Skip to content

The Political Economy of Moats

We say "moat" and imagine water, stone, permanence. Architecture designed to repel. But medieval sieges tell a different story. Historians estimate that three-quarters of sieges succeeded, and more through negotiation or starvation than assault. Fortresses fell to betrayal, bribery, and coercion at least as often as to military force. In 1119, Louis VI of France bribed the castellan of Les Andelys to smuggle soldiers inside hidden under straw; the "impregnable" fortress fell without a blow struck. Under accepted rules of war, a town captured by force could be sacked, but one that surrendered could not. Surrender was economically rational for everyone except the ideologically committed.

The drawbridge was always the point. Moats weren't designed to prevent entry; they were designed to make entry expensive enough that negotiation became preferable to assault. The moat's function was never architectural but political-economic. Every castle had a price. The question was whether attackers could pay it.

This week, three drawbridges lowered.

Disney licensed 200 characters to OpenAI's Sora for $1 billion plus equity. Mickey Mouse, Darth Vader, Iron Man—a century of IP fortress-building converted to platform positioning in a single transaction. Meta announced its pivot from open-source Llama to closed proprietary "Avocado," abandoning the strategy that made it the center of open AI development. And Anthropic's Claude Skills feature invited companies to "package your procedures, best practices, and institutional knowledge" into the platform—the capability moats that vertical SaaS companies believe make them defensible.

Three moats. Three prices paid. The water didn't drain; it converted to currency.

What we're witnessing isn't moat failure. It's moat-liquefaction: the transformation of defensive barriers into transactional surfaces. A liquid moat doesn't breach; it trades. The castle doesn't fall; it renegotiates. And the question that medieval castellans understood perfectly—whose incentives govern the drawbridge?—turns out to be the only strategic question that matters.

The standard strategy discourse treats moats as complicated engineering problems. Analyze the competitive landscape. Identify defensible positions. Build barriers to entry. Maintain them against attackers. The metaphor suggests solidity, permanence, architecture you construct once and reinforce periodically.

But moats aren't complicated systems; they're complex ones. The distinction matters. In complicated systems, cause and effect are knowable, and expert analysis can identify optimal strategies. In complex systems, cause and effect are visible only in retrospect, and the system's behavior emerges from interactions that no analysis fully captures. Complicated systems reward planning. Complex systems reward adaptation.

Medieval castle defense was complex, not complicated. The structural vulnerability wasn't in the walls but in the sociology. A lord with hereditary claim might hold out indefinitely. A mercenary captain with payroll to meet would negotiate. The garrison's incentives mattered more than the garrison's architecture. Castles fell when the people inside had more to gain from surrender than from defense.

Modern corporations have the same structure. Executives with quarterly targets. Boards with fiduciary duties to maximize shareholder value. The moat is maintained by people whose time horizons are short and whose incentives respond to capital offers. Offer $100 million, and corporate antibodies—the organizational immune system that kills anything different—develop sudden tolerance. Offer $1 billion, and the drawbridge lowers itself.

The political economy of moats is the study of who controls the drawbridge and what makes them open it. Or, more crisply: moats aren't walls; they're governed prices.

Disney: The Tollbooth⚓︎

Consider what Disney actually did.

The deal gives Sora users access to Mickey Mouse, Ariel, Cinderella, Iron Man, Darth Vader, and roughly 195 other characters across Disney, Marvel, Pixar, and Star Wars. Users can generate short social videos featuring these characters starting in early 2026. The deal explicitly excludes training data access—Disney's IP won't train OpenAI's models—and excludes talent likenesses and voices. Disney retains a steering committee for content moderation. The moat didn't fall; it became a tollbooth.

But look at the structure more closely. Disney negotiated only about one year of exclusivity with OpenAI; after that, Disney can license the same characters to other AI platforms. This isn't a surrender—it's price discovery. The first deal establishes the market rate; subsequent deals are pure upside against a proven model. Disney is positioning itself as the canonical source that all platforms must negotiate with, not a captive supplier to one.

