Forget AGI: the real prize is Enterprise Artificial General Intelligence

By SiliconANGLE · June 27, 2026.

Dave Vellante and George Gilbert, analysts at SiliconANGLE, publish one of their regular 'Breaking Analysis' pieces with a provocative thesis: most of the artificial intelligence industry is competing in the wrong race. While OpenAI, Anthropic and other frontier labs keep concentrating their efforts on ever more generalized models and on reaching superintelligence, the real economic prize lies elsewhere: in what the authors call **Enterprise AGI**.

**The central argument: general AGI is already solved**

The authors agree with Ali Ghodsi, CEO of Databricks, that the practical definition of artificial general intelligence has already been achieved. Moving the goalposts toward superintelligence —what in earlier analyses they had called 'Messiah AGI'— generates little differentiating value for real enterprises. The superintelligence debate is, in their view, a costly distraction that benefits the labs in terms of public narrative, but does not solve the concrete problems organizations face.

To illustrate this argument, they use the metaphor of the Coyote and the Road Runner: Sam Altman is depicted as the Coyote (Wile E. Coyote), running at full speed past the enterprise opportunity and tumbling off the cliff of superintelligence, while Ali Ghodsi (Databricks) and Satya Nadella (Microsoft) watch from the edge like the Road Runner, knowing that the real battle is not about having the smartest model, but about owning the enterprise intelligence layer.

**What exactly is Enterprise AGI?**

It is intelligence that is unique to and owned by each company. It is not simply a model plus enterprise data. The authors define it as the ability to turn proprietary data, business processes and tacit knowledge into governed, persistent and reusable assets over which models can reason and agents can act.

The concept pivots around what they call the **System of Intelligence (SoI)**, which they describe as something akin to an enterprise ontology or a digital twin. This system is the central link: it captures how a company actually works, encoding its business meaning, its metrics, policies, relationships and institutional knowledge. It is not an auxiliary technical layer, but the organizing core of all enterprise AI.

**The conceptual debate: data communism vs. data capitalism**

To articulate the difference between the two approaches, the authors introduce two highly graphic metaphors.

*Data communism* is the model followed by the frontier labs: the smartest people in the world contribute their reasoning traces to the models, which absorb that intelligence and redistribute it to all users and companies. The result is a common and powerful intelligence layer, but the problem is that everyone receives essentially the same built-in intelligence. The model becomes smarter, but no company differentiates itself.

The authors give the example of investment banking: to automate or augment the work of an analyst valuing an acquisition, a model needs examples of the analyst's workflow, their assumptions, their judgment calls, their data sources, their intermediate reasoning and their preferred output formats. That kind of reasoning trace is costly to generate, narrow in scope and deeply specialized. But once that specialized knowledge is captured and bottled into a frontier model, it becomes part of a shared model capability. It improves the model for everyone, but it creates no proprietary advantage for any one company in particular.

*Data capitalism* is the alternative model the authors advocate. In this paradigm, intelligence is not absorbed into a generalized model and redistributed to everyone. Instead, each company captures its own data, processes, policies and tacit knowledge, governs them according to its own corporate requirements, and turns them into durable assets over which models and agents can reason.

This is the key point: frontier models remain fundamental because they provide generalized reasoning, language understanding, code generation and multimodal capabilities. But if that generalized intelligence is widely available, it cannot be the basis of sustainable enterprise differentiation. The source of advantage shifts toward what is unique to each company: its customers, products, workflows, operating rules, institutional memory, regulatory constraints, decision rights and culture.

**The architecture of the Enterprise AGI stack**

The authors describe a layered stack for Enterprise AGI:

- **Bottom layer: data platforms and systems of record.** They remain essential. They tell the company what happened. They store transactions, events, logs, documents. But on their own they do not explain the business: they do not know why something happened, what is likely to happen next, or what action should be taken.

- **Central layer: the System of Intelligence (SoI).** It is the digital representation of the company, a living model of the state of the business. It connects governed data with business meaning, metrics, policies, processes, relationships and tacit knowledge. It is the equivalent of the organization's enterprise ontology or digital twin.

- **System of Agency.** This is where agents use the System of Intelligence to answer questions, analyze options, plan actions and operationalize decisions. The critical point is that agents must not act independently of the business context: they must act through the System of Intelligence, where rules, constraints, metrics and trusted data provide the guardrails.

- **System of Engagement (on the left side of the stack).** It is the new work surface for human and agent interaction. Here users express intent, ask questions, resolve ambiguities, approve actions and interact with insights, decisions and data. It is not simply a front-end: it becomes a learning surface that captures the language, questions, corrections and decisions that help the System of Intelligence understand how the company really works.

One of the most important insights the authors highlight is that the System of Engagement and the System of Intelligence must be co-designed: the client teaches the back-end, and the back-end improves the client. This continuous feedback is what makes the system more valuable with use.

**The design tension: top-down vs. bottom-up**

The authors point to a critical design tension in this architecture. A purely top-down model of the company can take too long to build and may become obsolete before it is completed. A purely bottom-up model can learn quickly from users, queries, workflows and behavioral signals, but it risks producing silos and inconsistencies.

The most promising architecture combines both approaches: it infers what it can bottom-up, governs what must be standardized top-down, and continuously reconciles the two. This is why data capitalism is not simply about owning data, but about owning the intelligence-production system.

**The competitive landscape: who is competing for the Enterprise AGI prize**

The authors identify a broad set of players converging on this opportunity from different starting points:

- **Databricks** and **Snowflake**: modern data platforms trying to turn governed data foundations into systems of intelligence. The analysis focuses especially on Databricks' latest developments, particularly its **Genie Ontology**, presented at the recent Data + AI Summit, as the most concrete step toward the System of Intelligence layer.

