A structural and crucial transition is underway in DeFi vault curation.
Sentora's position, developed over the past several months and formalised in our recent research paper, is that the protocols and vaults best positioned for the next market cycle are those that have solved a specific architectural problem: converting institutional investment policy into agentic, continuous risk execution.
When Financial Infrastructure Outpaced Its Operators
Traditional capital markets, for all their complexity, evolved at a pace that human organisations could track. Risk committees had time to deliberate, portfolio managers could review overnight positions, and the gap between a market event and a meaningful human response was tolerable because market structure was designed around that assumption.
DeFi is changing that forever.
The programmatic, always-on nature of DeFi markets changes the risk dynamic fundamentally, and within DeFi, the shift away from unified liquidity pools toward isolated, permissionless lending markets has multiplied the surface area that any curator must cover. Protocols like Morpho, Euler, and Kamino introduced a level of risk granularity the previous architecture could not express. Each market carries its own collateral configuration, oracle dependency, utilisation dynamics, and liquidation behaviour. Capital flowing across this lattice generates a continuous stream of interdependent signals, all of which bear on the safety and performance of a given vault.
As a consequence, curation has become a fundamentally different problem. What was once a periodic allocation decision is now a continuous optimization across a high-dimensional, rapidly shifting state space. Human judgment remains the appropriate tool for defining strategy and setting risk tolerance. Yet, it is no longer fit for executing against those parameters in real time.
“Scale in modular DeFi will belong to human-supervised agents, not to human operators.”
—Jesus Rodriguez, CPO Sentora
The Scaling Problem: Structural or Operational?
A natural response to growing complexity is to expand the team. More analysts, more monitoring tools, more frequent reviews. This approach has a ceiling, and DeFi has already pushed past it.
Consider what a vault curator is now required to track:
Collateral quality across multiple isolated markets and networks.
Oracle accuracy and divergence risk.
Withdrawal buffer adequacy at any given block.
Liquidity depth relative to position size.
Borrower concentration.
Cross-market correlation.
Governance-driven parameter changes.
Every one of these factors can alter risk profiles with little notice. Each new protocol integration, each additional asset type, each new dependency compounds the dimensionality of the problem. There is no staffing solution that keeps pace with that rate of expansion.
Three specific failure modes characterise manual curation at scale.
Attention is finite: no team can maintain genuine real-time awareness across a sufficiently large and fragmented market surface.
Response is episodic: human oversight happens between decisions, while DeFi risk accumulates between heartbeats.
And reputation, the traditional basis for trusting a curator, cannot substitute for the machine-readable, verifiable guarantees that programmatic capital increasingly demands.
Evolving Into an AI-Driven Operating Model
The agentic vault is a substantive architectural shift. The model is inverted: human-defined policy at the governance layer entirely delegates execution to software, which operates at protocol speed, rather than merely augmenting a human-led process.
What that means in practice is a system that monitors position health continuously, rebalances allocations as market conditions move, enforces mandate constraints without waiting for approval cycles, and flags anomalies before they become losses. The human role is not eliminated but elevated to a strategic level: investment objectives, risk tolerances, collateral preferences, drawdown limits, and liquidity requirements are all human decisions. The agent converts those decisions into a sustained operational posture that no committee could maintain on its own.
This mirrors a transition that has already occurred across other industries. Infrastructure provisioning, digital advertising allocation, and high-frequency market execution all followed the same arc: humans moved from managing the runtime to designing the rules that govern it. DeFi curation is following the same trajectory, and for the same underlying reason: the runtime became too complex and too fast for human-in-the-loop execution to remain viable.
Capital Itself Is Becoming Agentic
The urgency of this transition goes beyond a function of growing protocol complexity. The nature of the capital entering DeFi is also changing.
Increasingly, the entities allocating capital into on-chain markets are not individuals making discretionary decisions but software systems: autonomous treasury managers, embedded yield layers inside consumer financial products, AI-driven trading infrastructure, and programmatic allocation engines operated by institutions at scale.
These participants do not interact with a vault the way a human does. They require guaranteed response characteristics, consistent risk parameters, and machine-interpretable logic at every layer of the stack.
A curation model built around periodic human review is structurally incompatible with that demand. The infrastructure managing agentic capital must itself be agentic. This is not optional, but a compatibility requirement.
What Institutional Allocators Should Consider
For fintechs and financial platforms looking to incorporate on-chain yield, the strategic barrier has rarely been willingness. It has been the absence of a risk management layer that operates at sufficient speed and consistency to meet compliance and operational requirements.
Agentic curation addresses that gap directly. It provides the kind of continuous, governed, auditable risk management that institutional product teams need before embedding DeFi exposure into consumer-facing infrastructure.
For crypto-native institutions, the competitive dynamic is shifting. Differentiation through market access or interface design is narrowing. The durable advantage now accrues to those with superior execution infrastructure: tighter feedback loops, better risk telemetry, and the ability to act on signals before they become adverse outcomes. Model quality is becoming the primary driver of performance.
For DeFi more broadly, this transition reflects something the ecosystem has always implied but not fully acted on: programmable finance should extend its programmability to every layer, including curation. Keeping human operators inside the execution loop is a concession to the limits of earlier tooling.
We explore this vision in detail in our latest report. You can read the Smart Vault Paradigm Paper here.