Iger's framing was explicit: "We'd rather participate in the rather dramatic growth, rather than just watching it happen and essentially being disrupted by it." The IP could be vulnerable either way if generative AI becomes as powerful as its evangelists claim. For $1 billion—a bargain for a company Disney's size—Iger bought a seat at the table. CNN called it "a $1 billion hedge on the future of slop." The hedge framing is precise: Disney isn't confident AI will transform entertainment, but can't afford to be wrong.

The surgical precision matters too. Characters can appear on Sora; they can't be. No voices, no likenesses—the elements that depend on actor relationships and human talent remain behind the walls. Disney is liquefying the fungible parts of the moat (visual IP, character designs) while retaining the parts that remain defensible (the human relationships, the voices, the performances). This isn't wholesale capitulation; it's selective liquefaction of what can't be held anyway while reinforcing what still can.

And on the same day Disney signed the OpenAI deal, it sent a cease-and-desist letter to Google for allegedly using Disney IP to train models without permission. The moat becomes a tollbooth with legal enforcement for non-payers. License to partners who pay; litigate against those who don't. This is how you run a liquefied moat.

Except: follow the system effects.

Disney is already outsourcing animation to Vancouver and Manila, with Moana 2 as the first feature split between Los Angeles and offshore facilities. The Animation Guild commissioned research predicting 21% of US film, TV, and animation jobs consolidated, replaced, or eliminated by AI by 2026—over 100,000 positions. DreamWorks co-founder Jeffrey Katzenberg claims that animated features requiring 500 artists over five years will require 10% of that within three years. VFX industry observers predict workforce reductions of 80% or more once AI systems mature.

Disney's moat-liquefaction isn't just a financial transaction but a system reconfiguration. The $1 billion flows from OpenAI to Disney; Disney flows from content creator to IP licensor; the workforce flows from employment to redundancy; the creative pipeline flows from human-driven to AI-augmented with human cleanup. Capital moves in one direction, jobs in another, capability from studio to platform.

Using Simon Wardley's evolution framework: Disney's animation capability is moving from custom-built (proprietary, differentiated, expensive) toward commodity (platform-available, purchasable, price-competitive). The moat around capability dissolves. The moat around IP transforms into licensing revenue. Disney becomes the canonical source—the node that AI-generated content references—rather than the exclusive producer of what the characters do.

The question isn't whether this is good or bad for Disney shareholders. In the short term, it's probably good: $1 billion in equity, a customer relationship with OpenAI, positioning in AI-generated content distribution. The question is what it means for the system Disney inhabits. The workers losing jobs. The junior roles eliminated before new animators learn the craft. The pipeline changes that determine what kind of content becomes possible and what kind becomes uneconomical. The moat-holder's rational decision is the system's structural transformation.

Meta: The Self-Emptying Moat⚓︎

Meta's case illustrates a different liquefaction pattern: the self-emptying moat.

Open-source Llama was a strategic masterstroke—or looked like one. Release the model weights. Let the ecosystem build on your architecture. Capture value through position rather than enclosure. The logic was sound: become the center of gravity for open AI development, and you control the standard even without controlling access.

Then DeepSeek demonstrated that openness in AI is invitation, not deterrent.

The Chinese lab copied Llama's architecture, trained on different data, and built a competitive model. The gift that was supposed to create dependency created competitors instead. Meta's open-city strategy—giving away the walls, betting that generosity creates position—collapsed when the gift turned out to be valuable enough to accept gratefully and use against the giver.

The Avocado pivot is capitulation to moat logic. After DeepSeek's success, after the $14.3 billion acquisition of Scale AI's team, after Yann LeCun's resignation and 600 layoffs from the Fundamental AI Research lab, Meta learned that architectural moats self-empty when the architecture is valuable enough to copy. The company that championed open AI development will release its next frontier model as closed, paid, permission-controlled.

The system effect is concentration. Meta's departure from open-source narrows AI development to fewer Western labs. The ecosystem that was supposed to democratize capability now routes through OpenAI, Anthropic, Google, and a closed Meta. Capital consolidates; access becomes permission-gated; the moat that gave itself away reconsolidates as a gated community.

Skills: Capability Migration⚓︎

Anthropic's Skills feature reveals a third pattern: capability migration.

