a16z Frontier interprets the AI Agent track, discovering a new entrepreneurial opportunity
Author: Christian, a16z Frontier GTM
Compiled by: Gu Yu, ChainCatcher
Our work in the Frontier go-to-market team allows us to frequently engage with numerous leading companies in the AI infrastructure layer, covering everything from computing power suppliers, managed inference businesses to development teams building Agents. As our collaborations deepen, we have been exploring how cryptographic technology can assist these companies in their development.
Given that stablecoins have the potential to make payment processes faster, cheaper, and easier to manage, I initially thought payments would be the most straightforward entry point. However, computing power and inference suppliers have focused on the fundamentals of growth and business operations: ensuring computing power reserves, deployment, and profit margin management. For teams building and deploying Agents, reliability and controllability are mentioned far more frequently than payments: What can the Agent access? What operations is it authorized to perform? How can companies verify what actually happened?
Payments remain a promising opportunity for crypto x AI. But these conversations have raised another possibility: at least in the short term, verifiability is a more important application scenario for blockchain. As companies grant Agents more responsibilities, they need a mechanism to determine who authorized a particular operation, what restrictions apply, and whether that authorization is still valid.
For entrepreneurs, a more direct opportunity lies in helping companies maintain control over the actions of Agents while delegating more work to them. The rich experience of cryptographic teams in signing and authorization delegation opens the door for them to enter this market.
Empowering Agents with More Responsibilities
Empowering Agents with more responsibilities requires both reliable performance and clear limitations on their permissions. If an Agent consistently messes up tasks, mere permission control won't solve the problem. However, when an Agent's performance is already excellent and practical, companies still need to decide what it can access, which operations require approval, and how to revoke its access. This is where the expertise of cryptographic teams in signing and authorization delegation comes into play.
Before companies expand the responsibilities of Agents, they must clearly define what the Agent can access, what operations it can perform, and how to enforce these restrictions—especially in this new world where interfaces (and user control) are changing. This principle applies whether the Agent is reading company information for ordinary consumers or modifying production environment systems for developers.
Imagine a company that uses an Agent provided by an external software vendor to investigate a production incident. The Agent can read logs and propose fixes, but needs approval before writing to and modifying the system. The company grants permission, the vendor runs the Agent, and the receiving system must verify whether the operation is permitted. However, a survey conducted in January 2026 among 150 IT and security decision-makers showed that 69% of respondents indicated security concerns are slowing down the adoption of Agents. Only 27% of respondents believed their existing identity systems were fully capable of governing Agents at scale. Major concerns include data breaches and excessive permissions.
Meanwhile, research on deployed Agents also indicates that there is still much work to be done before companies grant them greater autonomy. In the article "Measuring Agents in Production" (which explores why organizations build Agents, how they build them, how they evaluate them, and the main development challenges they face), reliability is listed as the primary development challenge. Teams often restrict Agents to short workflows while maintaining human involvement in the process. Jellyfish's latest engineering data also indicates that Agent participation is increasing, but workflows that advance tasks from requirement specifications to code merge requests (PRs) without human intervention remain relatively rare.
Code-writing Agents are another example: they may generate very useful modified code but still require human approval for each deployment. Companies need to clarify which systems it can change, which operations require review, and how to revoke its access. Better control mechanisms will enable companies to confidently delegate more work to these two types of Agents.
Empowering Companies with Delegation Capabilities
When Agents work across multiple companies, the issues become more complex. In the previously mentioned production incident example, the company grants permission, the vendor runs the Agent, and the receiving system must enforce restrictions. If there is a verifiable record of Agent permissions, both the company and the vendor have a common basis to verify whether an operation was authorized.
Existing identity systems and application permissions can already address some aspects of this issue. For cryptographic technology or related companies, the entry point lies in leveraging their teams' experience in signing, authorization delegation, and user-controlled account management to make authorization, execution, revocation, and auditing delegation between these systems easier.
Cryptographic teams have begun applying this experience to the challenges faced by companies: enabling software to take action while not granting it unlimited permissions. In the Turnkey Wayfinder case study, Parallel Studios adopted a session-based signing mechanism that allows users to authorize multiple transactions at once without needing to sign each one individually. According to the case study, this approach retains user custody of their wallets.
Wayfinder applies this capability to on-chain transactions. What interests me and the enterprise teams I engage with is enabling software to act without granting it unrestricted power. The greater opportunity lies in allowing Agents to undertake more work within enforceable and revocable permissions set by the company while retaining records of all authorized content.
When independent parties need a shared permission record that is not controlled by any single entity, or when the controlled assets are already on-chain, blockchain can create value. However, parties still need to reach a consensus on who has the authority to grant permissions, what operations the permissions allow, and how the revocation mechanism operates—which is precisely the original intent of blockchain design.
For off-chain applications, companies can record the Agent's permissions and their expiration times, along with any revocation records, on-chain. This registration system can run in parallel with existing off-chain identity and application systems, which will verify the Agent's signing requests against the on-chain system before allowing an operation. Both companies and vendors can verify what has been authorized, while the receiving system continues to be responsible for enforcing these restrictions.
