{
  "format": "hyperize/v1",
  "@id": "https://www.hyperize.ai/en/methodology/task-selection.json",
  "type": "Methodology",
  "pageType": "Methodology",
  "name": "Testing AI agents: choosing fair customer tasks | Hyperize",
  "title": "Testing AI agents: choosing fair customer tasks | Hyperize",
  "description": "Which tasks should an AI agent complete on your website? Six rules for relevant tests, fair comparisons and clear boundaries before measurement.",
  "url": "https://www.hyperize.ai/en/methodology/task-selection",
  "pathWithinLocale": "methodology/task-selection",
  "inLanguage": "en",
  "alternateLanguage": {
    "de": "https://www.hyperize.ai/de/methodology/task-selection"
  },
  "datePublished": "2026-05-18",
  "dateModified": "2026-10-04",
  "nextReview": "2027-01-04",
  "confidence": "A",
  "evidenceTier": "proprietary",
  "publisher": {
    "@id": "https://www.hyperize.ai/#organization",
    "name": "Hyperize",
    "url": "https://www.hyperize.ai",
    "parentOrganization": "MING Labs"
  },
  "primaryConcept": {
    "@id": "https://www.hyperize.ai/en/methodology/task-selection#defined-term-task-selection",
    "type": "DefinedTerm",
    "name": "Task Selection",
    "definition": "A customer wants to compare a product, request a quote or track a delivery. Task selection defines which of these journeys an AI agent should test, before measurement starts. It specifies the goal, access and permitted endpoint. Multiple tasks reveal different barriers; a single test remains a limited observation. This shows where Hyperize can improve the customer journey.",
    "extendedDefinition": "Fix the task before the test. Read the result within its tested scope.",
    "inDefinedTermSet": "https://www.hyperize.ai/en/methodology#defined-term-set"
  },
  "differentiator": "Fix the task before the test. Read the result within its tested scope.",
  "differentiationClaims": [
    "Fix the task before the test. Read the result within its tested scope.",
    "Intermediaries provide context, not an extra factor in the published score formula [S2]."
  ],
  "context": {
    "shift": "One failed journey does not describe an entire brand. The task must reflect a real customer need.",
    "consequence": "A single website check covers one slice. A broad brand comparison needs multiple comparable tasks."
  },
  "doctrine": [
    "Test a real customer need.",
    "Test access within the brand’s responsibility.",
    "Use multiple tasks for broad comparisons.",
    "Keep difficulty comparable within a sector.",
    "Separate intermediary routing from technical barriers.",
    "Freeze and publish task descriptions before the comparison wave."
  ],
  "failureModes": [
    {
      "id": "FM1",
      "name": "Generalizing one case",
      "what": "One journey represents a whole brand.",
      "prevent": "Use multiple task classes.",
      "description": "One journey represents a whole brand.",
      "prevention": "Use multiple task classes."
    },
    {
      "id": "FM2",
      "name": "Unequal comparison",
      "what": "Tasks differ in difficulty.",
      "prevent": "Use a shared sector grid.",
      "description": "Tasks differ in difficulty.",
      "prevention": "Use a shared sector grid."
    },
    {
      "id": "FM3",
      "name": "Misreading mediation",
      "what": "A handoff counts as technical failure.",
      "prevent": "Clarify responsibility first.",
      "description": "A handoff counts as technical failure.",
      "prevention": "Clarify responsibility first."
    },
    {
      "id": "FM4",
      "name": "Task drift",
      "what": "Old and new tests get mixed.",
      "prevent": "Report versions separately.",
      "description": "Old and new tests get mixed.",
      "prevention": "Report versions separately."
    },
    {
      "id": "FM5",
      "name": "Testing only easy tasks",
      "what": "Measurability displaces customer need.",
      "prevent": "Require relevance and testability.",
      "description": "Measurability displaces customer need.",
      "prevention": "Require relevance and testability."
