Gartner’s new graphic on “The evolution of multiagent systems” stopped me mid‑scroll. It’s a clean three‑phase picture in 2025:
- Phase 1: Single platform – multiple agents created and hosted in one platform
- Phase 2: Cross‑platform – agents on different platforms interacting via protocols
- Phase 3: Internet of Agents – a global network of interconnected agents, discovering and interacting with one another
If you strip off the branding and the date stamp, it is almost a redraw of the Metaprise architecture I first sketched in 1998–99 and published publicly in 2007.
The good news: the mainstream is finally catching up to the architecture.
The bad news: the same critical gap is still missing.
Back in 2007, in “Dangerous Supply Chain Myths Revisited (Part 7): Enabling Technology – The Emergence of the Metaprise,” I defined the Metaprise this way in 1998/99TM:
“A Metaprise is a synchronized versus sequential architecture that simultaneously links or incorporates the unique operating attributes of all transactional stakeholders on a real‑world, real‑time basis.”
The core ideas were:
- Decentralized, agent‑based coordination across multiple organizations
- Real‑time synchronization instead of batch, sequential processes
- Centralized objectives achieved through decentralized operating control
If you overlay that on Gartner’s phases:
Phase 3, “a global network of interconnected agents, discovering and interacting with one another,” is the Metaprise by another name. The only real difference is vocabulary.
The Virginia eVA program, which I wrote about in “Yes Virginia! There is more to e‑procurement than software!” in 2007, showed this in production: a decentralized, agent‑based architecture delivering 80–90% spend throughput long before “Internet of Agents” made it onto a slide.
In February 2025, “What is the difference between a Metaprise and an orchestration and intake model in ProcureTech?” revisited this distinction explicitly.
- Orchestration & intake optimizes flows within one enterprise, even if it touches several platforms. This is Gartner’s Phase 1 and Phase 2: agents inside a suite, then agents crossing a few systems via protocols.
- Metaprise spans multiple businesses: suppliers, logistics partners, regulators, finance, and customers connected through a shared agent‑based fabric.
Orchestration is about coordination inside the firm.
Metaprise is about coordination across the ecosystem.
Gartner’s Phase 3 is precisely that cross‑ecosystem layer. Their language—“a global network of interconnected agents”—is the 2025 restatement of a 1998 idea that has already worked in the wild.
So far, so good. The architecture is converging.
The Missing Piece: Phase 0
Where Gartner’s graphic stops is where most implementations eventually fail.
The post talks about MAS “boosting efficiency and future‑proofing operations” and then adds a brief warning: “with new power comes new risks,” which governance and interoperability can mitigate. Governance, interoperability, and risk management are important. But they all live after the decision to deploy multi‑agent systems.
What is still missing is Phase 0:
- A pre‑decision readiness diagnostic that asks:
- Given our incentives, decision rights, exception patterns, data reality, and human capacity, can we absorb agents at Phase 1?
- Are we structurally ready to move from single‑platform automation to cross‑platform orchestration—and eventually to an Internet of Agents—without simply scaling failure?
- A fit model (Hansen Fit Score) that tells you where you are on that path, and what your next realistic step is, instead of assuming a straight‑line progression through all three phases.
Gartner’s phases are architecture and topology. The Metaprise adds orchestration and semantics. Phase 0 adds readiness and restraint.
Without Phase 0, “governance and interoperability” risk becoming what we have seen in every previous wave: guardrails around initiatives that should never have been launched in their current form.
Why This Matters Now
If this were only about labeling, it would be academic. It isn’t.
- In 2007, the Virginia eVA case showed that an agent‑based, decentralized architecture could deliver real results when it was designed around actual stakeholder behavior and readiness, not vendor roadmaps.
- In 1998, the DND project demonstrated how treating technicians, planners, and systems as agents in a shared operating model could deliver sustained 97+% fulfillment accuracy.
- From 2007 to 2025, the archives are full of examples where organizations skipped readiness, bought the next platform wave, and landed in the same 75–85% failure bucket we now see repeating with AI and agentic systems.
The Gartner graphic is not wrong. It is incomplete.
- Phase 1, Phase 2, and Phase 3 describe where technology can go.
- Phase 0 decides whether you should go there now, later, or not at all.
