Key Takeaways
- The rapid pace of AI development means custom-built orchestration layers often become obsolete before they are even deployed.
- Foundational AI capabilities improve quarterly. Organizations cannot replicate the learning curves or operational scale of major platforms like Meta or Microsoft, making custom infrastructure a high-risk investment.
- AI orchestration isn't just a technical task; it requires expert configuration to navigate specific regulatory environments, workflows, and company cultures—areas where managed services from expert consultancies provide the most value.
- Strategic value now lies in how AI is configured and governed to achieve business results, rather than in owning the underlying infrastructure. Success is measured by business outcomes, not the delivery of custom systems.
Meta's acquisition of Manus should be keeping some people up at night, and not the ones you might expect. If your organization spent the past two years and several million dollars following a major AI transformation consultancy's advice to build proprietary orchestration infrastructure, you just watched a platform giant pay $2 billion to commoditize exactly what you were told would be your competitive advantage.
The custom agent frameworks, the bespoke knowledge graphs, the multi-month pilots that consumed your best engineering talent; all of it is now competing against capabilities that Microsoft, Google, and Meta will bundle into your existing enterprise agreements for a fraction of what you have already spent.
This should be a wakeup call to everyone patronizing the Big Four leading AI transformation consultancies.
The orchestration layer was supposed to be your competitive advantage
If you have worked with any of the major consulting firms on AI transformation in the past eighteen months, you have likely heard some version of this pitch: the models are commoditizing, the real value is in the orchestration layer, and your organization needs to build proprietary agent infrastructure that turns raw AI capabilities into enterprise workflows. This guidance was not wrong in its diagnosis, but it was catastrophically expensive in its prescription.
I have watched this play out firsthand. A pharmaceutical client I advised had engaged one of the marquee AI consultancies to build an AI-driven marketing platform, and they were initially given nothing more than a few simple interfaces, an API connection to ChatGPT, and a spreadsheet of prompts. I had to break the news to them, which led to our development of the core concepts that drove what followed: they went back to the same marquee consultancy and hired them to build an internal agent orchestration system complete with custom knowledge graphs, retrieval-augmented generation pipelines, and multi-step workflow automation tuned to their specific regulatory environment. The project consumed millions in consulting fees, internal engineering resources, and opportunity cost as key personnel were diverted from other initiatives.
Meanwhile, the pilot was still wrapping up when Microsoft announced Copilot Studio capabilities that delivered similar functionality at a fraction of the cost, and Google followed shortly after with comparable offerings in Vertex AI Agent Builder. They could have just waited and achieved even more powerful results with vastly smaller costs.
The internal system they ultimately built with this consultancy's help was not bad; it was simply rendered obsolete before it could prove its value. The biggest piece of value that was provided, ultimately, was a white-labeled localization system with a very handsome attached markup. Had the client just done their research and gotten better advice, none of this would have happened.
The platform giants are now competing where consultancies told you to invest
Meta's acquisition of Manus signals that the execution layer, the orchestration infrastructure that sits between AI models and completed work, is no longer a greenfield opportunity for enterprise differentiation. It is becoming a platform commodity. Microsoft, Google, and now Meta are racing to own this layer because they understand that whoever controls orchestration controls the distribution of AI value.
This means that the custom orchestration systems, the proprietary agent frameworks, and the bespoke knowledge graphs that major consultancies advised clients to build at enormous expense are now competing against products that platform vendors can subsidize, continuously improve, and bundle with existing enterprise agreements. The economics are simply unwinnable for most organizations. A custom-built agent system requires ongoing maintenance, security updates, model migration as capabilities evolve, and specialized talent to operate. A platform offering requires a subscription and an afternoon of configuration.
The consultancies that guided clients into these multi-million-dollar investments were not acting in bad faith; they were applying a theory of competitive advantage that the market has since invalidated. The execution layer is indeed where value gets created, but that does not mean every enterprise needs to own its execution layer any more than every enterprise needs to own its own data centers. The cloud transition taught us this lesson already, and we are about to learn it again with AI orchestration.
The velocity problem makes custom builds untenable
What makes this moment particularly painful for organizations mid-implementation is the velocity of capability release from the major platforms. The pharma client I mentioned began their agent project when the state of the art required substantial custom engineering to achieve reliable multi-step task completion. By the time their pilot reached user acceptance testing, the baseline capabilities of off-the-shelf tools had leapfrogged their custom implementation, and the roadmap of forthcoming features suggested the gap would only widen.
