Designing CX for Non-Human Customers: AI Agents, APIs, and Machines as Users

The Rise of Non‑Human “Customers”

Traditional customer experience (CX) design assumes a human end-user, but an emerging reality is that many “customers” interacting with businesses today are not human at all – they are AI agents, software scripts, or connected machines. Analysts predict that by 2030, over half of consumer purchasing could be driven by AI-assisted buyers rather than direct human decisions. In B2B contexts, the shift is already underway. For example, Walmart’s internal AI procurement platform now closes nearly 70% of supplier contracts with over 2,000 vendors without human intervention. These autonomous systems act on behalf of people or organizations to find products, negotiate terms, and execute transactions at machine speed.

Even in everyday consumer life, non-human agents are making purchases. Consider the Internet of Things: Amazon’s Dash Replenishment Service (DRS) enabled devices like appliances and pet feeders to automatically reorder supplies when they run low. In such cases, the device itself becomes the customer initiating the order. The financial industry is adapting as well – payment networks are building infrastructure for “agentic commerce.” Visa, for instance, has launched its Visa Intelligent Commerce platform to let approved AI shopping agents securely find products and complete purchases on behalf of users or businesses. And enterprise software providers like SAP are infusing procurement systems with AI: the next generation of SAP Ariba includes embedded AI agents that sense, reason, and act autonomously across sourcing and purchasing workflows. These examples show that non-human customers – whether personal digital assistants, bots, or automated enterprise systems – are no longer science fiction; they are active participants in markets today.

1. How Machines “Experience” Products and Services

When the primary user is an algorithm rather than a person, the nature of “customer experience” changes fundamentally. An AI agent does not have feelings, aesthetic preferences, or the cognitive biases that humans do. Your next customer might not read your carefully crafted marketing copy or care about your brand story – it might be an algorithm that only cares about facts and efficiency. Instead of scrolling through a visual interface, a machine agent discovers and evaluates offerings through data feeds, API responses, and digital signals.

This means discovery and decision-making become highly logic-driven. Whereas humans respond to emotional cues and brand perception, AI agents perform ruthless logical matching of needs to specifications. For example, an AI shopping assistant tasked with restocking supplies will search product directories and schema-tagged content rather than relying on advertisements. If your product information isn’t in a machine-readable format (structured data, schemas, APIs), then to such an agent your offering essentially does not exist. Evaluation by a machine is equally unforgiving – the agent will compare objective parameters like price, quality ratings, compliance certifications, and performance metrics. Any inconsistency or missing data can lead to an automatic filter-out, with no second chances or “gut feeling” to save the sale. As one expert puts it, the purchase process shifts “from persuasion to verification”. In other words, a human customer might be swayed by branding or a friendly salesperson, but an AI customer simply checks whether your product meets the criteria and proofs it’s looking for (price thresholds, safety standards, uptime SLAs, etc.) and then makes a binary decision.

Notably, all of this happens at machine speed. Decisions that might take a human days of research or weeks of relationship-building can occur in milliseconds when algorithms are in charge. This compression of the buyer’s journey means companies must be ready with instant, data-driven responses. In summary, designing CX for a non-human user means ensuring that your business’s offerings are visible to algorithms, comprehensible to algorithms, and attractive to algorithms based on logical parameters rather than human emotion. To achieve that, experience designers need to embrace a new set of principles focused on the Agent Experience (AX) alongside the human User Experience.

2. Key Principles of Designing for AI Agents (AX)

Designing for AI agents and other machine customers has given rise to the concept of Agent Experience (AX) – a parallel discipline to traditional UX that focuses on making services easy for software to consume. While UX optimizes interfaces for humans, AX ensures a digital service is intelligible, reliable, and efficient for non-human agents. Below are core principles and best practices for AX design:

Structured, Unambiguous Data: Provide information in consistent, self-describing formats that algorithms can parse with certainty. All units, fields, and terms should be explicitly defined to eliminate any ambiguity. For instance, label data with standard schemas (e.g. using schema.org for product attributes) and use consistent units of measure and date formats across your system. If an AI agent has to guess or infer context, your data isn’t AX-optimized. Clear ontologies and metadata are the foundation of machine-readable CX.
Robust and Predictable APIs: From a machine’s perspective, the API is the product. Design APIs that are stable, well-documented, and behave consistently. This means strict versioning and contracts – once you define an endpoint’s format, don’t break it without notice. Use explicit machine-readable error codes and messages so that an agent can automatically handle errors without human help. In contrast to human developers who might tolerate quirks, AI agents “don’t forgive poor API design”. They expect consistent inputs/outputs every time. Predictability is valued over flexibility: an agent would rather have a rigid, always-consistent API than one that tries to be too clever or variable.
High-Throughput, Batch-Oriented Operations: Anticipate that AI agents will operate at volumes and speeds far beyond human usage. A single autonomous agent might make thousands of API requests per minute, something a human user or developer would rarely do. Design your system to handle bulk operations efficiently. Instead of forcing an agent to retrieve one record at a time in a loop, provide batch endpoints that return many results in one call. Paginate results and allow filtering or selective field returns so that agents can get exactly what they need with minimal calls. This not only caters to the agent’s need for efficiency but also prevents your own infrastructure from being overwhelmed by inefficient agent behavior. In short, optimize for automation at scale.
Consistency and Clarity Across All Touchpoints: Machines excel at spotting inconsistencies – and they won’t “work around” them like a human might. It’s crucial to maintain strict consistency in data formats and responses throughout your service. If one API returns a date-time in ISO 8601 format, every date-time in all your APIs should use that format. If a product catalog feed uses a field name price_usd in one place, don’t call it cost elsewhere. This uniformity allows algorithms to integrate with confidence. Even minor variations that a human could overlook (like one endpoint using userid and another using user_id) might break an automated client. AX design demands a disciplined, unified schema across the board.
Optimize for Machine Consumption: Beyond just data correctness, consider the payload and interaction patterns from a machine’s point of view. Remove bloat that humans might ignore but machines shouldn’t have to deal with. For example, allow agents to request only the fields or data elements they need, to minimize bandwidth and processing. An AI agent pulling customer info might not need high-resolution profile photos or verbose human-friendly descriptions – so your API could support query parameters to omit those. Similarly, think in terms of state management: implement idempotency for operations, so that if an agent repeats a request due to a lost response, it won’t accidentally perform the action twice. These patterns (using idempotent tokens for transactions, providing hooks for selective data retrieval, etc.) make your service more resilient and convenient for autonomous clients.
Support Standard Protocols and Interoperability: As the ecosystem of AI agents grows, common languages or protocols for agent-to-agent interactions are emerging. Forward-looking organizations should embrace industry standards for data exchange and agent communication. Whether it’s standardized schemas for product data, open banking APIs, or upcoming agent negotiation protocols, supporting these can make your service easily discoverable and usable by other machines. In the future, agents might use directories or marketplaces to find services; if your interfaces adhere to a known standard, you’re more likely to be included in an agent’s options. In essence, design with integration in mind, not as a closed system.
Comprehensive Observability and Testing: Designing for non-human users doesn’t end at launch – you need robust monitoring of how algorithms actually interact with your service. Implement logging and analytics focused on agent behavior: track unusual bursts of repeated calls (which might indicate an agent stuck in a loop or an inefficient integration). Measure latency and error rates at high granularity, since agents are less tolerant of slowdowns than humans. Many companies are establishing a dedicated sandbox environment to test with simulated agents. In a sandbox, you can unleash various AI agents or scripts to see how your system holds up under machine-driven usage patterns. This helps identify failure points or “weird” behaviors – for example, an agent misinterpreting data or overloading a certain endpoint – before they cause problems in production. Treat these agent interactions with the same attention you’d give to user experience testing or usability labs, because for machines this is their user experience.

By following these AX principles, organizations essentially treat their APIs and data as first-class products geared toward a new type of user. This often requires cross-functional effort – product managers, UX designers, and engineers collaborating to make services that are not only user-friendly for people but also machine-friendly for bots and AI consumers.

3. Trust and Transparency in an Algorithmic Market

Building trust with human customers often relies on brand reputation, emotional connection, and marketing messages. But what does “trust” look like to a non-human customer? In machine-to-machine interactions, trust is established through data: performance metrics, consistency, security credentials, and compliance records. An algorithm will not “take your word for it” – it needs evidence. As noted by one banking executive, for machine customers, reputation is measured in uptime stats, response times, and certifications, not in brand heritage. In practical terms, this means companies must expose their reliability and compliance in ways that algorithms can verify.

Consider reliability: a service might publish its API uptime percentage or latency metrics in a standardized format, or even have an endpoint that an agent can query for current system status. Likewise, compliance and quality credentials should be made machine-readable. If your product is ISO certified, that fact should be available as a data point an AI agent can check (for example, in the metadata of a product listing or via an API call). The same goes for security or sustainability credentials – some procurement bots might be programmed to select only vendors that meet certain sustainability benchmarks or cybersecurity standards. If you want to survive their filtering, you need to broadcast those attributes in the digital language they understand.

