Prior to the arrival of gen AI, analytical AI had already generated $11-18 trillion in value globally, largely through prediction and optimization. Adoption remained limited to experts, though. Gen AI changed this by democratizing access: 78% of companies now report using it in at least one function, adding an estimated $2.6-4.4 trillion in potential value.
Yet, this enthusiasm hasn’t translated into results. Over 80% of companies report no material earnings contribution from gen AI initiatives, and only 1% consider their strategy “mature.” Call it the “gen AI paradox”: tremendous energy, investment, and potential… but minimal impact at scale.
Many organizations have deployed “horizontal” AI use cases such as copilots and chatbots; nearly 70% of Fortune 500 companies, for example, use Microsoft 365 Copilot. But while these accessible tools enhance individual productivity, their impact is spread too thin to impact bottom lines.
By contrast, “vertical” use cases – embedded into specific business processes – show greater economic potential. But McKinsey research shows around 90% of such efforts to be stuck in the pilot stage. Even when deployed, they typically support isolated steps reactively rather than operating autonomously.
Adoption of vertical AI faces several obstacles: they are often isolated bottom-up initiatives, leading to fragmentation; a lack of packaged solutions often requires custom development; first-generation LLMs are hampered by their inherent passivity, unreliability, and inability to deal with complex, multi-step problems; data quality and accessibility is often uneven. Vertical AI can also face cultural resistance from business teams fearing disruption.
Despite its limited bottom-line impact, though, gen AI’s first wave has enabled broad experimentation, accelerated AI familiarity, and helped organizations build essential capabilities in prompt engineering, model evaluation, and governance. This lays the groundwork for a more integrated and transformative second phase: the emerging age of AI agents.
LLMs have revolutionized how organizations interact with data – enabling information synthesis, content generation, and natural language interaction – but in a fundamentally reactive way, isolated from other enterprise systems.
AI agents extend gen AI from reactive content generation to autonomous execution. They combine LLMs with memory, planning, and orchestration capabilities to understand goals, break them into sub-tasks, and execute actions with minimal human intervention. This expands the potential of horizontal solutions: agentic copilots become proactive teammates that can monitor dashboards, trigger workflows, follow up on open actions, and deliver insights in real time.
But the real breakthrough comes in the vertical realm, where agentic AI can automate complex business workflows involving multiple steps, actors, and systems – processes beyond the capabilities of first-generation gen AI tools.
Agents transform processes in five ways: by eliminating delays through parallel execution, adding real-time adaptability, enabling personalization at scale, bringing elastic capacity, and making operations more resilient through continuous monitoring.
For example, in supply chain management, an AI agent could continuously forecast demand, identify risks, and automatically reallocate inventory across warehouses while negotiating with external systems, improving service levels while reducing costs.
Agents can drive growth by amplifying existing revenue streams. For example: by proposing up- or cross-selling opportunities in an online store, based on analysis of user behavior, cart content, and other context. They can also unlock additional revenue streams: new automated offerings such as maintenance subscriptions for industrial companies, or bundled SaaS tools that provide interactive expertise to clients that need custom legal, tax, or procurement advice.
In short, agentic AI doesn’t just automate. It redefines how organizations operate, adapt, and create value.
When agents are simply embedded into legacy processes, they become faster assistants, typically offering productivity gains of 5-10%. True breakthrough gains, though, require reimagining processes from the ground up.
That includes reordering steps, reallocating responsibilities between humans and agents, and designing the process to exploit AI’s strengths, such as parallel execution, real-time adaptability, deep personalization at scale, and capacity that flexes instantly with demand.
Consider a hypothetical customer call center. Initially, human support staff use gen AI tools to retrieve articles from knowledge bases, summarize ticket histories, and help draft responses. Were the center to introduce AI agents but preserve existing workflows, impacts would be greater – an estimated 20-40% savings in time and 30-50% percent reduction in backlog.
But full process reinvention – where AI agents proactively detect issues, initiate automatic resolutions, and communicate directly with customers – could resolve up to 80% of incidents autonomously, leading to a 60-90% reduction in resolution time. Human agents in this scenario are now escalation managers and service quality overseers, brought in when agents detect uncertainty or exceptions to typical patterns.
Of course, not every business process requires full reinvention. Simple task automation is sufficient for standardized, repetitive workflows with limited variability, such as payroll processing or travel expense approvals. Complex, cross-functional processes, though – that are prone to exceptions or tightly linked to business performance – often warrant full redesign.
Agentic AI doesn’t just automate. It redefines how organizations operate, adapt, and create value.
Scaling agents requires overcoming three challenges: managing the new systemic risks they pose, blending custom and off-the-shelf systems, and staying agile (and avoiding vendor lock-in) amid rapidly evolving technology. These challenges cannot be addressed by merely bolting new components, such as memory stores or orchestration engines, on top of existing gen AI stacks.
What’s needed is a fundamental shift, from static, LLM-centric infrastructure to a dynamic, modular, and governed environment built specifically for agent-based intelligence – the agentic AI mesh.
The agentic AI mesh provides a unified framework that enables multiple agents (custom-built and off-the-shelf) to reason, collaborate, and act autonomously across a wide array of systems, tools, and language models – securely, at scale, and built to evolve with the technology. It draws on five interlinked design principles:
Organizations will also need to adapt their LLM strategies for agent-specific needs: deployments that demand real-time responses, for example, will need LLMs with low-latency inference; agents operating in regulated or knowledge-intensive domains (e.g. legal, finance, healthcare) require LLMs that can be fine-tuned and instrumented with external tools.
In the mid to long term, organizations must move past APIs to develop agent-first IT architectures, in which user interfaces, logic, and data access layers are natively designed for machines rather than humans. Instead of screens and forms, such systems are organized around machine-readable interfaces, autonomous workflows, and agent-led decision flows.
This shift is already underway. Microsoft is embedding agents into the core of Dynamics 365 and Microsoft 365 via Copilot Studio; Salesforce is expanding Agentforce into a multi-agent orchestration layer; SAP is rearchitecting its Business Technology Platform (BTP) to support agent integration through Joule. The future of enterprise software is not just AI-augmented, but agent-native.
As agents evolve and scale, they will introduce both technical and organizational complexity, creating challenges of coordination, judgment, and trust. This will play out in three critical ways:
To capitalize on the opportunity presented by agentic AI, organizations must fundamentally reshape their AI transformation approach across four dimensions:
The rise of AI agents presents a strategic inflection point that will redefine how companies operate, compete, and create value. This is a pivot that cannot be delegated – it must be initiated and led by the CEO. It will rely on three key actions:
Like any disruptive technology, AI agents offer laggards a chance to leapfrog their competitiveness. Deployed wrong – or not at all – they risk accelerating the decline of today’s market leaders.
Agentic AI continues to evolve, but is already mature enough to drive real change across industries. To realize its promise, CEOs must approach AI transformation as a series of focused, end-to-end reinvention efforts. That means identifying a few business domains with the greatest potential, then pulling every lever: from reimagining workflows to redistributing tasks between humans and machines to embracing new operating models.
Some leaders are already moving – not just deploying fleets of agents, but rewiring their organizations to harness their full disruptive potential. Moderna, for example, merged its HR and IT leadership, signaling that AI is now a workforce-shaping force. This is a structural move toward a new kind of enterprise.
Agentic AI is not an incremental step – it is the foundation of the next-generation operating model. CEOs who act now won’t just gain a performance edge. They will redefine how their organizations think, decide, and execute. The time for exploration is ending. The time for transformation is now.
This is an excerpt from QuantumBlack’s report Seizing the agentic AI advantage.
