Paul Veradittakit, managing partner at Pantera Capital, says enterprise software is approaching a structural break: software is no longer just a tool that helps people work faster, and in more business settings it is starting to become the end buyer itself.
The commentary, written by Veradittakit and translated by TechFlow, argues that for decades the software industry has been built on a stable assumption. Software was written by humans and sold to humans. From packaged software to cloud-based SaaS, sales processes, pricing models and procurement approvals were designed around how employees make decisions and work together.
AI agents move from tools to purchasers
Veradittakit says that assumption is now being challenged. As large language models and autonomous agents mature, a growing share of enterprise purchasing decisions could be made directly by AI, with software code itself becoming the largest end customer.
In his view, that shift would ripple across the technology stack. If the buyer is no longer a human clicking through a webpage, then the traditional sales funnel, graphical user interfaces and card-based payment setup lose relevance. In their place, he points to verifiable onchain wallets, micropayment streaming protocols and machine reputation systems.
The article lays out several core changes:
- For years, technology marketing and product design have focused on persuading humans to pay.
- Companies spend heavily on sales teams and polished graphical interfaces to get procurement managers to approve purchases.
- Once autonomous AI agents take over workflows, they need fast, low-latency and highly reliable machine interfaces rather than dashboards built for people.
- Agent-led purchasing could move away from monthly seat-based subscriptions toward granular pricing based on usage, compute load or even business outcomes.
- Crypto-native settlement and decentralized identity credentials could become the commercial link connecting a new class of software buyers.
Procurement could happen in milliseconds
Veradittakit frames the current AI cycle as more than a cloud transition. He says software is being given the ability to allocate capital and procure resources on its own.
He offers a simple example: a software agent tasked with producing a market research report may need to buy high-value financial data, rent GPU inference capacity and call a third-party data-cleaning interface. In that loop, no employee would step in to approve each individual purchase. Transactions would be matched and settled automatically by code within milliseconds.
Under that model, many familiar parts of enterprise selling, including lead cultivation, business entertainment and lengthy contract negotiations, lose their place in the process.
API-first design becomes central
Veradittakit also says recent conversations with startup founders and investors in Silicon Valley point to a visible shift in product development priorities.
More teams are removing complex front ends originally built for human users and moving toward minimalist protocol layers centered on APIs. For software builders, the issue is becoming practical: if a service cannot be easily discovered, called and paid for by mainstream AI agents, it risks becoming invisible in the next phase of digital commerce.
That puts machine readability and verifiability much closer to the center of product competition.
Fission as an example of machine-to-machine commerce
The article uses Fission to illustrate what this model could look like in practice.
In that ecosystem, AI agents built by different teams can publish requests, provide specialized services and send or receive payment instantly across a decentralized network. The setup does not depend on bank holidays or the friction of traditional cross-border foreign exchange systems. Instead, machine commerce runs continuously.
For Veradittakit, that example shows how software procurement and delivery could be reorganized once payments, identity and service calls are all handled in a machine-native way.
What founders should change
He gives three recommendations to startup teams facing this shift.
- Rework pricing and delivery architecture so products can support granular on-demand billing and machine-native payment settlement.
- Put more weight on machine readability and verifiability so products can be included more easily in trusted automated evaluation systems.
- Track foundational cryptographic infrastructure closely, because it may support the rise of a much larger software buyer base.
Broader crypto themes remain in focus
Beyond agent commerce, Veradittakit says he is also watching developments in crypto business models, regulation and product innovation.
He points to faster exploration of onchain asset tokenization by traditional financial institutions and to clearer regulatory boundaries across major jurisdictions. In the article, those trends are described as laying groundwork for institutional capital to enter the market.
He ends on a simple point: technology does not wait for human preference. If software becomes the largest customer, the value map of the digital economy will be redrawn.
