I keep coming back to the same quiet doubt every time someone claims AI agents will flood public blockchains with endless activity and, somehow, endless value. It sounds clean on a slide deck. In practice the path from agent to lasting network revenue looks a lot messier. A recent research note from a major digital-asset team put six specific risks on the table, and reading them felt less like a forecast and more like a checklist of things the market is currently choosing to ignore.
Why the AI Agent Narrative May Not Deliver What Crypto Expects
The core idea is simple enough. Autonomous software agents will need to pay for services, settle trades, manage capital, and interact with one another. Public blockchains already offer programmable settlement and open access. Therefore agents should prefer them, transaction volumes should climb, and native tokens should benefit. That chain of reasoning has powered a lot of recent enthusiasm. The problem is that each link can break independently.
What follows is not a prediction that the entire thesis fails. It is a closer look at the six pressure points that could keep blockchains from converting agent activity into durable economic value. Some of these risks feel obvious once stated. Others only become clear when you separate adoption metrics from actual revenue that reaches token holders.
Closed Systems Already Offer Better Conditions for Agents
Perhaps the largest single risk is competition from controlled environments. Large technology firms, banks, payment networks, and fintech platforms are building agent-ready infrastructure that stays inside their own walls. These systems can deliver lower latency, predictable costs, smoother user experience, and clearer regulatory footing. They already possess merchant reach, identity frameworks, and the ability to extend credit. Public blockchains cannot simply assume that openness and programmable settlement will win by default.
An agent might settle a machine payment on a blockchain while handling identity checks, credit lines, and compliance through a bank or fintech rail. The research calls this possible future “multi-fi.” In that world agents use several types of infrastructure side by side. Public chains capture only the slices that truly need permissionless settlement. Everything else flows through platforms that already control distribution.
I have watched similar dynamics play out in earlier technology waves. Open protocols often lose the bulk of economic activity to closed products that feel easier for the average user or enterprise. Agents are not romantic about decentralization. They optimize for speed, cost, reliability, and legal certainty. If closed systems score higher on those dimensions, they will attract the majority of agent volume even while public chains remain available.
Even if AI drives a substantial increase in overall digital economic activity, there is no guarantee that public blockchains will capture a meaningful share of it.
That sentence captures the heart of the first risk. Activity can rise everywhere while the share landing on open networks stays modest. Investors focused only on total agent transaction counts may miss this distinction until it is too late.
Higher Payment Volume Does Not Automatically Mean Higher Token Value
The second risk sits downstream of the first. Suppose agents do use public blockchains for payments in large numbers. Transaction counts climb. Does that translate into meaningful fee revenue for the base layer and its native token? The answer is not automatic.
Stablecoin issuers and payment service providers sit closer to the user and can capture more of the economic surplus. Low fees and intense competition among chains make agent payments useful without necessarily making them a major income stream for token holders. Recent on-chain experiments already illustrate the gap. Agents have completed large numbers of payments while generating only tiny amounts of network fees relative to the transaction count.
Trading activity tells a different story. Over a recent 180-day window, trading produced many times more base-layer revenue per dollar of volume than payments did. Trading also creates maximal extractable value opportunities for validators. The contrast is instructive. Large numbers of small payments may keep a network busy without enriching its token in any durable way. Agents that manage capital—trading, lending, borrowing, providing liquidity—look far more promising from a value-capture perspective.
In my view this is one of the cleaner lessons in the research. Volume is not value. Markets that celebrate every new payment metric without asking who captures the margin risk celebrating noise. Networks that attract capital-managing agents rather than pure payment bots stand a better chance of converting activity into tokenholder returns.
Faster Software Development Weakens Technical Differentiation
AI coding tools are already changing how software gets written. Studies involving large numbers of developers show sharp increases in commit frequency and production releases. Blockchain applications can be built, tested, and deployed faster than before. On the surface that sounds like pure upside for the ecosystem. The third risk is that the same speed works against lasting competitive advantage.
When features become cheaper and faster to reproduce, pure technical novelty loses its edge. Networks and applications that once differentiated themselves through unique smart-contract designs or clever protocol mechanics may find those edges eroded quickly. Competitors can copy or modify similar tools with far less effort. Human oversight remains essential for security-critical financial software, yet the overall cost of producing functional code continues to fall.
Durable advantages therefore shift toward qualities that cannot be cloned as easily: liquidity depth, distribution reach, proven security track records, and institutional trust. Established networks that already possess those attributes may benefit precisely because the software layer becomes less distinctive. Newer projects that relied on technical novelty alone face a harder path.
I have seen this pattern in other software markets. Once the cost of building drops, the winners tend to be those who control the surrounding ecosystem rather than those who merely ship the next feature. Blockchain may follow a similar trajectory once agents accelerate development cycles further.
Security Threats Grow as Attack Costs Fall
AI does not only help builders. It also lowers the cost of finding vulnerabilities and mounting attacks. The fourth risk is that security stops being a baseline requirement and becomes a central competitive battleground. Networks and applications that cannot demonstrate robust defenses may lose agent traffic even if their other features look attractive.