This new paradigm shifts the onus to “prove it” onto suppliers. Marketing to a human, you might claim to be the fastest or most reliable; marketing to an AI, you’d better show actual numbers. For example, an AI buyer might require a response time under 200ms for mission-critical integrations – if your service’s API documentation includes a published 95th-percentile response time of 180ms (and a history to back it up), the agent will trust and choose you. If not, you may be automatically passed over. One scenario described is an AI personal assistant filtering products by criteria like API uptime, data security, and absence of negative supply chain flags – it will “filter out instantly” any options that don’t meet the predefined thresholds, with no opportunity for a charm offensive. In short, machines are strict fact-checkers: they demand transparency and will compare hard data points where humans might be swayed by emotion or loyalty.

Designing CX for algorithms therefore requires a strategy of extreme transparency and verifiability. Some forward-thinking strategies include providing machine-readable summaries of your offerings and value proposition. For instance, a company might offer a public JSON file or API endpoint that summarizes its key offerings, prices, and service terms – effectively a datasheet that an AI agent could ingest directly, bypassing the need to scrape a website or interpret human-oriented text. This ensures you’re speaking the agent’s language. Another approach is embedding structured data on websites (using meta tags or JSON-LD) that conveys product info, inventory levels, or business credentials. This is analogous to search engine optimization (SEO), which was an early form of designing for non-human consumers (i.e. making web content legible to Google’s algorithms). Now the scope is broader: optimizing all aspects of the customer journey for algorithmic consumption.

Finally, consider security and fairness in a machine-driven market. When algorithms make rapid decisions or transactions, there’s potential for abuse or unintended consequences. Companies must incorporate safeguards like robust authentication (to ensure only legitimate agents transact) and rate limiting that can throttle a swarm of bots if needed. Imagine an influencer’s AI assistant triggers thousands of other agents to buy a hot product – your systems should distinguish genuine high-frequency agent activity from DDoS-like malicious traffic. Also, ethical considerations arise: algorithms might systematically favor lowest cost, which could drive suppliers into unsustainable practices unless the algorithm’s criteria include ethics and sustainability. Designing the rules that agents follow might partly be in the user’s hands (e.g., a consumer sets their AI to prioritize eco-friendly options). But businesses should be prepared for an environment of ruthless competition on measurable value. The focus shifts to quantifiable excellence: price-performance, reliability, compliance, user ratings – anything that can be encoded and compared. CX designers will need to ensure that the qualities that differentiate their brand are represented in data form, otherwise an agent will perceive no differentiation at all. The risk of a purely algorithmic market is a “race to the bottom” where emotional loyalty disappears; some experts suggest businesses will need dual strategies, preserving human emotional engagement while simultaneously catering to machine criteria, to avoid complete commoditization.

4. Organizational Shifts to Embrace Machine Customers

Adapting to AI and machine customers isn’t just a technical task – it’s a strategic and cultural transformation. Many organizations initially react to this concept with “Isn’t this just about having APIs?” or “That sounds like an IT problem, not CX. In reality, serving non-human customers spans multiple disciplines. Yes, it requires solid technical foundations (data architecture, APIs, etc.), but it also requires customer experience strategy to ensure that autonomous agents can discover, evaluate, and choose your company’s offerings in alignment with your business goals. In other words, someone needs to deliberately design how your services present themselves to algorithmic buyers – this is a role for CX professionals working hand-in-hand with technical teams.

Leading companies are beginning to create dedicated roles or teams focused on Agent Experience. For example, some have appointed AX strategists or “bot advocates” who effectively “speak for the bots” in product planning, making sure the needs of AI agents are considered alongside human users. API product managers are being treated as key UX designers – because an API that is delightful for a developer or an AI to use is now as critical as a slick mobile app is for an end-user. Data governance specialists ensure that data remains clean and consistently structured for machine consumption across the organization. Quality assurance teams are adding automation and simulation engineers who create test harnesses of simulated agents to probe the system’s AX in different scenarios. This cross-functional approach echoes the fact that machine customer experience bridges traditional silos: it’s part IT, part product, part marketing, part compliance. Building a coalition is crucial – “territorial thinking” (debating whether it’s an IT vs. CX responsibility) “misses the point entirely”. Successful adaptation will require breaking down barriers so that all stakeholders collaborate on making the business machine-friendly.

