Platforms Absorb L6 Orchestration, Squeezing the Middle Layer
Google and Microsoft are bundling L6 orchestration into their L7 surfaces and packaging L5 execution skills, while Anthropic enables the trend. This compresses pure-play L6 framework vendors by shifting distribution and TCO.
The News
In May 2024, major AI platforms including Google (at I/O with Gemini advancements in Workspace) and Microsoft (at Build with new Copilot integrations) unveiled significant enhancements to their agentic infrastructure. Anthropic concurrently released improved tool-use functionality in its Claude 3 model family. This synchronized move signals a strategic shift from foundational model improvements to creating captive ecosystems where AI can perform multi-step tasks, directly competing with and absorbing the value of standalone orchestration frameworks.
Layer Scoring
Sublayer Impact Map
Which of the 50 sublayers this move actually touches, the magnitude of impact, and who plays that slice today.
Intelligence Cube · 2D
The move's footprint across the three Cube axes, Functions, Verticals, Layers, flattened into two readable 2D projections.
Layers × Verticals
14 cells · 7×2
Layers × Functions
21 cells · 7×3
Two 2D projections of the Intelligence Cube (Functions × Verticals × Layers). Filled cells = this move occupies that intersection.
Why Now
This move is happening now for three reasons. First, model capabilities (L2) just crossed a critical threshold in reasoning and reliability, making complex tool-using workflows viable at scale. Price drops on flagship API tiers and improved latency make these multi-step chains tolerable in user-facing products. Second, this is a direct manifestation of multiple structural laws: Law I is in full effect as platforms, facing commoditization of L2 access, must move up the stack to capture value at L5 and L6. This is concurrent with Law II, as they seek to establish L6 as a new value bottleneck, and Law III, by leveraging their vast L7 surfaces for unmatched distribution. Third, the cost of inference has fallen below the economic threshold required to bundle these powerful, multi-step queries into existing subscriptions, changing the TCO for customers.
The Structural Take
This convergence is a classic case of Law I — Intelligence Commoditizes Downward. The major platforms (Google, Microsoft) recognize that pure L2 model access is becoming a low-margin utility. To escape this gravity, they are pushing up the stack to absorb the emerging L6 (Orchestration) layer, packaging it with L5 (Execution) skills. Their goal is to make "agent building" a feature of their respective platforms, not a standalone product category. This directly threatens any company whose primary value is generic workflow orchestration. This also demonstrates Law III — The Surface Captures Attention; the Chain Captures Power. By embedding these L5+L6 capabilities directly into their dominant L7 Surfaces (Windows, Office 365, Google Workspace), they create an unmatched distribution channel. This allows them to push their proprietary L6 orchestration frameworks to a massive, captive audience of developers and users, creating powerful ecosystem effects and raising customer switching costs. The scarce resource being contested is developer mindshare for the next generation of AI applications. Owning the orchestration layer is the key to that kingdom.
Second-Order Effects
The primary downstream effect is a capital crunch for startups building horizontal L6 orchestration tools. Venture allocations to this category will likely decline over the next 12-18 months, with pre-money valuations for seed rounds compressing. We will see an 'acquihire' wave as standalone L6 startups with talent but no defensible GTM are absorbed by application-layer companies, mirroring the valuation reset of AI writing 'wrappers' like Jasper after ChatGPT's launch. This forces a flight to defensibility: new investment will pivot towards companies with deep vertical expertise (L5a) or proprietary data (L1b), as those are the only durable moats against this platform envelopment.
- Who Wins
- Microsoft & Google. They successfully move up the stack, commoditizing the L6 orchestration layer and tying developers and users deeper into their cloud and application ecosystems.
- Companies with L1b Proprietary Data. Their unique data becomes the essential, non-commoditizable fuel for these new agentic systems, increasing its value and their strategic leverage.
- End Users. They gain access to more powerful, integrated AI capabilities within the software they use daily, often for little to no marginal cost.