Evidence already points in both directions. AI systems have identified genuine flaws in Ethereum-related software, including issues later assigned formal vulnerability identifiers. At the same time, human researchers still need to separate valid findings from convincing false positives. The net effect is an arms race. Attack surfaces expand while the tools available to both defenders and attackers improve.
For agents managing real capital, security is not optional. A single high-profile exploit can shift preferences toward more conservative infrastructure almost overnight. Public blockchains that treat security as an afterthought rather than a core product feature risk being sidelined precisely when agent activity begins to scale.
Perhaps the most interesting aspect here is the asymmetry. Building useful agent applications gets cheaper. Breaking them also gets cheaper. The institutions that succeed will be those that invest early and heavily in making the second activity expensive again.
Regulatory Friction Favors Permissioned Alternatives
The fifth risk is regulatory. Institutions that deploy or interact with agents often require clear identity controls, permissioning, and legal accountability. Fully permissionless networks can struggle to provide those guarantees. When an autonomous agent needs to interface with regulated financial services, the preference for systems that already satisfy compliance expectations becomes strong.
This does not mean public blockchains disappear. It does mean that certain high-value use cases may migrate toward hybrid or closed environments that can more easily demonstrate regulatory alignment. Agents operating at institutional scale are unlikely to ignore legal risk simply because a protocol is open.
Recent product launches already show the tension. Some platforms enable businesses to accept stablecoin payments from autonomous agents. Others develop competing or complementary systems that stay closer to traditional rails. The market is still testing which approach wins for different use cases. The regulatory overlay will shape those outcomes more than pure technical capability in many instances.
Value Capture May Concentrate Away from Base Layers
The sixth risk ties several of the previous points together. Even when agents use public blockchains, the economic value they create may concentrate at the application or service layer rather than at the base protocol. Stablecoin issuers, payment processors, specialized agent platforms, and liquidity providers can sit between the agent and the chain and extract most of the surplus.
Native tokens benefit most when the base layer itself captures meaningful fees or when demand for block space rises in a sustained way. If agent activity is dominated by low-fee payments or by applications that settle only occasionally, the base layer may see volume without corresponding revenue growth. Investors who treat every rise in agent transaction counts as a direct bullish signal for a given token may overestimate the eventual economic impact.
I find this distinction useful when evaluating claims about AI-driven demand. Ask first where the agent is deploying capital and who is collecting the margin. Networks that combine deep liquidity, strong distribution, proven security, and regulatory integration stand a better chance of converting agent activity into lasting demand for their native assets. Networks that offer only cheap settlement face a thinner path.
Putting the Six Risks Side by Side
Taken together the six risks form a coherent picture rather than a list of isolated concerns. Closed systems compete on performance and compliance. Payment volume may not translate into token value. Faster development erodes technical moats. Security becomes a sharper differentiator. Regulation pushes certain activity toward permissioned rails. And residual value may concentrate away from base layers even when agents do appear on-chain.
None of these outcomes is guaranteed. They are possible paths that depend on how agents, platforms, and regulators actually behave over the next few years. The research presents them as scenarios to watch rather than as firm predictions. That framing feels honest. Markets often treat any positive narrative as inevitable until evidence forces a reassessment.
One practical implication stands out. Investors and builders should focus less on total agent transaction counts and more on where agents choose to deploy meaningful capital. Capital-managing agents that trade, lend, borrow, and provide liquidity create different fee and security dynamics than pure payment agents. Tracking the former may give a clearer signal of durable demand.
What the Market Is Currently Overlooking
Much of the public conversation still treats the AI-agent-plus-blockchain combination as an almost automatic growth story. The risks outlined above suggest the story is more conditional. Openness alone does not guarantee share of activity. Volume alone does not guarantee revenue. Technical novelty alone does not guarantee lasting advantage. Security and regulatory posture matter more as agents handle real money. And the layer that captures value may not be the layer that processes the most transactions.
I have found that the most useful mental model is to treat public blockchains as one possible infrastructure choice among several rather than as the default destination for agent economic activity. That shift in framing changes which metrics look important and which claims look overstated. It also clarifies why some networks and applications may benefit disproportionately while others see limited upside even if overall agent usage rises.
The research does not argue that AI agents will avoid blockchains entirely. It argues that increased agent activity does not automatically produce lasting value for blockchain networks or their native tokens. That distinction is worth keeping in view as the technology continues to develop.
Practical Signals Worth Monitoring
Rather than waiting for a single definitive outcome, it is more productive to watch a set of observable signals. Where are agents currently choosing to settle payments versus manage capital? How much fee revenue reaches base layers relative to transaction counts? Are closed platforms winning the majority of enterprise agent deployments? How quickly are security incidents involving agent systems rising or falling? Do regulated institutions show a clear preference for permissioned or hybrid rails?
