Valley, who previously faced long delays for expert review of their cases.
In customer experience, companies are creating assistants that can analyze product images, screenshots, or videos to provide more contextual help. Rather than asking customers to describe technical issues, these systems can directly analyze visual information to diagnose problems.
Organizations should start exploring multimodal applications now, even if just for internal use cases, to build familiarity with these powerful new capabilities. Google is even embedding multimodal AI capabilities directly into devices like Chrome browsers and Pixel phones, allowing for local processing without sending data to servers.
Beyond responding to direct queries, the next generation of AI systems can actively pursue goals through multi-step processes. In “Enterprise GenAI Delivery Patterns,” Damon Edwards and colleagues describe how agents can follow strategies to achieve objectives, going beyond simple question-answering to more sophisticated task completion.
These agentic systems are characterized by:
Tool usage and function calling: They can interact with external systems, databases, and APIs based on user needs. Instead of just providing information, they can take actions like searching records, creating tickets, or updating systems.
Orchestrated workflows: Rather than relying on a single model, these systems coordinate multiple specialized components. Cisco’s implementation routes queries to different services based on intent, combining their outputs to deliver comprehensive solutions.
Memory and persistence: They maintain context across interactions, learning from past experiences to improve future performance. MongoDB’s incident response system tracks the state of ongoing issues, progressively refining its analysis as more information becomes available.
Organizations are already deploying agentic systems for specific use cases. At Google, Paige Bailey describes how they’ve developed software agents that can decompose a product requirements document into subtasks, create tickets for each component, and assign them to appropriate team members or other agents. This automates much of the project management overhead that traditionally slows development.
While still emerging, these capabilities represent the next frontier in enterprise AI, moving beyond passive assistance to proactive collaboration.
A critical limitation of early GenAI systems was their restricted context window—how much information they could process at once. The latest models have dramatically expanded these capabilities.
This expanded capacity allows models to process:
This expanded capability enables entirely new use cases. Paige Bailey describes how she was able to upload two versions of a software framework to Google’s AI Studio and ask Gemini to create a comprehensive blog post summarizing all the changes. The model processed over 750,000 tokens and completed the task in just 30 seconds.
This capability reduces the need for complex engineering that was previously required to work around context limitations. Instead of building sophisticated data chunking and retrieval systems, organizations can often directly feed relevant information to the model, simplifying implementation, and accelerating development.
As GenAI becomes more deeply integrated into enterprise operations, organizational structures and roles are evolving to support this transformation.
Swyx Wang articulates how AI engineering differs from traditional software development. Traditional engineering focuses on deterministic systems where you make it work, make it right, and make it fast. AI engineering, however, requires a different approach since systems can never be completely “right” in the traditional sense—they’ll always have some level of non-determinism.
Instead, AI engineers focus on three different phases:
Wang emphasizes that successful AI engineers focus on extracting practical utility from AI models. While they consider ethical issues and safety concerns, they maintain a pragmatic focus on delivering business value. His perspective is that even if model development were to stop advancing today, organizations would still have decades of work ahead implementing the capabilities we already have.
James Wickett argues persuasively that AI will transform security in the same way the cloud transformed operations. Traditional security approaches have often focused on ”shifting left”—pushing security responsibilities to developers earlier in the process. While well-intentioned, this has added complexity to development workflows that many teams struggle to manage effectively.
AI offers a better approach by making security contextual and accessible. Instead of forcing developers to learn complex pattern matching and security rules, natural language interfaces can help them understand potential issues in plain English. For example, rather than writing complex rules to detect encryption vulnerabilities, teams can simply ask, “Is this code change modifying encryption for sensitive data?” and receive clear, actionable responses.
This approach helps both development and security teams. Developers can more easily understand and address security concerns, while security professionals can focus on strategic issues rather than tedious manual reviews.
As organizations scale their AI initiatives, many are establishing dedicated platform teams to accelerate adoption while ensuring governance. Adobe’s approach, as described by Brian Scott, emphasizes process efficiency and resource optimization.
Adobe found that without coordination, they quickly overwhelmed stakeholders with too many use cases to review. By implementing a structured intake process and prioritization framework, they’ve been able to focus on high-value initiatives while maintaining appropriate governance.
Their platform approach includes:
As GenAI initiatives scale, organizations must address several economic considerations that differ from traditional software.