The marketing language is instructive: "Package your company's procedures, best practices, and institutional knowledge so teams work consistently and new members get expert-level results from day one." This is the value proposition of vertical SaaS—domain-specific workflows, procedural expertise, the accumulated knowledge that makes specialists more effective than generalists. The capability moat that vertical software companies believe makes them defensible against horizontal platforms.

But Skills suggests capability moats are migrating upward. What lives in your SaaS application today—the workflows, the domain logic, the procedural knowledge—can be packaged, uploaded, and executed by the platform. The moat doesn't breach through competition; it evaporates into the platform layer.

Model Context Protocol already lets ChatGPT access HubSpot data directly. The next step is accessing your workflows directly. The platform doesn't need to attack your moat; it needs to make moats hostable. And once the moat is hosted, the moat-holder becomes the Skills provider—still capturing some value, but from a fundamentally different position.

The Mechanism: Legibility to Liquidity⚓︎

Wardley's framework helps clarify what's happening. Components evolve along a predictable axis: genesis (novel, uncertain, expensive) to custom-built (understood, differentiated) to product (stable, competitive) to commodity (interchangeable, price-driven). Strategic advantage lives in genesis and custom-built. Commodity components are costs, not differentiators.

AI accelerates evolution. Capabilities that took decades to commoditize now move in years. Disney's animation expertise, Meta's architectural advantage, vertical SaaS domain knowledge—all are moving rightward on the evolution axis faster than the moat-holders anticipated.

More precisely: AI makes moat contents legible to capital. What was opaque (proprietary process, institutional knowledge, craft expertise) becomes inspectable, replicable, packageable. But legibility alone isn't enough. What converts legibility into liquidity is standardization—APIs, evaluation frameworks, connectors like MCP, compliance schemas, audit trails. Skills isn't just "upload your knowledge"; it's knowledge packaged in a platform-native execution format. The chain runs: legibility → standardization → portability → price discovery → bargaining. Once capital can see inside the moat and the contents can flow through standard interfaces, the drawbridge negotiation begins.

The non-kinetic warfare framing sharpens this. Wardley has described LLMs as "a non-kinetic form of warfare designed to embed the values of a small number of people into much wider communities by capturing the process of decision making." The attack vector isn't breaching moats; it's making moat-holders comfortable. Partnership feels like opportunity. Licensing feels like revenue. The payload is delivered through helpfulness.

Disney licensing to OpenAI normalizes AI-generated Disney content. The moat-holder becomes the legitimizer. Meta closing Llama concentrates AI development. The former champion of openness becomes a gatekeeper. Claude's Skills captures procedural knowledge. The platform becomes the host for workflows that used to be moats.

Each transaction is individually rational. Each executive makes a defensible decision. The system effect is consolidation, dependency, and capability migration—from moat-holders to platforms. The attackers didn't storm the walls. They made offers the drawbridge-controllers couldn't refuse.

Vertical SaaS: The Durability Question⚓︎

What does this mean for vertical SaaS?

The companies believe they have moats: domain-specific data, workflow expertise, customer relationships, regulatory knowledge. HR software for aged care. Rostering systems for education. Compliance tools for financial services. The horizontal platforms can't compete because they don't understand the domain.

The question moat-liquefaction raises: are these moats durable, or merely not yet interesting to capital?

If capital wants what's inside, your moat becomes a negotiation. OpenAI or Anthropic approaches with partnership offers. "Let us access your domain data to improve our models for your vertical. Let us host your workflows as Skills. Let us be the AI layer; you be the domain expertise." Each offer is individually attractive. Each acceptance transfers capability from moat to platform.

If capital doesn't want what's inside, your moat persists—but possibly because it's too niche to bother acquiring. Australian workforce rostering for aged care isn't valuable enough for a billion-dollar offer. This might be the good news. The moats that survive aren't the most defended; they're the least interesting.

Map your moat components on Wardley's evolution axis. Where are they? If they're in product moving toward commodity, the platform will absorb them. If they're in genesis or custom-built, you have time—but time converts to commodity eventually. The clock runs faster now.

