Soon after taking over IBM in 1993, Lou Gerstner made a judgment that would shape its future: IBM’s competitive strength should no longer be scattered across independent products and business units, but reorganized into integrated capabilities serving customers.
IBM subsequently restructured its organization, sales and core processes, reintegrating its capabilities as ‘One IBM’. What makes this case worth revisiting is less the difficulties IBM faced than how its top leader redesigned the company’s way of operating amid a shift in technology and industry paradigms.
Business history offers repeated examples. Electricity transformed manufacturing not on the day factories replaced steam engines with electric motors, but when they redesigned plants, production lines and management around the new power system. ERP similarly did more than digitize finance and inventory: unified data and processes enabled enterprises to reorganize supply chains, finance, sales and operations.
Truly general-purpose technological revolutions eventually move beyond upgrading tools to restructuring enterprises.
AI is approaching the same threshold today.
If AI only helps employees write copy, prepare meeting minutes or generate code, it remains a tool upgrade. But when agents enter procurement, inventory, contracts, scheduling, risk management, customer operations and business decisions, the question changes. Enterprises must decide which operating powers can be delegated to machines and which people must retain; which capabilities can rely on external models and which must become their own intelligence assets.
As AI moves from assistance into core operations, a harder question arises: will enterprises entrust it with real business?
That confidence depends on governance. The deeper AI goes into critical operations, the more complete business semantics, permission systems, auditing and ultimate control must be. AI cannot enter the operating core as a black box outside the organization. It must know what it may do, who authorizes it, how people intervene and roll back deviations, and how each critical action records a traceable process and rationale.
Enterprise AGI should therefore pursue autonomy and control together, rather than autonomy alone.
Once AI touches operating authority, it naturally becomes a top-leadership project. Which processes should be rewritten, which job boundaries can be crossed, and which capabilities must remain internal over the long term are questions beyond any single department.
To judge whether an enterprise has entered a deeper AI transformation, consider a more fundamental question:
Has its top leader begun redesigning the company?
Enterprises differ in AI adoption depth, business foundations and governance requirements. In practice, three types of development needs commonly coexist.
The first is AI SaaS and individual productivity. Lightweight tools improve writing, search, meetings, coding, sales and customer service, but the core remains people using tools. The key is not merely providing more applications; it is connecting those tools to trustworthy, unified enterprise data and business semantics.
The second is collective intelligence and mixed human–agent teams. As agents enter existing IT systems and work across departments and processes, the challenge becomes orchestrating, authorizing, scheduling and governing multiple agents, and forming new work networks with people. At this level, intelligence enters organizational processes.
The third, our central focus, is enterprise intelligence transformation. AI goes beyond individual efficiency and linking processes to enter sensing, analysis, decision-making and execution. It progressively forms organizational intelligence with a shared business world, context, memory and rules for action.
The first two improve efficiency and collaboration, but only the third begins to generate the enterprise’s own intelligence.
Although AHS’s enterprise AI system Agentrix supports all three paths, we believe the central question for enterprise AGI is not how many agents are deployed. It is whether an enterprise can develop an intelligence hub that understands itself, learns continuously and participates in business action.
The stronger large models become, the more important an enterprise’s own ontology and semantic structure become. As general models turn into capabilities that can be bought, replaced or switched, differentiation lies in what external models cannot inherently possess: operating data, business relationships, tacit experience, decision logic and continuous feedback.
AI moves from knowing many things to understanding a company only when customers, orders, equipment, organizations, rules, permissions, processes and their relationships become machine-understandable business semantics. Agents then act in real operations, generating feedback; business outcomes become new experience. The system progressively understands the enterprise’s state, the potential consequences of choices and, eventually, how to simulate its business world.
We believe enterprise AGI will grow gradually from within enterprises, rather than begin with the sudden arrival of an all-powerful supermodel.
First, machines must understand the enterprise; then agents can enter it. Collective intelligence precedes a shared business world. Ultimately, enterprises gain the ability to simulate business changes and continually correct their actions.
If IBM’s challenge was to turn a company divided by products and departments back into ‘One IBM’, today’s challenge is more fundamental:
Reorganizing an enterprise divided by systems, jobs, processes and data into a business world that intelligence can understand and coordinate as a whole.
This is why AI will ultimately transform organizations.
Traditional organizational structures rest on people’s limited cognitive, communication and management capacity. Departments, hierarchies and processes make large-scale collaboration possible. But as intelligence becomes a factor of production, boundaries once taken for granted can shift from sources of efficiency to sources of friction.
Future organizations may evolve from fixed departments to task-based collaboration and then swarm networks comprising people and agents. AHS’s recent organizational changes reflect the same assessment.
Product evolution and organizational evolution will ultimately converge.
From this perspective, AGI’s most significant impact on business may be less how human-like machines become than the first redesign by intelligence of the company as an organizational form that has endured for more than a century.
Tomorrow’s leaders will not simply be the companies using the most AI or agents. They will be those that first entrust real business to intelligence while retaining ultimate control; turn experience dispersed across minds and systems into shared organizational intelligence; and recognize that transformation starts by redesigning the company itself, rather than adding an AI layer to its current form.
Enterprise AGI truly begins when an enterprise becomes an intelligent system that understands itself, organizes action, absorbs feedback and continuously evolves.
AHS Future Intelligence Research Institute