These questions do not require predicting the future. They simply require tracking the evidence as it appears. Networks that score well across liquidity, distribution, security, and regulatory integration will be better positioned if agent activity does scale. Networks that rely primarily on low-fee settlement or pure technical novelty face steeper odds.
In the end the AI agent thesis for crypto is not dead. It is simply more contingent than many current narratives suggest. The six risks identified by the research team provide a useful map of the contingencies. Paying attention to them does not require abandoning the opportunity. It does require measuring success by value captured rather than by activity observed.
That measurement difference may turn out to be the most important distinction of all.
Why Capital-Managing Agents Matter More Than Payment Agents
One theme that keeps resurfacing is the economic difference between agents that simply move small amounts of money and agents that actively manage portfolios of capital. The former can generate impressive transaction counts while contributing little to network revenue or maximal extractable value. The latter tend to interact more intensively with lending protocols, automated market makers, and trading venues. Those interactions create both fee income and opportunities for validators and liquidity providers.
If the bulk of future agent activity concentrates on micro-payments for services, the upside for native tokens may remain modest. If a meaningful share of agents instead takes on roles as automated traders, lenders, or liquidity managers, the revenue picture improves. Watching which type of agent gains traction first can therefore serve as an early indicator of whether the broader thesis is likely to deliver for token holders.
This is not a prediction that capital-managing agents will dominate. It is simply a recognition that their economic footprint looks different. Builders and investors who ignore the distinction risk misjudging the eventual impact of rising agent numbers.
The Multi-Fi Future and Its Implications
The concept of multi-fi deserves more attention than it currently receives. Agents will not be forced to choose a single infrastructure type for every task. They can route payments through one system, identity verification through another, credit through a third, and settlement through a fourth. In that environment public blockchains become one specialized tool rather than the universal settlement layer.
Such fragmentation is already visible in early product experiments. Some companies are building agent payment systems that span card networks, bank rails, and blockchain options simultaneously. Agents will select the rail that best matches the specific requirements of each transaction. Open networks that assume exclusive use will underperform relative to networks that position themselves as the best choice for particular high-value or high-programmability use cases.
I suspect multi-fi will become the default operating model for sophisticated agents. The implication is that pure “blockchain or nothing” thinking underestimates the competitive landscape. Public chains still have roles to play. Those roles may be narrower and more specialized than early narratives assumed.
Security as a Competitive Moat Rather Than a Checklist Item
As attack tools improve, the networks and applications that treat security as a continuous investment rather than a one-time audit will pull ahead. Agents that control real capital will prefer environments where the cost of successful exploitation remains high. That preference can create a self-reinforcing advantage for platforms that demonstrate strong defenses over time.
The same dynamic works in reverse. A high-profile agent-related exploit can shift capital and attention away from affected systems quickly. Because AI lowers the cost of both finding and exploiting weaknesses, the window between vulnerability discovery and potential damage may shrink. Continuous monitoring, formal verification where practical, and rapid response capabilities become more valuable under those conditions.
Security therefore moves from a cost center to a potential source of differentiation. Networks that can credibly claim superior resilience may attract a disproportionate share of capital-managing agents even if their pure technical features are comparable to competitors.
Regulatory Integration as a Quiet Advantage
Permissionless design remains a core strength of public blockchains. It also creates friction when agents need to interact with regulated entities. Identity, accountability, and auditability requirements do not disappear simply because software is autonomous. Systems that can satisfy those requirements while still offering useful programmability may capture institutional agent activity that pure permissionless networks struggle to reach.
Hybrid approaches already appear in the market. Some platforms combine on-chain settlement with off-chain identity and compliance layers. Others keep the entire flow inside more controlled environments. The eventual mix will depend on how regulators treat autonomous agents and on how much legal risk institutions are willing to accept. In the meantime, networks that invest in regulatory compatibility gain an option that pure technical openness does not automatically provide.
This does not mean every public chain must become permissioned. It does mean that the ability to interface cleanly with regulated systems becomes a practical advantage for certain use cases. Ignoring that reality underestimates how capital actually moves in the real world.
Measuring Success by Value Rather Than Volume
The final practical takeaway is methodological. Metrics that count agent transactions or total payment volume capture only part of the story. Metrics that track fee revenue reaching the base layer, capital locked by agents in productive protocols, and the share of agent activity that involves active capital management tell a more complete picture. Focusing on the latter set of indicators reduces the risk of mistaking busy networks for valuable ones.
Markets love simple narratives. The AI-agent-plus-blockchain story is simple and appealing. The six risks outlined in the research complicate that story in useful ways. They do not destroy the opportunity. They clarify the conditions under which the opportunity becomes real for token holders and network participants.
Staying attentive to those conditions is less exciting than celebrating every new agent metric. It is also more likely to produce accurate assessments as the technology matures. In a space that has repeatedly overestimated the speed and certainty of value capture, that kind of attention remains valuable.
The next phase of agent development will show which of these risks prove binding and which prove manageable. Until clearer evidence arrives, treating the six points as live possibilities rather than distant hypotheticals seems the more prudent stance.
