#188 Agentic AI and Creative Destruction

Key Takeaways:

• Agentic AI is reshaping enterprise software, introducing "service-as-a-software" meme and challenging traditional models.
• In this new paradigm, human roles are shifting to focus on strategic input and final decision-making, while AI manages intermediate processes.
• Traditional enterprise software schemas, like those in ERP and CRM systems, have long dictated business operations. AI and Large Language Models challenge this paradigm, potentially offering more accurate and flexible representations of business realities without rigid predefined structures.
• Software delivery is evolving from scheduled releases to continuous evolution, with AI systems learning and adapting in real-time.
• User interfaces are becoming more intuitive, even "invisible," as AI interprets user intent and executes complex tasks with minimal direction.
• The consulting industry is adapting to AI capabilities, moving towards outcome-based pricing and redefining the role of human expertise.
• Practitioners are driving this transformation, motivated by competitive pressures to leverage AI for business advantage.

The enterprise software landscape has long been a complex ecosystem, regardless of whether it's delivered through traditional on-premises installations or modern software-as-a-service (SaaS) platforms. Giant software ecosystems like SAP and Salesforce, while powerful, often require extensive customization to align with the unique requirements of each business. This complexity has given birth to a parallel industry: a multi-billion dollar ecosystem of consultants, specialists, and integrators dedicated to implementing, customizing, and maintaining these intricate systems.

Agentic AI is poised to disrupt this established order with a concept that sounds like a meme (because it is one): "service-as-a-software." Despite being meme, it encapsulates a profound shift in how we approach business solutions, effectively turning the traditional software-as-a-service model on its head.

Humans pushed to the corner

This paradigm shift redefines the role of human expertise, concentrating it at two critical junctures: the strategic input stage and the final decision-making and oversight phase. Consider legal research, for instance. Lawyers might specify case parameters and required precedents, while AI agents conducts exhaustive research and analysis. The lawyers then review AI-generated findings, interpret their relevance, and make strategic decisions based on this synthesized information.

This transformation isn't limited to the legal sector. In finance, human experts might set investment goals and risk parameters, leaving AI to handle complex calculations and market analyses. Customer service sees a similar evolution, with humans designing overarching service strategies and managing escalated issues, while AI handles routine interactions seamlessly.

In the software development life cycle (SDLC) domain, the shift is equally profound.?At the outset, Product Managers and stakeholders will focus solely on crafting detailed, real-world scenarios that vividly capture user interactions and business needs. These scenarios will serve as the primary input for AI agents, which will then dominate the intermediate stages. AI will analyze these scenarios to derive user stories and epics, generate detailed specifications and code, conduct testing, and even create initial user interfaces. Human expertise will then re-emerge at the final stage, where developers, designers, and quality assurance teams will thoroughly review and refine the AI-generated application.

The Creative Destruction of Software as We Know It

Data Model or Schema

The schema, or data model, in enterprise software systems acts as a hidden puppeteer, invisibly pulling the strings that control entire business ecosystems. Enterprise Resource Planning (ERP) and Customer Relationship Management (CRM) systems like SAP and Salesforce have built empires on their highly opinionated data models. These schemas don't merely organize data; they impose a specific worldview on how businesses should operate, compelling adopters to conform to their structure and logic. However, beyond regulatory and accounting requirements, there's little evidence that these entrenched data models offer the best representation of business reality.

The rise of Large Language Models (LLMs) and AI-powered applications challenges this paradigm. These advanced systems can process vast amounts of unstructured data, potentially uncovering insights and relationships that rigid schemas might miss. By freeing businesses from the constraints of predefined data models, AI-driven approaches could offer more accurate, flexible, and adaptive representations of business realities. This shift promises not just to improve efficiency, but to fundamentally transform how we conceptualize and interact with business data in software applications.

Software Delivery

The traditional concept of software releases, rooted in the need for controlled change management, is being fundamentally challenged by the rise of AI-driven systems. Conventionally, releases served to introduce new features, fix bugs, and maintain security in a structured manner. This approach reflected an opinionated view of software development, emphasizing deliberate, scheduled, and comprehensive changes.

However, in an AI-driven world, the notion of a "release" becomes fluid and potentially obsolete. AI agents, particularly those powered by machine learning, are in a constant state of evolution, learning from new data and interactions. Adding new functions might involve fine-tuning models or adjusting parameters rather than writing new code. Significant improvements often come from better data rather than code changes. This shift means that "change" in AI systems is more about continuous model refinement, knowledge base expansion, and adaptive interfaces that evolve based on usage patterns and context.

Human Interface & User Experience

Traditional software interfaces are cluttered with buttons, drop-downs, and checkboxes for a specific reason: they reflect points in workflows where human agency was required. These interfaces were designed to accommodate human decision-making and input at various stages of a process. As a result, we've ended up with complex, often cumbersome user experiences that guide users through predetermined steps, each requiring explicit human interaction.

The rise of AI agents is set to dramatically change this paradigm. These agents, named for their ability to make decisions and act independently, can take over many of the intermediate steps that previously required human input. By interpreting user intent, accessing relevant information, and making decisions autonomously, AI can eliminate the need for numerous interaction points. This shift promises to simplify interfaces significantly, potentially leading to more intuitive or even "invisible" experiences.

The Creative Destruction of Consulting as We Know It

The software consulting and systems integration (SI) industry stands on the brink of a profound transformation, simultaneously facing disruption and enabling it. Generative AI and agentic workflows introduce a new layer of complexity that plays to the strengths of the consulting industry. However, this very technology also challenges the industry's long-standing business models and operational paradigms.

AI's growing capability to generate code, conduct tests, and perform basic debugging challenges the traditional billing model based on developer headcount and hours worked. Clients will likely shift away from paying for developer time and instead demand outcome-based pricing. This transition to result-oriented models will be challenging for consulting firms, requiring them to develop new metrics for measuring the effectiveness and efficiency of AI-augmented development teams.

The role of human developers in consulting firms will likely evolve as AI takes over more routine coding tasks. Consultancies may need to focus on providing expertise in AI integration, complex problem-solving, and ensuring AI-generated code meets business requirements and quality standards. This shift will likely happen gradually and vary across different sectors and project types. Successful firms will be those that can effectively balance AI capabilities with human expertise, adapting their services and workforce skills to meet changing client needs in this new landscape.

Practitioners: The Pacers in the Creative Destruction

Practitioners - companies leveraging technology for their core business functions - will likely drive the pace of creative destruction in both the Independent Software Vendor (ISV) and Systems Integrator (SI) spaces. These practitioners control the budgets that ISVs and SIs depend on.

While there's a rush to invest in AI technologies, with companies rapidly purchasing GPUs, many practitioners, such as banks, are not in a hurry to implement these technologies immediately. What prompts practitioners to bridge this gap? Although cost reduction and increased efficiency are often cited as key drivers, the true motivating factor is competition. As some companies successfully leverage AI to gain competitive advantages, others will feel the pressure to follow suit.

Conclusion

As intelligence gets infused into every step of digital workflows, the true power of creative destruction will come into play. This AI-driven juggernaut will reveal which steps in current processes have been unnecessary or inefficient. Many workflow stages were originally designed to accommodate human intervention, decision-making, or error-checking. With AI's capabilities, these stages may become obsolete or radically transformed. The integration of AI into workflows will likely lead to a fundamental reimagining of business processes, resulting in faster, more efficient operations.