That is not something a vendor can answer for you, and it’s not something an “Internet of Agents” will magically fix. It is an organizational discipline.
The Opportunity: Use Their Map, Bring Your Operating System
For practitioners, the practical move is simple:
- Use Gartner’s phases as a shared language. It’s a useful way to talk with IT, vendors, and boards about where the architecture is heading.
- Anchor the conversation in Metaprise, not in suites. Make it clear that the goal is a cross‑enterprise agent ecosystem, not a bigger proprietary platform.
- Insert Phase 0 explicitly. Before anyone draws a roadmap from Phase 1 to Phase 3, run a readiness and fit assessment. Ask, in plain language, “Can we absorb this?” and be prepared for the answer to be “not yet.”
- Treat governance as downstream of readiness. ISO 42001, NIST, IEEE, and multi‑agent risk frameworks matter—but they cannot salvage a system the organization was never ready to deploy.
Gartner has finally drawn the same map the Metaprise sketched 27 years ago. The job now is to make sure organizations do not mistake a map for a guarantee.
The architecture is converging. Whether the outcomes do will depend on whether we add the one phase that is still missing from every pretty diagram: the unglamorous, non‑negotiable work of Phase 0.
-30-
Procurement Insights
When Gartner Rediscovers The Metaprise (And Still Skips Phase 0)
Posted on December 30, 2025
0
Gartner’s new graphic on “The evolution of multiagent systems” stopped me mid‑scroll. It’s a clean three‑phase picture in 2025:
If you strip off the branding and the date stamp, it is almost a redraw of the Metaprise architecture I first sketched in 1998–99 and published publicly in 2007.
The good news: the mainstream is finally catching up to the architecture.
The bad news: the same critical gap is still missing.
The Metaprise, 2007: An Internet of Agents Before It Had a Name
Back in 2007, in “Dangerous Supply Chain Myths Revisited (Part 7): Enabling Technology – The Emergence of the Metaprise,” I defined the Metaprise this way in 1998/99TM:
The core ideas were:
If you overlay that on Gartner’s phases:
Phase 3, “a global network of interconnected agents, discovering and interacting with one another,” is the Metaprise by another name. The only real difference is vocabulary.
The Virginia eVA program, which I wrote about in “Yes Virginia! There is more to e‑procurement than software!” in 2007, showed this in production: a decentralized, agent‑based architecture delivering 80–90% spend throughput long before “Internet of Agents” made it onto a slide.
Orchestration vs. Metaprise: Why Phase 3 Is Different
In February 2025, “What is the difference between a Metaprise and an orchestration and intake model in ProcureTech?” revisited this distinction explicitly.
Orchestration is about coordination inside the firm.
Metaprise is about coordination across the ecosystem.
Gartner’s Phase 3 is precisely that cross‑ecosystem layer. Their language—“a global network of interconnected agents”—is the 2025 restatement of a 1998 idea that has already worked in the wild.
So far, so good. The architecture is converging.
The Missing Piece: Phase 0
Where Gartner’s graphic stops is where most implementations eventually fail.
The post talks about MAS “boosting efficiency and future‑proofing operations” and then adds a brief warning: “with new power comes new risks,” which governance and interoperability can mitigate. Governance, interoperability, and risk management are important. But they all live after the decision to deploy multi‑agent systems.
What is still missing is Phase 0:
Gartner’s phases are architecture and topology. The Metaprise adds orchestration and semantics. Phase 0 adds readiness and restraint.
Without Phase 0, “governance and interoperability” risk becoming what we have seen in every previous wave: guardrails around initiatives that should never have been launched in their current form.
Why This Matters Now
If this were only about labeling, it would be academic. It isn’t.
The Gartner graphic is not wrong. It is incomplete.
That is not something a vendor can answer for you, and it’s not something an “Internet of Agents” will magically fix. It is an organizational discipline.
The Opportunity: Use Their Map, Bring Your Operating System
For practitioners, the practical move is simple:
Gartner has finally drawn the same map the Metaprise sketched 27 years ago. The job now is to make sure organizations do not mistake a map for a guarantee.
The architecture is converging. Whether the outcomes do will depend on whether we add the one phase that is still missing from every pretty diagram: the unglamorous, non‑negotiable work of Phase 0.
-30-
Share this:
Related