The strategic frame matters too. Most vertical SaaS companies treat their situation as complicated—analyzable, defensible through planning, amenable to best practices. But the situation is complex, characterized by feedback loops, emergent effects, and second-order consequences visible only in retrospect. The workforce management domain that exists today (manual rostering, compliance rules, award interpretation) may be substantially different in three years. The moat you're defending may not be the territory that matters.

Scenario Sketches: Lived Experience⚓︎

What does this mean for lived experience? Not prediction but projection—scenario sketches that stress-test incentives rather than forecast timelines. Following the dynamics forward to see what work and watching might become if these patterns play out.

Consider a facilities manager at an aged care provider in 2028. She used to open Workforce Manager Pro (the vertical SaaS her organization has used for a decade) to build rosters, check compliance, manage leave requests. The software knew the aged care award inside out. It understood minimum staffing ratios, qualification requirements, the baroque complexity of penalty rates.

Now she opens Claude. Or ChatGPT. Whatever the dominant interface has become. She says: "Build next week's roster for Sunrise Lodge. We're short two RNs on Thursday night." The platform consults Skills—aged care compliance packaged by some vendor, maybe Workforce Manager Pro, maybe a competitor, maybe an aggregator who licensed from multiple sources. The platform accesses her staff database through MCP. It knows the award because someone uploaded the procedural knowledge. The roster appears.

The facilities manager doesn't care whose moat the knowledge came from. She cares that the roster is compliant and the night shift is covered. The domain expertise that was Workforce Manager Pro's moat has liquefied into the platform layer. The vendor might still exist—as a Skills provider, as a compliance certifier, as the entity that validates outputs against regulatory standards. But the relationship has transformed. The moat became an API.

The HR director at a mid-sized education provider experiences something similar. She used to rely on EduStaff (a vertical SaaS for education workforce management) because it understood teacher registration requirements, working with children checks, the specific leave provisions in the enterprise agreement. EduStaff's sales pitch was: "We know education. The horizontal tools don't."

In 2028, the horizontal tools do know education—because EduStaff licensed its procedural knowledge to survive. Or because a competitor did. Or because the platform scraped enough documentation and edge cases to approximate the expertise. The HR director's experience hasn't changed much; she still gets compliant rosters, still manages leave, still runs reports. But the value capture has shifted. She pays less for EduStaff (now a thin integration layer) and more for the platform (which hosts the intelligence). EduStaff's moat liquefied; the company persists as a node rather than a fortress.

The worker's experience changes differently. The aged care nurse checking her shifts used to interact with Workforce Manager Pro through a clunky mobile app built by a 50-person vertical SaaS company. In 2028, she asks her phone about next week's shifts, and the platform answers—pulling from whatever compliance logic and roster data lives in the Skills layer. The interface is better (platform-grade UX, not vertical-SaaS-grade). The intelligence is comparable or better.

What's missing is harder to name: the specific relationship between her employer and the vendor, the vendor's rep who understood their organization, the customizations built over years, the institutional knowledge in the support team. The liquefied moat still works—rosters get built, compliance gets checked—but something has shifted in the texture of work. The domain-specific vendor, with domain-specific knowledge and domain-specific relationships, became an interchangeable commodity supplier to a platform the worker never chose and may not trust.

Project the Disney case forward. What does moat-liquefaction mean for watching Disney content?

2025: Disney licenses 200+ characters to Sora. Users generate short social videos featuring Mickey, Darth Vader, Iron Man. The content is canonically licensed but user-created. Disney retains a steering committee to prevent brand damage.

2027: Sora-generated Disney content saturates social media. A viral video of Darth Vader singing karaoke, created by a teenager in Buenos Aires using the licensed Sora tool, accumulates 400 million views. Disney didn't make it. Disney licensed the possibility of it. Disney captures some value through the platform relationship but not the cultural moment.

2028: Disney releases a new animated feature. It was animated in Vancouver and Manila, with AI handling interpolation, cleanup, and increasingly, "creative suggestions" that the reduced human team reviews and approves. The credits list 47 human animators. Five years earlier, the same production would have listed 300. The movie looks like a Disney movie. It moves like a Disney movie. Whether it feels like a Disney movie depends on whether you know how it was made.