- Who's Exposed
- LangChain & CrewAI. Framework vendors focused purely on L6 orchestration face significant compression. Their core open-source value is being absorbed into the platform layer, forcing them to find new monetization paths. The winners in this category will successfully pivot to enterprise-grade L4 connectors, managed services, hosting, private deployments, or by building defensible L8 memory and L5 vertical execution features.
- Simple 'Wrapper' Startups. Companies building thin UIs on top of model APIs with simple prompting will be made obsolete by integrated, platform-native L5+L6 agent capabilities.
- Zapier & Make.com. Their core L6 orchestration of simple workflows, monetized via a vast L4 integration network, is directly threatened. Platforms will subsume the most common workflows with native L6 capabilities that leverage their own deep L4 connections (e.g., Microsoft Graph) and integrated L7 distribution, rendering the 'connector tax' of the incumbents unnecessary for a large segment of the market.
Deep Product Lens
The product shipping here is a bundled set of developer APIs and user-facing copilots. The core primitive is no longer just a model, but a "tool-using agent" that can be invoked with a goal. The key packaging choice is abstracting the L6 agent loop (e.g., ReAct framework) away from the developer, bundling it into the API call itself. This dramatically lowers the barrier to creating agentic software. The system prompts for these agents are where the magic lies, defining reasoning scaffolds, tool-use policies, and fallback behaviors. The v2 roadmap is clear: move from transient, single-session agents to persistent L8 Memory, allowing agents to learn from interactions. The wedge-to-lock-in motion is classic: start with developer APIs (wedge), expand with embedded user-facing copilots (expand), and achieve lock-in via ecosystem-specific toolsets and compounding L8 institutional knowledge.
Deep Strategy Lens
This is a textbook platform envelopment strategy. The tech giants are using their power at L0 (Cloud Compute) and L2 (Foundation Models) to subsidize an aggressive move into the adjacent L6 (Orchestration) market. By offering orchestration as a cheap, integrated feature, they raise the competitive cost for any pure-play L6 startup, which must bear the full cost of sales, marketing, and R&D for that single layer. The scarce resource being claimed is not just developer mindshare, but control over the emerging "agent" runtime environment. By dictating the protocols for agent interaction (L4b) and the tooling for orchestration (L6), they establish a powerful gatekeeping position. This forces rivals into a difficult choice: adopt the dominant platform (accepting a platform take rate and limited pricing power) or attempt to build a defensible moat in a vertical (L5a) or data (L1b) niche, which requires battling the platform's enormous CAC advantage from its L7 distribution.
The Horizontal Lens
In the SaaS vertical, this move fundamentally alters the calculus for workflow automation. Consider a mid-market company's marketing operations team. Their current buyer journey for automation involves identifying a gap, evaluating L6 players like Zapier or Make, and purchasing a tiered plan based on task volume. This budget line is distinct from their Microsoft 365 or Google Workspace spend. The new journey evaporates this process. The marketing lead simply describes a multi-step task in a Copilot prompt: 'When a new lead hits HubSpot, enrich it with Clearbit data, check our CRM for existing contacts from that domain, draft a personalized welcome email referencing their industry, and assign a follow-up task in Asana.' The platform's native L6 + L5 capabilities execute this. The incumbent's defense, Zapier's 6,000+ app connectors, is an L4 access asset that the platforms will erode by focusing on the top 200 enterprise apps first, capturing 80% of the value.
- Steelman: The Counter-Thesis
The counter-argument is that platform-native agent builders will be too generic and restrictive. Serious developers will always prefer the flexibility, control, and faster innovation cycles of open-source, neutral frameworks like LangChain. Enterprises, wary of vendor lock-in, will opt for a multi-cloud L6 layer that can orchestrate models from any provider. However, I maintain my position because the platforms' structural advantage in distribution is overwhelming. For the 90% of use cases, the convenience and zero-TCO of a deeply integrated L6/L7 stack inside Teams or Google Workspace will trump the flexibility of a standalone solution.
