Navigate the Agentic AI maze by making customers - not technology - the focal point of discussing adopting innovation

How should organizations think about adopting innovation in the age of the feeding frenzy around AI? Finding their bearings on adopting innovations such as Generative and Agentic AI is immensely challenging for organizations. Beyond the sheer noise levels, Microsoft’s CEO Satya Nadella is touting the end of Saas because Agentic AI is said to replace large parts of the segment. Google’s CEO Sundar Pichai, in a recent earnings call, suggested that 25% of its new code is being created by AI, then reviewed and accepted by engineers. If that was not enough to highlight the immense compression of innovation cycles, Accenture, in its 2025 Technology Vision, proclaims that by 2030, agents, not people, will be the primary users of most enterprises’ internal digital systems. Is software eating the world, but will AI eat software, as Nvidia’s Jensen Huang put it? Are we witnessing a fundamental change in how organizations run their IT and processes?

First and foremost, there are no simple answers to those questions. Enterprise IT equates to complexity, often compounded by technology, process, and cultural debt. I was reminded of those issues when I caught up with a good friend with whom I worked at the AI pioneer Arago. First, he shared his enthusiasm for exploring new capabilities of OpenAI o3 in terms of the advancement of reasoning and the progress toward non-deterministic automation—two issues that were central to the research efforts at Arago back in the day. However, as our discussion continued, he also shared his frustration that a fair few of his customers migrated SAP to HANA on-prem rather than embracing its cloud capabilities, given concerns about the business case and that changes are irreversible. Therefore, how can we reconcile those two vastly different perspectives on adopting innovation? One way to shed more light on those challenges is to reflect on the history of other secular technology shifts, such as automation and cloud.

1. Intelligent Automation: On face value, the promises of Agentic AI with autonomous decision-making and actions on those decisions throw us back to the early discussions on automation. We started those discussions by deliberating whether automation could conceivably lead to the end of outsourcing and even offshoring. When the dust of this excitement settled, the industry landed largely on task automation and employee productivity. While RPA was meant to be the conduit for the discussions on Digital Labor, what organizations got was mostly attended automation of largely static processes. Many deployments were brittle, and the algorithms didn’t understand contextual change. Consequently, for most organizations, the outcome was not transformation but a band-aid for underperforming processes. The gap between claims and reality is clear to see. What that means for organizations is that we must be transparent about the scope of automation programs. Are organizations looking to progress toward end-to-end process automation (thus transformation), or are they quite content with point solutions and task automation? And what are the levers to progress to goal-oriented, autonomous execution of insights delivered by LLMs?

So, what can we learn from those discussions? We urgently need to return the discussions to outcomes, business value, and how innovations can accelerate transformational journeys. To do that, we need to understand better whether Agentic AI can overcome the limitations of traditional automation approaches such as RPA or Runbooks. How can we reference progress with reasoning? Can agentic systems really understand context in environments that are not standardized? Finally, we must demystify what autonomous execution really means. In our view, there is a continuum from the next best action derived from insights from approaches like Observability and AIOps to goal-oriented and autonomous automation, culminating in the self-healing of complex systems and processes. We can only demystify the discourse by discussing business outcomes, use cases, and metrics rather than reciting the capabilities of foundational models.

2. Cloud transformation: Beyond the specifics of automation and the progress of Agentic AI, lessons from the many cloud transformation journeys can provide more valuable pointers for reframing the AI narratives. Fundamentally, progress with AI deployments can only happen through the interplay of cloud, data, and AI. While a completely different set of technologies, cloud was the most profound secular shift before the emergence of Generative and Agentic AI. Yet, the sobering reality is that more than two-thirds of cloud-native transformations are not fully achieving their strategic priorities. The[TR1]  biggest challenge in capturing value from investments in cloud-native transformation lies predominantly in aligning technology and business priorities to develop a coherent strategy.

We can’t interpret this to mean that all their transformations were failures, but it shows that significant business value is left on the table. Many mid-level IT stakeholders focus on technology transformation, while senior-level management and the C-suite aim for business transformation. Here, the disconnect in cloud discussions becomes tangible. Cloud is a lever for efficiency gains but not yet for broader business transformation.