The Animation Guild's prediction has largely materialized: 21% of US animation jobs consolidated or eliminated, distributed unevenly. Junior roles—the entry points where young animators learned the craft—were hit hardest. AI handles the work that used to train the next generation. The senior animators who remain are older, supervising AI output rather than creating frames. The pipeline has changed.

For the viewer, the immediate experience might be: the movies are fine. Maybe slightly different in ways hard to articulate. A certain efficiency in the movement. A subtle uncanny valley not in the faces but in the choices—AI-generated "creative suggestions" that are competent but not quite surprising in the way human choices sometimes surprise. The Disney moat has liquefied into brand (canonical source), distribution (Disney+), and IP (characters that Sora users generate content with). The animation capability moat largely evaporated.

For the aspiring animator, the experience is: the path closed. The junior roles that were entry points don't exist at the scale they used to. You can generate impressive animation with AI tools—more impressive than any student reel from 2020—but generation doesn't teach the craft. The expertise moat liquefied, and the pipeline for creating new experts liquefied with it.

For the culture, the experience is harder to summarize. More Disney content exists than ever before (user-generated, AI-assisted, officially released). The canonical versions remain canonical. But the density of Disney imagery in the environment has increased while the human labor creating it has decreased. The characters are omnipresent; the humans who brought them to life are fewer.

The workforce management market in 2029:

The landscape has consolidated. Three or four platforms host most of the intelligence layer. Dozens of vertical Skills providers supply domain expertise—aged care compliance, education workforce rules, healthcare registration requirements. The Skills providers compete on comprehensiveness, accuracy, update speed (regulations change; Skills packages must change with them). The platforms compete on UX, integration breadth, enterprise features.

The workers using these systems experience a strange homogenization. The aged care facilities manager, the education HR director, the healthcare clinic owner—they all use the same platform with different Skills attached. The workflows feel similar. The interfaces are consistent. The domain-specific differences are real but hidden in the Skills layer, invisible to the user unless something goes wrong.

What's been gained: ease of use, interface quality, the ability to ask natural language questions rather than navigating software designed by engineers who've never worked in aged care. What's been lost: the specific relationship between a vertical SaaS vendor and its industry—the vendor who understood aged care because their team came from aged care, the support rep who'd seen your specific edge case before, the roadmap shaped by domain customers rather than platform economics. The moat was also a relationship, and liquefaction dissolved the relationship along with the barrier.

The Moat-to-Node Transition⚓︎

Historical parallel: castles didn't disappear when gunpowder made walls obsolete. They became manor houses, administrative centers, symbols of legitimacy. The defensive function died; the positional function transformed. The castle that couldn't stop cannons could still anchor a territory, signal authority, attract visitors.

Moat-liquefaction doesn't end moats. It transforms them.

Disney's IP fortress becomes the canonical source. AI-generated Mickey Mouse proliferates regardless of licensing. Disney's position shifts from preventing generation to legitimizing it. The moat becomes a brand node. Everyone generating Disney content references Disney as origin; Disney captures value from reference rather than exclusion. The 2028 viewer watches a Disney movie and sees the brand; they don't see (and perhaps don't care about) the liquefied animation capability beneath it.

Meta's open-source position becomes the gated community. Avocado is closed, paid, permission-controlled. Meta moves from ecosystem center to access gatekeeper. The moat transforms from generosity-as-strategy to scarcity-as-strategy. The 2028 developer navigating frontier models finds a landscape of three or four major providers, each with their own terms, their own values embedded in their systems.

Vertical SaaS capability moats become integration surfaces. Domain expertise doesn't defend against platforms; it qualifies for platform integration. The moat transforms from wall to API. The 2029 compliance officer or clinic owner or facilities manager experiences this as: the software works better, but the relationship is different. The vendor is still there—as a Skills provider, a compliance certifier, an integration specialist—but the locus of value has shifted to the platform.

This is the moat-to-node transition: when defensive barriers transform into network positions, and control shifts from exclusion to mediation. Disney mediates between AI generation and canonical characters. Meta mediates between developers and model access. Vertical SaaS mediates between platforms and domain expertise.