On a practical level, organizations can start with small steps: identify where autonomous or algorithmic interactions already touch your business, and improve those touchpoints. Chances are you already have some machine “users” today – common examples include search engine crawlers indexing your site, pricing bots comparing your products, or clients integrating via your APIs. Treat these not as peripheral, but as proto-customers to learn from. Analyze your logs for telltale signs of agent traffic (e.g. repetitive patterns or high-frequency bursts at odd hours). Engage with your developer ecosystem to see how many integration use cases are fully automated. These insights can highlight pain points to fix early (perhaps your API has an inflexible format causing developers’ bots to crash, or your lack of bulk operations is forcing inefficient calls). By smoothing these interactions, you not only support current partners better but also lay the groundwork for the coming wave of machine customers.

Another key organizational shift is mindset: expanding the definition of “customer-centricity.” Teams need to internalize that “customer” now includes algorithms. As one CX leader noted, we must redefine customer experience as “designing optimizable interaction journeys with measurable outcomes, regardless of whether the customer processes information emotionally or logically. In practice, this means setting KPIs and success metrics for machine-facing services too – e.g. how quickly can an AI agent onboard to our API? What percentage of machine-initiated inquiries result in successful transactions? Just as we measure human customer satisfaction, we might track agent integration success rates or algorithmic retention (does the AI keep using our service or switch to a competitor?). By treating algorithmic agents as a legitimate customer segment, companies can begin to craft tailored experiences for them. This might even extend to marketing and sales – for instance, ensuring your product data is syndicated to all the places an AI might look (kind of like SEO, but across multiple agent ecosystems), or negotiating placement with AI platform providers so that your service is recommended by default.

Crucially, none of this implies humans are written out of the picture. We are heading toward a hybrid reality where human and AI customers coexist. In B2B sales, you might still have human relationship managers and negotiations, but in parallel the client’s AI system is crunching numbers and checking your every claim in the background. Or in consumer retail, a human might ultimately click “buy,” but an AI concierge might have curated the shortlist of options. Therefore, companies must design joint experiences where the machine-to-machine interactions are seamless and the human touchpoints remain supportive. One challenge is orchestrating these dual layers – ensuring that your human-facing channels and machine-facing channels are consistent (e.g. an AI might pull price info via API that differs from what a human sees on the website – those inconsistencies could erode trust from both sides). Aligning these and coordinating handoffs between human and AI touchpoints will be a new frontier for CX strategy.

5. Examples of Early Adopters and Innovations

Despite being a nascent area, there are already notable examples of how businesses are adapting to non-human customers:

Walmart’s Autonomous Procurement: As mentioned, Walmart utilizes an AI-driven procurement platform that automates the majority of its purchasing negotiations with suppliers (handling ~70% of contracts with no human in the loop). This system can automatically evaluate vendor offers against Walmart’s criteria and finalize deals, demonstrating huge efficiency gains in B2B transactions. Suppliers hoping to do business with Walmart increasingly must ensure their catalogs, prices, and compliance info are available in formats that Walmart’s algorithms can process. This is a prime example of a large enterprise expecting its partners to cater to a machine as the customer.
Amazon’s IoT Reorders: On the consumer side, Amazon’s Dash Replenishment and similar IoT services allow devices to be customers. For instance, Brita water filters, printers, or pet food dispensers have been enabled to sense low supply and directly place an order through Amazon on behalf of the user. Here the “CX” is about invisibility – a perfectly designed machine experience results in the user simply never running out of the product, without any manual action. Amazon’s system provides the APIs and certification for device makers to integrate this capability, effectively turning consumables purchasing into a machine-to-machine subscription. The success of such programs underlines that convenience and loyalty can be achieved by appealing to an AI’s efficiency (and by extension, the human’s desire for hassle-free service).
Visa’s Agentic Payments: The financial sector sees agent-based purchasing as a major trend and is gearing up for it. Visa’s Intelligent Commerce initiative provides developer tools and APIs for AI agents to conduct secure payments and even bundle services for a user. In one Visa example, a corporate AI assistant detecting low inventory could automatically source a new supplier, negotiate pricing, and pay via Visa APIs without a human buyer involved. Mastercard has similarly piloted “Agent Pay” concepts. These efforts show how critical infrastructure players are adapting their CX (or rather “DX” – developer experience) to suit AI agents, knowing that millions of transactions may soon be initiated by algorithms. The trust and safety components (like tokenized payments, fraud detection tuned for bots, and strict authentication of AI identities) are especially important here to prevent abuse when non-humans start transacting in volume.
SAP Autonomous Sourcing Agents: Enterprise software providers are embedding AX features into their products. SAP’s latest upgrades to its Ariba procurement suite boast AI assistants (branded “Joule”) that autonomously handle sourcing tasks – from scanning supplier databases to analyzing bids and compliance. This means in a large company’s purchasing department, an AI agent could automatically evaluate suppliers on predefined criteria (cost, quality scores, ESG ratings, etc.) and either make recommendations to a human manager or even execute routine purchases end-to-end. The implication is that suppliers dealing with such a system will need to present their information (product specs, prices, certifications) in the precise structured formats the AI expects, or risk being overlooked. SAP’s move also underscores that user experience within software now includes the experience of AI helpers – the software must be designed to cooperate with and inform these agent assistants, not just the human operator.
Standard Chartered’s Machine-Readable Banking Services: In the banking sector, Standard Chartered’s corporate banking division has recognized the rise of “machine customers” in areas like algorithmic trading, treasury management systems, and autonomous procurement by clients. The bank has been re-evaluating how to make its products and services “discoverable and trustworthy to algorithms.” This includes publishing detailed API documentation, exposing performance and reliability metrics, and ensuring data outputs are extremely consistent and well-structured. The bank’s CX team views this as designing trust at scale – whereas a human client might be impressed by brand history and personal relationships, an AI client will “trust” Standard Chartered only if its digital interfaces perform flawlessly and its data proves its track record. By proactively building out these capabilities, Standard Chartered aims to remain the go-to choice when clients deploy smart agents to handle financial interactions.