- **Microsoft**: with Satya Nadella portrayed as a player who understands that the battle is over the enterprise intelligence layer, not over the smartest frontier model.

- **Palantir**: highlighted as a pioneer in this difficult work, connecting proprietary data, operational processes and domain knowledge into a model of how the company works. The authors credit it with having done this hard work before the rest of the market.

- **Google, AWS, Salesforce, SAP**: the hyperscalers see the control point forming above the infrastructure; the SaaS vendors are under pressure to move beyond systems of record.

Convergence is the main message: all these players are arriving at the same destination from different angles, which makes the competition especially intense.

**The implications for agentic AI**

This analysis has profound implications for the development of agentic systems. The authors argue that agents can only operate effectively when they are anchored in an enterprise system of intelligence. Without that context, agents act generically or, worse still, incorrectly for the organization's specific context.

The Enterprise AGI framework suggests that agents will not simply replace individuals, but that, when grounded in a system of intelligence, they will be able to support planning, control, coordination, resource allocation and organizational alignment. These are the core functions of management. The more complete the underlying intelligence layer, the more humans and agents will be able to coordinate at scale.

Challenging the myth of the one-employee company with an army of agents, the authors argue that this captures individual productivity but ignores the broader organizational opportunity. In their view, Enterprise AGI points toward larger and more complex forms of economic organization, not toward smaller companies with fewer people.

**The Amazon analogy and scale as an operating platform**

The authors turn to the example of Amazon to illustrate their thesis: the scale advantage increasingly comes from an operating platform that allows humans, software and data to coordinate across many domains. Enterprise AGI extends that logic by adding agents that can reason and act through the enterprise model.

In the industrial era, companies were organized around physical assets such as railways, warehouses, factories and assembly lines. In the AI era, companies will increasingly be organized around intelligence assets: the modeled representation of how the company works and, specifically, its processes.

**The assessment of Databricks and Genie Ontology**

The analysis uses Databricks' Data + AI Summit and the announcement of **Genie Ontology** as the primary use case for assessing where the market is heading. The authors propose that this tool is one of the most concrete attempts to build the System of Intelligence layer, although they note that work remains to be done and that the product's maturity is still being built.

The central question the analysis tries to answer is which vendors can help companies turn their unique data, processes and tacit knowledge into governed and composable intelligence assets, and whether Databricks is well positioned to lead that race.

**Risks and limitations of the proposed framework**

Although the authors present the Enterprise AGI framework with considerable conviction, the analysis itself suggests several unresolved tensions:

1. **The SoI construction problem**: building a complete enterprise ontology is an enormously costly and complex project. The authors acknowledge the tension between the top-down approach (it takes long and may become obsolete) and the bottom-up one (it generates silos). The hybrid solution they propose is conceptually sound but operationally difficult.

2. **The fragmentation risk**: if each company builds its own proprietary system of intelligence, integration costs across organizations may soar. Data capitalism may create individual competitive advantage but hinder inter-company collaboration.

3. **The threat of frontier models improving**: if labs like OpenAI and Anthropic manage to get their models to absorb enough specialized knowledge and to customize efficiently for each company, the gap with the data-capitalism approach could narrow.

4. **Tacit knowledge remains hard to capture**: the promise of turning tacit knowledge into governed assets is attractive but extremely difficult to execute. Tacit knowledge, by definition, is what experts cannot fully articulate.

**Regulatory perspective**

As industry context, the European regulatory framework —especially the EU AI Act— to some extent favors the Enterprise AGI approach by requiring transparency, traceability and accountability in high-risk AI systems. An enterprise system of intelligence with governed data, explicit rules and guardrails for agents aligns better with the EU AI Act's documentation and audit requirements than a black-box frontier model applied directly to critical business processes.

**Impact on developers and engineering teams**

For engineering teams and solution architects, the Enterprise AGI framework has important practical implications:

- The priority is no longer simply to integrate the best available model, but to build the System of Intelligence layer that anchors agents in the business context. - Investments in enterprise ontologies, knowledge graphs and governed data architectures move from being secondary projects to being projects of strategic importance. - The joint design of the system of engagement and the system of intelligence requires closer collaboration between product, data and AI engineering teams. - Proprietary training data and domain-specific reasoning traces become strategic assets that should not be shared with the frontier labs if competitive advantage is to be maintained.

**Outlook**

If the Vellante and Gilbert framework is correct, the implications for the market are significant. Frontier labs like OpenAI and Anthropic will face growing pressure to demonstrate that their models are not only the smartest in abstract terms, but that they can be natively integrated into the enterprise System of Intelligence layer. Model power alone will no longer be enough.

Data platforms like Databricks and Snowflake have a historic opportunity to become the central nervous system of Enterprise AGI, given that they already control large volumes of governed enterprise data and have trusted relationships with organizations' data teams.

The major enterprise software vendors —SAP, Salesforce— are in an ambivalent position: they have privileged access to business processes and record data, but their legacy architectures may make it difficult to build a fluid System of Intelligence.

And Palantir, which has spent years building exactly this kind of operational intelligence system, could see the market finally arrive at the place where they have long been working.

The analysis concludes with a clear warning: companies that simply rent generalized intelligence from the frontier labs without building their own proprietary system of intelligence will be at a disadvantage against competitors that turn their data, processes and tacit knowledge into composable and governed assets. In the AI economy, the intelligence everyone can buy is not an advantage. The advantage is the intelligence only you own.