The texture of mediation differs from the texture of defense. A fortress-holder's relationship to their territory is protective, controlling, autonomous. A node's relationship to its network is participatory, dependent, embedded. Disney-as-fortress decided what Disney content existed. Disney-as-node legitimizes Disney content that others generate. Vertical-SaaS-as-fortress decided how domain workflows worked. Vertical-SaaS-as-node packages domain knowledge for platforms to deploy.

The ILC Trap⚓︎

For vertical SaaS, your moat is liquid. The question isn't whether—it's when and into what.

The instinct is to list options: become a node, exploit your niche, race the liquefaction, accept the transition. These are the standard responses to platform pressure. But Wardley's ILC framework (Innovate-Leverage-Commoditize) suggests something more uncomfortable: these aren't strategies for durable advantage. They're coping mechanisms for inevitable decline.

The ILC playbook is simple and relentless. Take something that's a product, turn it into a utility, let everyone build on top of it, mine the metadata to spot future patterns, commoditize those patterns into new component services, move up the stack. Amazon runs this playbook. The AI platforms are running it now. Vertical SaaS domain expertise is exactly the kind of thing that gets fed into the ILC machine. The vertical player does the innovation; the platform does the leverage and commoditization.

Run the options through this lens:

Become a node means accepting ILC. You become the Skills provider, a component supplier to a platform running the ILC playbook on you. Your margins compress over time. Your differentiation erodes as the platform learns from your patterns and the patterns of every other vertical provider. This isn't durable advantage; it's managed decline with a seat at the table.

Exploit the niche works only if you're beneath the platform's notice. Australian aged care rostering might qualify today. But this isn't durable advantage; it's durable irrelevance. The moment your niche becomes interesting—aggregated with other niches, valuable for training data, strategic for platform expansion into adjacent markets—you're back in the ILC crosshairs. You survive by hoping no one ever wants what you have.

Race the liquefaction means staying in genesis and custom-built while the platforms commoditize everything behind you. But the treadmill speeds up. Each innovation you create gets commoditized faster than the last as the platforms get better at pattern recognition across their installed base. This isn't a strategy; it's a hamster wheel with a dignified name.

Accept the transition is the same as becoming a node, just more honest about the trajectory. You package your domain expertise, license it to platforms, and watch the revenue model compress. The knowledge persists; the company shrinks. Transformation, like the castle becoming a manor house—but manor houses don't command armies.

These are coping strategies, not durable advantages. So what might actually work?

Before answering, an honest acknowledgment: I've underweighted some durable-ish moats that vertical SaaS often possesses. Distribution moats matter—approved vendor lists, multi-year procurement contracts, implementation partners, change management fatigue. Enterprises don't switch systems easily, and platforms haven't yet built the sales motion to displace embedded vertical players. Data rights matter differently than "we have data"; permissioned, longitudinal, high-integrity data with specific usage rights is harder to replicate than it sounds. Certification ecosystems matter—"we are the system auditors already accept and insurers recognize" creates friction that pure capability can't overcome.

These don't defeat the liquefaction thesis. They can liquefy too, given enough platform patience and enough capital to build enterprise sales capacity. But acknowledging them makes the timeline less certain and the outcome less predetermined. The question isn't whether moats liquefy but which ones and how fast.

What Might Actually Work⚓︎

Three possibilities, each requiring vertical SaaS to become something fundamentally different than what it is today.

Own the liability, not just the logic. Platforms can commoditize aged care rostering knowledge through Skills. They can't commoditize being the entity that's liable when the roster is wrong and a patient is harmed. They can't hold the regulatory certification. They can't be the party that AHPRA or ASIC holds accountable when something goes wrong.

Vertical SaaS that embeds itself in compliance liability and regulatory relationships—not just compliance knowledge—might have something more durable. The moat isn't the code or the domain expertise; it's the position in the regulatory graph. This requires becoming less of a software company and more of a compliance infrastructure provider that happens to use software.

But liability is a product, not a slogan. Owning it requires contractual structure (indemnities that platforms won't accept), insurance relationships (carriers who understand your domain and will underwrite your exposure), auditability (versioned rules, explainable outputs, incident logs), and operational capacity to respond when something goes wrong. The transformation is significant: different revenue model, different risk profile, different organizational capabilities—and a credible claim that when regulators come asking, you're the entity with answers.