Extending that thought process to the early deployments of Generative and Agentic AI, the efficiency gains from cloud can be compared to the productivity gains from AI. Yet, as an industry, we haven’t articulated the different ways of working well. In that regard, AI is directly comparable to cloud. The good news is that a compressed learning curve hopefully comes with the compression of innovation cycles. That learning curve must pivot to depicting the re-invention of business strategies, re-imagining operating models, and providing insights into how to successfully drive cultural change to progress transformation while being realistic in sourcing talent.

Summarizing the learnings from those examples, looking back at other innovation cycles, history shows that technological revolutions are not binary in nature. These transformations rarely result in the complete displacement of established paradigms. Instead, they foster the development of diverse ecosystems characterized by a blend of legacy and emerging models, each optimizing its specific role within a broader framework. This heterogeneity underpins innovation, enabling businesses to effectively leverage traditional strengths and cutting-edge advancements to navigate complex transitions. To make this point more tangible, we shouldn’t ask whether Generative and Agentic AI is replacing SaaS or any other paradigm. Rather, we should try to understand how innovation is optimizing and enhancing existing capabilities to transform business and operating models.

3. The Innovation/AI Adoption Navigator:

Having just joined the brilliant folks at PAC, those issues describe much of my research agenda over the coming months. I don’t profess to have all the answers, but asking relevant questions and challenging the noise on social media should lead to more realistic and relevant insights. My main motivation is to help our clients and stakeholders articulate the success factors for transformation journeys more successfully and better understand the change agents transforming and disrupting those journeys. Figure 1 points out how we can shape these narratives to make progress with adoption innovation and ultimately be more successful with those challenging transformations.

Figure 1: The AI/Innovation Navigator

Source: PAC 2025

When evaluating Agentic AI, organizations should focus on two fundamental things. First, they should reflect a continuum of autonomy, goal orientation, and adaptability of those agents, while the execution of insights increasingly gained from LLMs ranges from largely rules-based automation to agentic orchestration to systems with full agency and autonomy. Second, we must flip the discussions on AI from technology capabilities to business outcomes. Technology availability does not equate to adoption by enterprises. Specifically, organizations must consider the following issues:

Business strategy: The acceleration of AI innovation is challenging the business strategies of most organizations. Therefore, they must redefine their business objectives and understand the opportunities and pitfalls of AI to stay relevant. Their AI evaluations should focus on value creation, not just productivity gains. Most organizations struggle to measure and harness productivity. To be able to drive effective change, companies need to articulate and communicate those redefined objectives to reach all relevant stakeholders. As such, align the technology with the business strategy – not the other way around.

Operating model: Enterprises need to heed the lessons from cloud transformation. Just like cloud, AI is about different ways of working. Depending on their assessment of Agentic AI, enterprise leaders need to prepare their organizations for new hybrid work patterns where humans and technology collaborate. Thus, the discussions on Digital Labor will take on a completely new quality. The biggest mistake organizations can make is trying to retrofit all those wonderful innovations into traditional operating models.

Cultural change: It is here where most transformations fail. If we want to change how organizations fundamentally function and leverage technology, we will have to deal with the people part, and that’s always the hardest. More than ever before, we must take the concerns of employees seriously. Agentic AI has the potential to disrupt whole industries, such as contact centers and marketing. We are hearing system integrators talking about flattening their talent pyramids and taking less entry-level staff on. The big unanswered question in that regard is how organizations will ensure they have the necessary “humans-in-the-loop” and provide effective governance if staff training is drastically reduced.                                                                                                                                                                              

Talent: AI will impact talent more than any technology evolution before it. Perhaps most fundamentally, Generative and Agentic AI is not yet taught comprehensively at universities, so where is all the talent coming from to drive that disruptive change that so many pundits talk about? We have seen with cloud how difficult it is to attract and retain talent; with AI talent so scarce, this will be evidently more pronounced. Lastly, innovation cycles are becoming so compressed that talent strategies need a new agile approach.

If Agentic AI is really to become the main lever to transform our industry into services as a software model, as many pundits proclaim, these four issues and many of the topics that we did raise have to come front and center of our research. All of us at PAC are keen to make sure we help on your transformational journey.