This is not a failure of execution; it is a structural problem with the build-versus-buy calculus in a market moving this quickly. When foundational capabilities improve on a quarterly basis, and new product announcements arrive monthly, any custom implementation that requires more than a few months to deploy risks becoming irrelevant before deployment. The consultancies that sold these engagements priced in eighteen-month implementation timelines because that is how enterprise transformation has always worked, but AI does not respect enterprise transformation timelines.
Manus processed over 147 trillion tokens and created more than 80 million virtual computers in its short history, and Meta acquired that operational scale and institutional knowledge for approximately $2 billion. Your organization cannot replicate that learning curve, and the consultancy advising you cannot either. The question is not whether your custom orchestration layer can eventually match platform capabilities; the question is whether it can match them before the next platform release renders the comparison moot.
The case for managed orchestration by expert consultancy
If building custom orchestration infrastructure is increasingly untenable and platform vendors are racing to commoditize the execution layer, what role remains for consulting expertise in enterprise AI?
The answer lies in recognizing that orchestration is not merely a technical problem; it is an organizational problem with technical components. The platforms can provide the infrastructure for agent execution, but they cannot provide the judgment about how that infrastructure should be configured for a specific regulatory environment, a particular approval workflow, or an organizational culture with implicit rules that no documentation captures. They cannot diagnose whether a communications bottleneck is a technology problem, a process problem, or a people problem, and AI only solves the first.
Managed orchestration by expert consultancy offers a different value proposition than the build-it-yourself model that has consumed so much enterprise investment. Rather than constructing proprietary infrastructure that will require perpetual maintenance and eventual replacement, managed orchestration focuses on configuring platform capabilities to organizational requirements, maintaining those configurations as both platforms and requirements evolve, and providing the institutional knowledge that makes AI systems actually work in complex environments.
This approach eliminates the complexity that has made so many enterprise AI initiatives stall. It ensures consistency across use cases because configuration and governance are centralized with experts who understand the full landscape. It converts the capital expenditure of custom development into the operational expenditure of managed services, which aligns costs with value delivery rather than front-loading investment into systems that may never reach production.
Most importantly, managed orchestration by experts who understand your domain transfers the burden of keeping pace with platform evolution from your organization to a partner whose core competency is exactly that. When Microsoft releases new Copilot capabilities or Google updates Vertex AI or Meta deploys Manus technology into enterprise offerings, your internal team does not need to evaluate, test, and integrate those changes. Your consulting partner does, and you benefit from their accumulated experience doing so across multiple clients and contexts.
Choosing consulting partners wisely in the post-Manus landscape
The Meta acquisition should prompt enterprise leaders to ask difficult questions about any ongoing or planned AI transformation engagements. If your consulting partner is advising you to build custom orchestration infrastructure, ask them how that investment will compete with platform offerings that improve quarterly. If they are recommending proprietary agent frameworks, ask them what the maintenance burden looks like in year three when the original implementation team has moved on. If they are scoping knowledge graph implementations that require months of development, ask them why the same capabilities cannot be achieved with platform-native tools in weeks.
The right consulting partners in this environment are those who understand that their value lies in configuration, governance, and organizational change rather than in custom development that platforms will inevitably subsume. They are the partners who can help you navigate the increasingly complex landscape of platform offerings, selecting the right tools for specific use cases rather than defaulting to building because building is more lucrative for the consultancy. They are the partners who measure success by business outcomes achieved rather than by systems delivered.
Meta paid $2 billion for Manus because execution infrastructure is strategically valuable. That same strategic value is why enterprises should be cautious about building it themselves and thoughtful about choosing partners who can help them leverage platform capabilities rather than compete with them. The execution layer is the new strategic frontier, and the organizations that navigate it wisely will be those who recognize that owning infrastructure is less important than owning outcomes.
About the author
Stephen C. Webster is Senior Director of Integrated Intelligence at Aquent Studios, where he leads AI transformation for Fortune 500 clients. He previously trained frontier AI models created by Google, Meta, and Amazon, and spent over 20 years as a journalist. He believes the toughest challenges in AI are not technological, they're organizational.
Latest.
The best marketing automation tools for 2026.
Engineering & Technology, Innovation & Emerging Tech, Marketing & Analytics
Reimagining org design in the fast-paced new world of AI.
Insights from InsideOut, Leadership & Management, Content & Creative
Meet the new emerging role: AI Trainer
Career Advice, Leadership & Management, Engineering & Technology, Innovation & Emerging Tech, Talent Acquisition & Recruitment