These examples illustrate a mix of B2C and B2B scenarios, all converging on the same theme: to serve non-human customers, companies must rethink how they package and deliver value. The payoff for those who get it right can be significant. Early adopters are already seeing competitive advantages – e-commerce platforms that embraced agent-friendly APIs are powering the next generation of shopping assistants, financial firms with streamlined data access are enabling advanced trading algorithms, and healthcare providers with open data interfaces are supporting AI-driven diagnostics. By catering to machine intermediaries, businesses can tap into new automation-driven use cases and even create demand from AI systems themselves (imagine an AI that knows to always reorder from a vendor because it’s the easiest integration). On the flip side, companies that ignore this shift risk losing visibility. If your product isn’t picked up by the algorithms that filter options for consumers or businesses, it may as well not exist in those future funnels. Or as one strategist bluntly warns, those who do not make their services legible and trustworthy to AI agents “will become invisible to a substantial and growing segment of the market.”

Conclusion: Preparing for the Agentic Future

Designing CX for non-human customers is rapidly moving from an avant-garde idea to an essential business practice. We’re entering an era of the “agentic internet,” where swarms of AI agents transact, negotiate, and cooperate on behalf of humans. In this future, customer experience and developer experience merge, focused on enabling smooth machine-to-machine interactions alongside human ones. Organizations must evolve now, by making data transparency, API excellence, and interoperability core to their CX strategy. This doesn’t diminish the importance of human-centered design – rather, it adds a new dimension to it. The companies that thrive will be those that can serve both humans and algorithms effectively, leveraging the unique strengths of each.

In practical terms, being ready for non-human customers means investing in the “plumbing” of digital experiences: structured data, rigorous API design, monitoring tools, and cross-functional teams to continually improve these underpinnings. It also means reimagining the customer journey with algorithms in mind – from how an AI might learn about your offering, to how it decides and acts on a purchase, to how it evaluates post-purchase service. In all these stages, speed, clarity, and factual proof will trump the emotional and experiential facets we traditionally consider in CX. The agent economy is not a distant speculation; it’s already here in its early steps. Adapting to it is not just a defensive move but an opportunity: it forces businesses to become more data-driven, efficient, and transparent, which ultimately benefits all customers, human or not. The concept of the Customer Machine was first articulated by Don Scheibenreif and Mark Raskino in When Machines Become Customers, where they outlined how AI-enabled non-human customers would fundamentally change how businesses design and deliver value. Another CX visionary Katja Forbes the author of very new book, Machine Customers: The Evolution has Begun: How AI that buys is changing everything, which I highly recommend, noted, this is likely the biggest transformation in customer experience since the field was invented, and those who step up can help shape the next era of commerce. In summary, designing for AI agents and machine customers is about future-proofing your customer experience – ensuring that as your customers increasingly delegate decisions to machines, your business remains front-and-center in that machine-mediated world. The brands that embrace this challenge will continue to be “seen” and chosen by customers’ digital proxies, while others risk fading into algorithmic obscurity. The time to evolve your CX for non-human users is now, before the gap between machine-optimized businesses and the rest becomes too great to bridge.

If this article resonated with you, feel free to share it — and let’s connect on LinkedIn for more insights and future posts: Ricardo Saltz Gulko

We invite you to subscribe to the open ECXO.org (European Customer Experience Organization) and be part of shaping our Global Business Network, now open for individual access: https://ecxo.org/individuals/

Ricardo Saltz Gulko, columns in several respected CX publications.