Harnessing the Stable: Why Orchestration Beats Model Wars
Owning the workflow, not the model, is the real venture prize.
Growing up as I did, more often covered in horse manure than not, one lesson was impossible to miss – a horse is only as powerful as the rider guiding it. I learned the hard way that I could be sitting on a 1,200-pound engine of muscle and speed, but without the right cues, tack, and timing, the potential went nowhere (or worse, straight over a fence, as I am often reminded by my now lopsided collar bone). The magic wasn’t just in the animal; it was in the orchestration between horse, human, and tools.
Three years into the AI revolution, it’s clear models aren’t so different from horses. Each is immensely powerful yet specialized – some quick and nimble, others steady and precise. Their true potential emerges only when directed well. That’s why knowing which to use for a given task, when to switch mounts, and how to make them work as a team has become a superpower and a store of great value.
Just as no horse combines the perfect temperament, alignment, and athleticism for every discipline, no single model will dominate every AI task. Some will excel at reasoning (jumping), others at creativity (cross-country), and still others at code or design (dressage). Users will need to toggle between them, whether consciously or invisibly, and with each new model added to the stable, the entire system grows more capable. Since these toggles can happen instantly and as often or as seldom needed with the right intelligent orchestration, why pick one horse when you can pick them all?
Cursor: A Burgeoning Blueprint for Orchestration
Cursor is a prime example of this orchestration-first approach – in this case, for developers. As an AI-native IDE, it doesn’t try to crown a single model as the best. Instead, it gives users the power to route tasks, like code generation vs. complex refactoring, to whichever model is best suited. By embedding this optionality in a familiar interface, the orchestration layer becomes the core value proposition. It manages context, preserves state, and chains outputs into a seamless experience, agnostic to the level of user sophistication.
Developers don’t adopt Cursor just because they care about the specific model under the hood. They adopt it because they want the best possible result. Cursor is only the beginning of a wave of orchestration-first products, but the pattern applies far beyond code.
Why Orchestration Isn’t a One-Trick Pony
For years, investors have been fixated on model providers across text, image, and video, and it’s easy to see why. The raw capabilities are dazzling. But this remains a risky bet. The “model wars” are capital-intensive, technically demanding, and ultimately a winner-takes-few game.
Orchestration platforms, by contrast, gain strength directly from this arms race. These platforms build increasingly defensible moats through four compounding advantages:
Model agnosticism: A new, more powerful model isn't a threat; it's an upgrade that improves the entire platform’s functionality and output
Proprietary routing intelligence: Every user interaction generates data about which models excel at which tasks, under what conditions, and for which user types. This intelligence becomes increasingly valuable and defensible
User stickiness, via trust: Users develop confidence in the platform's judgment, creating loyalty that transcends any individual model's performance
Network effects: Each new model integrated makes the platform more valuable for all users, while each new user type teaches the system to route more effectively
In venture terms, model builders are in the business of breeding faster horses. Orchestrators are in the business of training the steed, grooming the riders, and guiding them around the course. The long-term value isn’t tied to one animal; it’s tied to the stable and the program.
Where Orchestration Matters Most: Two Camps
The orchestration opportunity doesn’t look the same everywhere. It varies based on user sophistication and desired controls, which leads to a bifurcation into 2 distinct camps:
Camp 1: Orchestration-as-a-Service (Replacing the BPO)
In certain legacy industries, customers won’t care which models are used – or even that models are used at all. Here, orchestration platforms will own workflows end-to-end, often with humans in the loop, and deliver results as packaged services. Think orchestration as an outcome (“selling the work”), not a toolkit (“selling software”). These are AI-native service businesses (a huge labor capture opportunity) that replace traditional outsourcing.
A few opportunities that stand out:
Legal & Compliance: Firms drowning in repetitive tasks don’t want to 10x pre-litigation staff; they want work completed and handed back ready for settlement or litigation. Platforms that orchestrate models (and, where needed, human review) for intake, medical record retrieval, demand letter drafting, and case prep can deliver turnkey services, earning the right to expand across the org over time. With consumer-facing legal services generating $450B+ annually in the U.S., the opportunity is enormous.
Healthcare Administration: Providers need phones answered, referrals processed, schedules set, claims adjudicated, and prior authorizations approved before they can treat patients. Orchestration can unify OCR, data extraction, scheduling, and compliance checks into one service that automates these bottlenecks, replacing manual work often offshored to BPOs. With healthcare administration consuming over $800B annually, integration with clinical data will over time enable platforms to take on sequential units of work, compounding their value.
Financial Services Back Office: Loan processing, KYC, compliance checks, fraud detection, and underwriting remain highly manual, requiring armies of analysts. AI-native orchestration platforms can step in as contractors, weaving together document ingestion, identity verification, risk models, and regulatory workflows to deliver faster, cheaper, more reliable outcomes. Starting with narrow, high-value workflows opens the door to broader ownership across banks, fintechs, and insurers.
In these markets, the winners will replace entire BPO categories with AI-native alternatives. For industries long accustomed to outsourcing or staffing up, this won’t read as disruption so much as business-as-usual, just faster and smarter.
Camp 2: The Orchestration OS (Empowering the Expert)
For sophisticated users who want control over their creative or analytical process, orchestration platforms will become the operating system for professional work. The winners will be orchestration platforms that directly embed into professional workflows and become daily tools of record. A few opportunities that stand out:
Design & Creative Tools: Creative teams constantly juggle imagery, copy, video, and motion graphics. Modern orchestration embedded into design platforms will enable seamless switching across modalities – generating visuals, copy, and motion assets without leaving the canvas. With design today requiring countless hours of manual iteration across tools, orchestration becomes the natural AI-native extension, reducing grind while expanding creative output.
Marketing & Content: Marketers today manage fragmented workflows spanning campaign ideation, copywriting, SEO optimization, localization, and performance analytics. Orchestration platforms can integrate these steps, routing across models for each function (e.g., one for drafting, another for translation, another for analytics), then delivering back a cohesive campaign package that enables faster campaign cycles and easier to measure ROI.
Scientific Research: Scientists often synthesize literature, structure datasets, run statistical or simulation models, and visualize findings across disconnected tools. Modern orchestration can unify these tasks, automating literature review, proposing methods, executing analyses, and generating publish-ready figures in a single, interoperable platform. Instead of a “paper factory,” researchers get a sandbox that accelerates discovery while preserving flexibility and rigor.
In these markets, orchestration value accrues to the platform that becomes indispensable to daily work.
The Road Ahead
As agentic workflows mature, orchestration won’t stop at model switching. The best platforms will become the control layer for how work gets done – and ultimately, the system of record for how entire industries adapt to AI capabilities.
Every tech wave tempts us to chase star performers, but history shows value accrues to the connective tissue. AI will be no different. Models will keep getting faster and more specialized, but the real leverage lies with the rider (or platform!) that knows how to harness them all.
The stable masters, not the fastest horses, will capture the greatest returns.
If you’re as excited about the orchestration opportunity – we’d love to hear from you. Subscribe to this Substack or find me on LinkedIn for more!