Since many such Agents operate off-chain, reaching enterprise customers means that any built-in control mechanisms must seamlessly integrate with the existing systems these customers are already using. For cryptographic teams, this often means extending their products beyond on-chain applications. To bridge this go-to-market gap, these teams need to demonstrate how their experience can help companies empower Agents with more responsibilities without granting them excessive access. The most pragmatic test is: can the target customer allow an Agent to handle a task that was previously unable to be delegated due to permission issues?
Simply telling companies to "trust their Agents" is hardly a compelling reason to purchase your product. Instead, entrepreneurs should first target specific tasks that companies want to delegate but cannot due to permission or approval requirements. Then demonstrate how your product eliminates this barrier. Based on this, further position why this improvement is worth adopting a new system—does it provide enough value? Can it improve the company's bottom line, or help it explore new business areas?
Who Will Pay for This?
In exploring enterprise-level go-to-market strategies, another key aspect is identifying the right buyers:
For companies deploying Agents internally, the buyer is likely to be the identity or security engineering manager, with the CISO as a potential sponsor. The problems they want to solve may include Agent permission sprawl, slow revocation, or the need to reconstruct behavioral trails across multiple systems for investigations. You should refine your pitch around a specific deployment scenario that your product can unlock and demonstrate how it will integrate with the company's existing security systems.
For software companies selling Agents to enterprises, the entry point should be the CTO or the head of Agent platform engineering. Consider asking them how much authorization development work is required for each deployment for specific customers, and whether this work is slowing down product launch timelines or draining engineering team resources. Show them how your product can meet these customer needs with less customization work and accelerate deployment timelines.
In either case, founders need to understand who is maintaining the current alternatives, what the costs are, and whether they are more likely to solve it themselves or are willing to pay existing vendors to handle it.
Existing identity providers have an inherent advantage because their systems are already deployed. Even if new products can offer better control features, target customers will always weigh the benefits against the costs of integrating and maintaining another tool. (Some enterprise salespeople liken this internal resistance to fighting gravity.) Founders must understand the shortcomings of existing systems and why customers would choose your product instead of simply expanding their current systems. The hurdle to overcome is not just filling necessary gaps but also creating an enhancement that clearly outweighs the effort required to introduce a new tool. Unlocking a significant Agent deployment may justify the investment in extensive integration work; however, a minor operational improvement must possess characteristics that make it easy to implement. Start with a very clear workflow and take on as much of the integration work as possible yourself, so the target company does not have to go through great lengths to get everything sorted.
If these conversations reveal such opportunity gaps, start small and make the value tangible. Allow the Agent to perform a specific action under executable and revocable permissions from the customer, and return records to the customer's existing auditing tools. You can validate through pilot projects whether this can save the customer integration work or enable them to delegate tasks that still require manual review while preventing unauthorized actions.
To assess potential customer opportunities, you or your sales team need to ask the following questions:
What tasks does the customer want to delegate, and what currently hinders the delegation of those tasks?
Who maintains the current alternative solution, and what is the cost?
What kind of improvement is sufficient to justify the introduction of a new system?
What does adopting the solution require from the customer, and how much of that work can you take on?
What functionalities does your solution provide that the customer's existing tools cannot achieve or cannot easily achieve?
You should clarify the answers to these questions before approaching the customer, not after.
Next, let's talk about payments
New payment infrastructure will only be truly adopted when the resistance to achieving the customer's goals is eliminated. Simply increasing the adoption rate of Agents does not tell us which workflows need new payment infrastructure. The exact needs depend on specific workflows:
An Agent purchasing access to an unknown API may need a way to complete payments immediately.
A coding Agent operating within the company's existing subscription framework may not encounter any payment resistance at all.
Compute providers may benefit from a different settlement method, but they also have more pressing operational challenges that need to be prioritized.
Optimizing payments only becomes a top priority when existing systems hinder the goals customers want to achieve. For example, an Agent may need to manually set up an account before purchasing access to a service; a certain billing model may make small purchases impractical; the settlement process may impose high operational burdens. These are all sufficient reasons to consider adopting different payment systems.
Payments also require control mechanisms. For example, AP2 uses cryptographically verifiable authorization directives (mandates) to record actions authorized by users for Agents to perform, including purchases made in the user's absence. However, customers may very well need more robust control mechanisms without requiring an entirely new payment method.
Simply increasing transaction volume is not enough to establish a demand for new payment infrastructure unless you clarify what customers are actually purchasing and why existing payment methods fail to meet their needs.
As we enter the era of the Agent economy, the core question is not how to introduce stablecoins into AI businesses. Rather, what can enable these companies to confidently delegate more responsibilities to Agents? Founders in the crypto industry and other cutting-edge fields should engage more with these teams, trying to understand this gap and address it. Selling cutting-edge products to enterprises is no different: you must always start from where the customer currently is, not just stay at the place you hope they will reach.
ChainCatcher reminds readers to view blockchain rationally, enhance risk awareness, and be cautious of various virtual token issuances and speculations. All content on this site is solely market information or related party opinions, and does not constitute any form of investment advice. If you find sensitive information in the content, please click , and we will handle it promptly.
Popular articles