What to Watch (Next 90 Days)
- 01Track DAU and number of deployed automation flows on platform-native agent builders like Microsoft Copilot Studio.
- 02Changes in pricing and packaging from incumbents like Zapier or Make.com, specifically if they begin to unbundle their L4 connector network from their L6 workflow builder.
- 03The number of enterprise pilots for platform agent builders that convert to paid license upgrades within 90 days.
- 04Microsoft or Google acquiring a major data integration or API company (a pure L4 player) to bolster connectivity for their agents.
What This Means for You
Product Leader
Pick a side: deepen a layer of your own, or attach cleanly to whoever does. The middle position tends to get ground out over 12–18 months.
Investor
Position-size for binary outcomes. Track who consolidates the L4 distribution above this layer.
Operator
Run a 90-day bake-off. Hold off on lock-in until the L4 winner is clearer.
Sources
Written by Supply Chain of Intelligence™ analysis engine, reviewed weekly. By Anand Arivukkarasu · Ex-Meta Product Leader.
Share kit
Take this to LinkedIn
Three artifacts, one argument. The image carries the diagram, the short post stops the scroll, and the detailed article copies as rich text, so headings, bold lead-ins, italic standfirsts, pull-quotes and bulleted lists land in LinkedIn's Pulse editor already styled. No markdown markers, no tables, nothing to reformat by hand.
Supply Chain of Intelligence™ · Battle Card
May 31, 2024
Platforms Absorb L6 Orchestration, Squeezing the Middle Layer
Territory taken: L6 Orchestration · L7 Surface · L1 Data — This is the core battleground. Platforms are making orchestration a feature, not a…
- Microsoft & Google — They successfully move up the stack, commoditizing the L6 orchest…
- Companies with L1b Proprietary Data — Their unique data becomes the essential, non-commoditizable fuel…
- LangChain & CrewAI — Framework vendors focused purely on L6 orchestration face signifi…
- Simple 'Wrapper' Startups — Companies building thin UIs on top of model APIs with simple prom…
Expected counter-moveThe counter-argument is that platform-native agent builders will be too generic and restrictive. Serious developers will always pre…
Anand Arivukkarasu
supplychainofai.com
↑ hover the card and hit PNG to download
The L6 Orchestration layer is being compressed into the platform. The synchronized moves from Google (Gemini in Workspace) and Microsoft (Copilot) aren't just new features. It's a structural shift to absorb L6 (Orchestration), making it a feature of their existing L7 (Surface) products. This is Law I (Intelligence Commoditizes Downward) playing out in real-time. Startups that built powerful L6 agent loops are discovering they were building on rented land. The platforms, by bundling orchestration, shift the competitive landscape. It's no longer about who has the best L6 framework; it's about who owns the L7 distribution channel. The surviving companies will be those with deep moats in proprietary data (L1b), defensible vertical workflows (L5a), or compounding memory (L8). The middle is getting squeezed. This isn't an 'agent layer' war. It's a battle for ecosystem control, fought by bundling L5 skills and L6 orchestration into dominant L7 surfaces. The question is whether neutral, open frameworks can maintain relevance against the gravity of the platforms. #AIStrategy #PlatformEconomy #SupplyChainOfIntelligence #GenAI Full breakdown, with the layer map: https://supplychainofai.com/live/platforms-absorb-agent-orchestration-layer #AI #Strategy #SupplyChainOfIntelligence #ProductStrategy #VentureCapital
Get the next teardown in your inbox.
One issue when something structurally important happens, usually weekly. No spam, no filler, unsubscribe anytime.
Worth sharing? Pull-quote: "Google and Microsoft are bundling L6 orchestration into their L7 surfaces and packaging L5 execution skills, while Anthropic enables the trend. This compresses pure-play L6 framework vendors by shifting distribution and TCO."