Enterprise AI is moving from experimentation toward a more difficult phase: redesigning how organizations actually operate around increasingly autonomous systems. At the ISG AI Impact Summit London, executives from Lloyds Banking Group, NatWest, AstraZeneca, Diageo, Shell and other major companies will examine AI ROI, agentic systems, sovereign AI, data readiness and the governance challenges emerging as businesses move beyond pilots.

The enterprise AI conversation is changing.

After two years of generative AI pilots, companies are increasingly confronting a less glamorous but more consequential problem: how to redesign the organization around AI rather than simply adding AI to existing workflows.

That transition will be a central theme at the 2026 ISG AI Impact Summit London, taking place September 9–10 at Park Plaza Victoria London. Information Services Group (ISG), the global technology research and advisory firm hosting the event, says executives will focus on operating models, workforce strategies and data foundations needed to translate AI investment into measurable business outcomes.

Executives from Lloyds Banking Group, NatWest, Ogilvy, AstraZeneca, Diageo, Reckitt, Carlsberg and Shell are among the organizations represented on the agenda.

The focus is notable because enterprise AI adoption has reached a point where technical capability is no longer the only constraint. Organizations must now determine which decisions should be automated, which should remain under human control, how AI systems should be governed and whether the underlying data infrastructure is sufficiently reliable.

ISG describes the emerging destination as an autonomous enterprise, where AI increasingly governs execution while humans retain responsibility for outcomes.

That distinction could become one of the defining management questions of the next stage of enterprise AI.

From AI pilots to measurable ROI

One of the summit’s sessions, “From Pilot to Payback: Closing the AI ROI and Maturity Gap,” will bring together executives from Lloyds Banking Group, NatWest and Ogilvy to examine why AI experimentation does not automatically translate into financial returns.

The problem is widespread.

Companies can demonstrate that a large language model can summarize documents, generate software or automate customer interactions. The harder task is integrating those capabilities into processes with measurable productivity, revenue or cost outcomes.

That requires more than deploying an AI model.

Enterprise teams need data pipelines, identity controls, model governance, workflow integration, employee training and measurement frameworks. In many cases, they also need to redesign the underlying process rather than automate an inefficient one.

This is where the next generation of enterprise AI platforms is likely to compete with established technology ecosystems from Microsoft, Google, Amazon, Salesforce and Adobe. The differentiator will increasingly be how well AI integrates into business operations, not simply how impressive a model’s benchmark performance appears.

Agentic AI raises a different governance problem

The summit will also focus heavily on agentic AI.

Unlike conventional chatbots, AI agents can be designed to take actions across enterprise systems. An agent might retrieve information, initiate a transaction, update a customer record, negotiate a workflow or interact with another software system.

That creates significant productivity potential—but also changes the security model.

An AI system that can act needs permissions. Those permissions need to be constrained, monitored and auditable.

The summit’s startup challenge illustrates how quickly this category is developing. Participants include Ralio, which is developing a trust layer for agentic business payments; M11, an agentic trust and intelligence platform; and Envisioned AI, which describes its product as an operating system for running AI safely at scale.

The emergence of these companies reflects a broader ecosystem forming around agent identity, authorization, observability and trust.

As autonomous systems gain access to corporate data and financial workflows, conventional application security models may not be sufficient. Enterprises will need to understand not only who accessed a system, but potentially which AI agent acted, under whose authority, using what data and according to which policy.

Sovereign AI moves from policy debate to enterprise issue

Another major theme will be sovereign AI.

The concept has often been discussed at the national level, particularly around domestic compute capacity, data sovereignty and dependence on foreign AI models. For enterprises, however, sovereignty is becoming a practical technology question.

The summit’s “Sovereign AI: Who Actually Owns Your Intelligence?” panel will examine ownership, security, compliance and long-term value creation, with participation from Lewis Silkin and AstraZeneca.

The issue is particularly relevant to multinational businesses operating across jurisdictions with different privacy, cybersecurity and AI regulations.

European companies face another layer of complexity as enforcement of the EU AI Act develops in 2026. Organizations deploying AI systems therefore need to consider regulatory obligations alongside model performance, data governance and operational risk.

Data remains the foundation

Despite the attention surrounding generative and agentic AI, enterprise data remains a fundamental bottleneck.

The summit will address that issue in a panel titled “Knowing Where to Bend and Where to Hold the Line: Data Readiness in the AI Era.” Executives from VML, Sector Alarm Group, Carlsberg Group and Shell will discuss the tension between imperfect data and the governance required for trustworthy AI.

That tension is difficult to avoid.

Large enterprises often have decades of data distributed across ERP systems, customer platforms, data warehouses, spreadsheets, proprietary applications and cloud services. AI can make that information more accessible, but it does not automatically make it accurate.

An AI agent connected to unreliable data can simply automate bad decisions faster.

Consequently, enterprise AI architecture is increasingly converging around several foundational layers: data governance, identity, security, observability, model management and workflow orchestration.

The economics of AI are also changing

The summit will examine another emerging issue: the economics of AI itself.

The session “Economics of the Possible: When AI Makes the Uneconomical Viable” will feature Reckitt’s global director of Data and Shared Services, German Faraoni Heidenreich.

Falling inference costs and increasingly capable models could make previously uneconomical applications viable. Tasks that once required human labor may become candidates for automation, particularly when AI agents can perform multiple steps rather than simply generate content.

But lower technical costs do not necessarily guarantee economic value.

Enterprises still need to account for integration, governance, security, infrastructure and change-management costs. The real measure of AI maturity will therefore be whether organizations can redesign processes so that the productivity gains exceed the cost of operating the new system.

That is the larger question behind the ISG summit.

Enterprise AI is moving beyond the phase where companies ask “What can this model do?”

The next question is more difficult: “How should the organization change because the model can do it?”

The answer will determine which AI investments become durable enterprise capabilities—and which remain expensive pilots.

Market Landscape

Enterprise AI is shifting from standalone generative AI applications toward AI platforms, copilots, agentic systems and autonomous workflows.

Major technology companies are pursuing this transition from different directions. Microsoft is embedding AI into its enterprise software ecosystem, Google is combining Gemini with cloud infrastructure, Amazon is integrating generative AI and agents into AWS, while Salesforce and Adobe are incorporating AI agents into CRM and marketing workflows.

The competitive landscape is consequently expanding beyond LLM performance.

Enterprise buyers increasingly need AI governance, security, data readiness, agent orchestration, observability and integration. Regulatory requirements are reinforcing the trend, particularly in Europe.

The emerging autonomous-enterprise model also changes workforce planning. AI may automate individual tasks, but organizations still need humans to establish objectives, manage exceptions, validate outcomes and accept accountability.

The ISG summit’s emphasis on operating-model redesign reflects this shift. The next phase of enterprise AI adoption is likely to be measured less by the number of pilots launched and more by business processes redesigned, decisions improved and measurable economic value created.

Top Insights

• ISG’s London summit will examine how enterprises are redesigning operating models as generative AI evolves into autonomous and agentic systems.
• Executives from Lloyds, NatWest, AstraZeneca and major consumer companies will focus on closing the gap between AI investment and measurable ROI.
• Agentic AI introduces new requirements for identity, authorization, governance and observability as software systems gain greater ability to act independently.
• Sovereign AI is becoming an enterprise concern involving data ownership, regulatory compliance, security and long-term dependence on external AI providers.
• Data quality and governance remain critical constraints as companies connect AI models and agents to increasingly consequential business decisions.

Power Tomorrow’s Intelligence — Build It with TechEdgeAI