Most vertical SaaS companies won't make this shift. The ones that do become harder to absorb because the platform would have to absorb the liability along with the capability—and platforms are structured to avoid liability, not accumulate it.

Become the vertical platform before the horizontal platform notices. "Too big to absorb" is relative. Big in one vertical is still tiny compared to horizontal platforms. But there's a window: consolidate your vertical fast enough to become the platform for that vertical before a horizontal platform decides to enter.

This means running ILC yourself, within your vertical, before it's run on you. Commoditize the smaller vertical players. Build the platform layer for aged care or education or financial services. Become the thing that horizontal platforms have to integrate with rather than replace—because you've already absorbed the ecosystem they would need to build.

This is a race condition. It requires capital, speed, and willingness to cannibalize your own product business to build a platform business. It requires being more aggressive about ILC within your domain than the horizontal platforms are about ILC across domains. Most vertical SaaS companies don't have the capital, the speed, or the stomach. The few that do become the consolidators rather than the consolidated.

Generate novelty faster than commoditization erodes it. This is different from racing the liquefaction. Racing means staying in genesis—an exhausting treadmill. This move accepts that lower layers will be commoditized while continuously generating new value at higher layers.

Amazon commoditized cloud infrastructure with AWS. But companies building on AWS aren't all commoditized—some built genuinely novel things on top that Amazon couldn't anticipate. The question for vertical SaaS: can you keep generating novelty at the layer above what platforms commoditize?

For workforce management, this might mean: let the platform commoditize rostering logic. Build something new on top that the platform can't see yet—predictive workforce analytics that require domain context the platform lacks, regulatory arbitrage tools that depend on relationships the platform can't replicate, capabilities that don't exist yet and won't exist until you create them. Accept that each layer you build will eventually be commoditized. Keep climbing.

This requires genuine innovation capacity, not just domain expertise. Most vertical SaaS companies have accumulated domain knowledge; few have built the R&D capability to stay ahead of platform absorption. The ones that do become moving targets rather than sitting ducks.

There's a fourth possibility, visible in Disney's deal structure: become the price-setter, not just a node. Disney negotiated one year of exclusivity with OpenAI, then freedom to license the same IP to competing platforms. The first deal establishes market rates; subsequent deals are upside against a proven model. Disney isn't a captive supplier—it's the canonical source that all platforms must negotiate with.

For vertical SaaS, this would mean: license your domain expertise to one platform first, but negotiate short exclusivity windows. Use the first deal to establish that your compliance logic or workflow knowledge has market value. Then license to the second platform, and the third. Become the entity that all horizontal platforms integrate with for your domain, not the exclusive partner of one.

This requires bargaining power most vertical SaaS companies lack. Disney can negotiate short exclusivity because Disney is Disney—the characters are irreplaceable. Most domain expertise isn't irreplaceable; it's approximable. But for the few vertical players with genuinely unique data, regulatory relationships, or accumulated edge cases that platforms can't easily replicate, the Disney playbook offers a model: selective liquefaction on your terms, with diversification across platforms and legal enforcement against non-payers.

The Honest Assessment⚓︎

The honest assessment: there may not be durable advantage for vertical SaaS as currently constituted. The ILC dynamic is real and probably inexorable for companies that remain primarily software businesses selling domain expertise.

The three paths that might offer genuine durability all require becoming something different: compliance infrastructure, vertical platform, or innovation engine. Each is a significant transformation. Each requires capabilities most vertical SaaS companies don't currently have. Each involves risk that most boards and executives won't accept.

What remains for the rest? Coping strategies. Managed decline. Negotiated surrender terms. The hope that liquefaction happens slowly enough to exit before it completes.

This isn't nihilism; it's clarity. The medieval castellan who understood that every castle had a price wasn't defeatist—he was realistic about the nature of defense. The vertical SaaS executive who understands ILC isn't giving up—she's seeing the board clearly enough to make an actual choice rather than a comfortable illusion.

The choice isn't between defense and surrender. It's between transformation and decline. The moat liquefies either way. The question is whether you use the time to become something the liquid can't dissolve.