AI systems have unique cost structures that organizations need to manage carefully. Unlike traditional software, where costs are primarily driven by development and infrastructure, AI systems typically incur ongoing usage costs based on the volume of data processed.
Patrick Debois notes that even large companies like Microsoft are feeling the pressure of these costs and looking for ways to optimize. The challenge is particularly acute for applications with large context windows or high-volume usage, where costs can quickly exceed budgets if not carefully managed.
Organizations are developing various strategies to address these challenges, including:
As organizations incorporate AI into their products and services, they need to evolve their pricing and business models. Fernando Cornago describes how adidas has reimagined its architecture to achieve greater efficiency, ensuring that a 10% growth in sales creates only a 1-2% increase in running costs.
This focus on sustainable, scalable cost structures is essential for maintaining competitive margins as AI becomes table stakes in many industries.
Based on the experiences of enterprise leaders, several strategies emerge for positioning your organization for the next wave of AI innovation:
Anand Raghavan emphasizes the importance of designing AI infrastructure with metrics and feedback in mind. Traditional usage metrics like daily active users may not tell the full story for AI systems. Instead, organizations should consider more nuanced measures like the percentage of interactions that involve AI assistance or the number of queries per session.
Effective feedback mechanisms are equally critical. Users should be able to easily indicate when AI responses are helpful or unhelpful, with this feedback feeding directly into improvement cycles. This creates a virtuous cycle where systems continuously improve based on real-world usage.
Google has taken a comprehensive approach to building AI capabilities across its organization. Rather than treating AI as a separate technology silo, they’ve integrated it throughout their workflows and tools. As Paige Bailey explains, this extends far beyond code editors to encompass everything from email and documents to specialized development tools.
This holistic approach recognizes that knowledge workers, including developers, perform many different tasks throughout their day. By augmenting all of these activities with AI, organizations can achieve greater productivity improvements than by focusing on narrow use cases.
Brian Scott shares a key lesson from Adobe’s experience: perfectionism can be the enemy of progress when it comes to AI governance. Rather than attempting to design the perfect process from the beginning, organizations should start with a minimum viable approach and iterate based on feedback.
This allows governance to evolve alongside the technology itself, addressing real issues as they emerge rather than attempting to anticipate every possible scenario in advance.
Fernando Cornago describes how adidas has created a learning culture by tracking team health metrics and explicitly focusing on learning and enjoyment. This approach recognizes that AI adoption is as much about people and culture as it is about technology.
By making learning and experimentation part of the core values, organizations can maintain momentum and engagement as the technology continues to evolve.
While the technical aspects of GenAI continue to evolve rapidly, the most successful organizations recognize that the ultimate goal is not to replace human creativity and judgment but to enhance it.
Paige Bailey emphasizes that AI’s greatest value may be in reducing the friction between having an idea and bringing it to life. Many people have experienced the frustration of getting 85% of the way through a project only to stall on the final details. AI can help bridge this gap, allowing more ideas to become reality.
Fernando Cornago raises a thought-provoking question about the future of software interfaces: are we witnessing a fundamental shift from transactional applications to conversational interfaces? This shift could be as significant as the move from command-line to graphical interfaces, reshaping how people interact with technology across the enterprise.
As we look ahead, several principles can guide enterprise leaders in navigating the GenAI landscape:
The enterprises that thrive in this new era will be those that skillfully blend human and artificial intelligence—leveraging AI for what it does best while empowering humans to focus on innovation, critical thinking, and strategic decision-making.
By building on the foundations outlined in this series, organizations can navigate the challenges and opportunities of enterprise GenAI with confidence, creating sustainable value while preparing for whatever comes next.
Managing Editor at IT Revolution working on publishing books and guidance papers for the modern business leader. I also oversee the production of the IT Revolution blog, combining the best of responsible, human-centered content with the assistance of AI tools.
No comments found
Your email address will not be published.
First Name Last Name
Δ
As enterprises build and scale their GenAI implementations, forward-thinking leaders are already anticipating the…
The Public Sector's Unique Challenge Government agencies face a fundamental problem: Imagine if you…
Moving from isolated pilots to enterprise-wide GenAI implementation requires thoughtful strategies that balance innovation…
In product development, the quest for better flow has been a constant for nearly…
