The Machine Customer

Machine customers can be classified into three categories, each demanding a different design approach. The Bound Agent operates within tight, specific parameters, like a sophisticated comparison engine with execution authority (e.g., finding the cheapest insurance that meets a specific criteria). The Adaptive Agent learns a principal's preferences over time, adjusting its behaviour and making decisions that fall within the spirit, but not the letter, of its initial delegation. Finally, the Autonomous Agent exercises broad discretionary judgment, such as managing a financial portfolio, requiring a robust accountability architecture to ensure its decisions are legible and responsible. Designing for these agents requires moving from human-centred design to systems design focused on protocols, APIs, and machine-readable value propositions.

For the banking sector, the machine customer is an immediate existential threat and a profound opportunity. The traditional business model, which relies on customer friction, complexity, and inertia to retain business, is rendered obsolete by agents that are relentless, logical optimisers. A machine customer will switch a savings account to a competitor for a fractional interest rate improvement in milliseconds. This demolishes loyalty and brand value as retention mechanisms. Banks must therefore shift from competing on perception and inconvenience to competing on genuine, machine-readable value: superior products, better terms, and seamless, API-driven service integration. Those that do will thrive by attracting a new class of hyper-efficient customers; those that do not will see their customer base optimised away.

The core design challenge of the machine customer era is twofold, operating at both the machine and human levels. At the machine level, the work is about creating the protocols and infrastructure for agentic commerce, enabling agents to discover, evaluate, and transact. This involves building what the essay terms Agent Identity Architecture - a verifiable system for establishing an agent's principal, scope, and provenance. At the human level, the challenge is designing for oversight. Since the human principal does not participate in the transaction, the experience is defined by the delegation design and the outcome narratives - the structured reports the agent provides. This requires a new form of communication that builds trust and makes autonomous actions legible, ensuring the human can understand, assess, and control their agent's behaviour without micromanaging it.

The emergence of the machine customer compels a strategic pivot for all commercial organisations. Product leaders must prioritise making their offerings machine-readable, which means investing in structured data, comprehensive APIs, and transparent, algorithmically negotiable pricing. The value proposition must be re-engineered to be legible and superior not to a human, but to an optimising agent. This shifts the competitive basis from brand marketing and user experience to verifiable performance and quantifiable value. Businesses must prepare for a future where their most valuable customer is an API call, and their primary user interface is a protocol.

Designers and product teams must expand their focus from human-centred design to a dual-track approach that encompasses both machine-to-machine interaction and human-agent oversight. This means developing new skills in systems design, protocol design, and what AXD calls trust architecture. The key is to create systems that are both efficient for agents and trustworthy for the humans who delegate authority to them. This involves designing clear outcome narratives that explain an agent's decisions and creating intuitive interfaces for setting constraints and reviewing performance. Ultimately, organisations must stop designing just for human attention and start designing for delegated, autonomous action. The institutions that master this will not just survive the transition to agentic commerce; they will define its architecture.