From experimentation to transforming customer value

Since ChatGPT was released nearly three years ago, individuals and companies have experimented with reactive AI, composing AI prompts to create articles, tables, translations, to-do lists, and chatbots to resolve questions.

While these tasks offer practical value, especially boosting efficiency, the next leap isn’t just technical. Agentic AI systems can enable organizations to deploy autonomous service agents capable of managing entire organizational processes end to end.

The value of agentic AI lies in collaboration, not replacement, where humans and machines each play to their strengths. When workflows combine human judgment with machine precision, organizations can streamline tasks, innovate personalization and insights, sharpen decisions, scale, and measure outcomes that inform ongoing process upgrades.

Unlike reactive AI, agentic systems function as digital colleagues, taking initiative, pursuing goals, having memory and context, using tools to learn from outcomes, and adapting in real time. This shift unlocks incremental gains and breakthrough innovation in customer and user experience through reimagined workflows for operational excellence.

The catch?

Redefining how an organization operates to deliver differentiating value requires careful human orchestration. While agentic AI acts autonomously, it depends on human or enterprise oversight to anchor purpose, set guardrails, and ensure alignment. Effectively implementing agentic AI prioritizes the role of human employees in value creation to a higher level, ensuring transparency, ethical standards, and responsible strategic oversight at every level.

Enable seamless integration: The Model Context Protocol

To realize agentic AI’s full potential, organizations need to connect AI agents to multiple tools and data sources without building custom integrations for each one. Enter the Model Context Protocol (MCP)—an open standard that replaces fragmented integrations with a single, universal protocol.

Think of MCP as a USB-C port for AI agents. Just as USB-C standardized device connections, MCP standardizes how AI systems access databases, applications, and external services. No more writing separate code for each integration.

For businesses, this means autonomous agents can seamlessly access customer databases, CRM systems, knowledge repositories, and execute actions across platforms—all through one standardized protocol. As the ecosystem matures, AI systems maintain context while moving between tools and datasets, creating a sustainable architecture.

The result? Dramatically reduced technical complexity and agents with the contextual awareness to deliver transformative customer value.

Change management for cross-functional process reengineering

Implementing the transformative value of agentic AI requires organizational change management—redesigning organizational processes that efficiently yield quality outcomes.

Not just a new tool, impactful agentic AI implementation requires an AI expert as a vital ongoing member of cross-functional mission-based teams focused on a specific process selected for re-engineering. AI experts should not be siloed in the technical function. They should be embedded with the functional process content and outcomes experts, learning from each other as they expand organizational expertise.

As the number of reimagined process teams grows, so does the organizational expertise, extending gains achieved and staying ahead of a continuously evolving AI.    All of this requires careful orchestration of data, strategy, and organizational readiness focused on the functions to which agentic AI is applied and a work culture that adapts to discover new opportunities. This is transformational enterprise change. It is not an event but a new way of working. The potential, however, is substantial.

If your organization is still only implementing prompt engineering, you are lagging.

Prioritize process re-engineering targets

Another key factor is how to prioritize which re-engineered processes customers and users will value most by seeing how they use the product or service. In the early 1980s, then-NCR Corporation used observational research to identify the most time-consuming challenges their retail cash registers could automate. NCR collaboratively developed the Small Computer System Interface (SCSI) protocol and developed a SCSI computer chip that enabled scanning charges to replace hand entry.

Similarly standing in their customers’ shoes, Intuit engineers and product managers spur innovation by regularly engaging in “follow-me-home” with customers seeing how users apply their product features in their daily lives. This institutionalizes technical experts’ insights into customer usage to feed innovative ideas for more transformation.

Train for an AI world

Companies recognizing employees’ AI skill gaps are providing in-house or commercial training for workers. Higher ed institutions and their nonacademic competitors offer a variety of online courses. With AI continuously evolving, the next generation and their teachers also need training. The American Federation of Teachers (AFT), America’s second largest teachers’ union, is starting a training hub with $23 million in funding from Microsoft, OpenAI, and Anthropic to focus on training teachers to generate lesson plans with AI wisely, safely and ethically.

AFT’s Share My Lesson is now Beta testing an OpenAI-powered teaching assistant, TRYEdBrAIn, that can adapt lesson plans to change grade levels, translate into different languages and many other options. Beta testing is now underway to understand user experience. The Khan Academy is testing an AI-powered teacher assistant as a student tutor in various school districts.

As digital transformation accelerates, leading organizations will see agentic AI not as a tool, but as a catalyst for new paradigms of teamwork, value creation, and enterprise agility.

Barie Carmichael is senior counselor and David Sanchez is director, AI Business Solutions at APCO Worldwide.