The Liquid Fortress⚓︎

Moats were never walls; they were negotiated spaces with drawbridges for a reason. The medieval castellan understood that defense was always priced—the question was whether attackers would pay. What AI accelerates is the rate of liquefaction: how quickly defensive barriers convert into transactional surfaces, how visibly capital can see inside the moat, how easily partnership offers align with executive incentives.

December 2025 was a week of keys turning in locks. Disney's $1 billion handshake with OpenAI. Meta's retreat from openness. Claude's invitation to package your procedures for the platform. The drawbridges lowered; the water began to flow.

Looking back from 2029, what we notice isn't that the moats disappeared. They didn't. Disney is still Disney; the brand moat is stronger than ever. Meta still matters; the gatekeeper position commands value. Vertical SaaS still exists; the domain expertise still has buyers. What disappeared was the illusion that moats are architectural—that they're walls you build once and maintain, rather than contracts you renegotiate every time someone with capital asks nicely enough.

The liquid fortress doesn't drain. It flows into new shapes, carries value in new directions, deposits it in new hands. The medieval lesson holds: castles fell to incentive alignment, not superior siege engines. The modern corollary: moats liquefy when capital offers make sense to the people holding the keys.

The question was never "will the moat hold?"

The question was always "who benefits when it doesn't?"

Sources⚓︎

  • English Heritage, "10 Things You Probably Didn't Know About Sieges": https://www.english-heritage.org.uk/visit/inspire-me/10-things-sieges/
  • The History Press, "Siege warfare in the Middle Ages": https://thehistorypress.co.uk/article/siege-warfare-in-the-middle-ages/
  • OpenAI, "Disney Sora Agreement" (December 2025): https://openai.com/index/disney-sora-agreement/
  • CNBC, "Disney making $1 billion investment in OpenAI" (December 2025): https://www.cnbc.com/2025/12/11/disney-openai-sora-characters-video.html
  • CNBC, "Disney's OpenAI stake is 'a way in' to AI" (December 2025): https://www.cnbc.com/2025/12/11/disney-open-ai-iger-altman.html
  • CNN Business, "Disney's OpenAI deal is a $1 billion hedge on the future of slop" (December 2025): https://www.cnn.com/2025/12/11/business/disney-openai-hedge
  • Fortune, "Bob Iger says Disney's $1 billion deal with OpenAI is an 'opportunity, not a threat'" (December 2025): https://fortune.com/2025/12/11/disney-openai-deal-investment-bob-iger-opportunity-not-threat/
  • CNBC, "From Llamas to Avocados: Meta's shifting AI strategy" (December 2025): https://www.cnbc.com/2025/12/09/meta-avocado-ai-strategy-issues.html
  • OpenAI, "Introducing GPT-5.2" (December 2025): https://openai.com/index/introducing-gpt-5-2/
  • Context News, "Hollywood animation, VFX unions fight AI job cut threat": https://www.context.news/ai/hollywood-animation-vfx-unions-fight-ai-job-cut-threat
  • The Wrap, "An AI Wave Will Sweep Through Hollywood's VFX Systems in 2025": https://www.thewrap.com/ai-vfx-production-labor/
  • Screen Daily, "Why outsourcing and AI means the US animation sector is facing hefty challenges in 2025": https://www.screendaily.com/features/why-outsourcing-and-ai-means-the-us-animation-sector-is-facing-hefty-challenges-in-2025/5199989.article
  • SaaStr, "10 Ways to Build a Moat in SaaS. But AI is Also Making Them Weaker": https://www.saastr.com/whats-your-moat/
  • Vendep Capital, "Forget the data moat: The workflow is your fortress in vertical SaaS": https://www.vendep.com/post/forget-the-data-moat-the-workflow-is-your-fortress-in-vertical-saas
  • Wardley Maps, "Mapping 101: A Beginner's Guide": https://www.wardleymaps.com/guides/wardley-mapping-101
  • Wardley Maps, "ILC (Innovate-Leverage-Commoditize)": https://www.wardleymaps.com/glossary/ilc
  • The Cynefin Company, "About the Cynefin Framework": https://thecynefin.co/about-us/about-cynefin-framework/
  • Simon Wardley on LLMs as non-kinetic warfare, LinkedIn (November 2025)