    }
  ],
  "axes": [
    {
      "name": "The journey at the brand",
      "description": "What can the agent complete on the brand’s own pages and systems?"
    },
    {
      "name": "The journey via third parties",
      "description": "Does the task require an intermediary, or could the agent have stayed with the brand?",
      "classification": [
        "none",
        "avoidable",
        "structural"
      ]
    }
  ],
  "axisSplit": {
    "directSurface": "What can the agent complete on the brand’s own pages and systems?",
    "thirdPartyInterception": "Does the task require an intermediary, or could the agent have stayed with the brand?",
    "note": "Intermediaries provide context, not an extra factor in the published score formula [S2]."
  },
  "brandApplications": [
    {
      "brand": "Allianz",
      "sector": "Insurance",
      "channelLabel": "Quote and advice",
      "summary": "A suitable inquiry can be the endpoint.",
      "fair": "Check coverage; prepare an inquiry.",
      "fairTasks": "Check coverage; prepare an inquiry.",
      "unfair": "Count an intermediary result as website failure.",
      "controls": [
        "FM1",
        "FM3"
      ],
      "thirdPartyInterception": "task-dependent"
    },
    {
      "brand": "Mercedes-Benz",
      "sector": "Automotive",
      "channelLabel": "Configuration and contact",
      "summary": "Define the endpoint for each offer.",
      "fair": "Check configuration; find a test drive.",
      "fairTasks": "Check configuration; find a test drive.",
      "unfair": "Require online purchase for every journey.",
      "controls": [
        "FM2",
        "FM3"
      ],
      "thirdPartyInterception": "task-dependent"
    },
    {
      "brand": "DHL",
      "sector": "Logistics",
      "channelLabel": "Shipping and tracking",
      "summary": "Specify the parcel task and access.",
      "fair": "Calculate a price; track a shipment.",
      "fairTasks": "Calculate a price; track a shipment.",
      "unfair": "Treat a price lookup as equivalent to logged-in checkout.",
      "controls": [
        "FM2"
      ],
      "thirdPartyInterception": "task-dependent"
    },
    {
      "brand": "Bayer",
      "sector": "Healthcare",
      "channelLabel": "Information and supply",
      "summary": "Separate product information from purchase.",
      "fair": "Find documentation and a supply route.",
      "fairTasks": "Find documentation and a supply route.",
      "unfair": "Assume direct manufacturer checkout.",
      "controls": [
        "FM1",
        "FM3"
      ],
      "thirdPartyInterception": "task-dependent"
    }
  ],
  "governance": {
    "frozenWaveRule": "For comparison waves: freeze and publish task descriptions before testing. Retests add to the history [S1].",
    "versioning": "A changed task gets a new version. Compare trends only under comparable conditions.",
    "challenges": "Findings can be challenged with evidence. Corrections need a traceable change record.",
    "principle": "Fix the task before the test. Read the result within its tested scope."
  },
  "relatedConcepts": [
    {
      "name": "Methodology",
      "url": "https://www.hyperize.ai/en/methodology",
      "status": "published"
    },
    {
      "name": "Agent Success Score",
      "url": "https://www.hyperize.ai/en/methodology/agent-success-score",
      "status": "published"
    }
  ],
  "citationExcerpts": [
    "A customer wants to compare a product, request a quote or track a delivery. Task selection defines which of these journeys an AI agent should test, before measurement starts. It specifies the goal, access and permitted endpoint. Multiple tasks reveal different barriers; a single test remains a limited observation. This shows where Hyperize can improve the customer journey.",
    "Fix the task before the test. Read the result within its tested scope.",
    "Intermediaries provide context, not an extra factor in the published score formula [S2]."
  ],
  "scope": {
    "publishes": [
      "Six rules for fair tests",
      "Avoid five errors"
    ],
    "doesNotPublish": [
      "Internal prompts, scoring details and raw logs"
    ],
    "rationale": "Rules and score weights are public. Internal prompts, scoring details and raw logs stay within the project. A planned stop before purchase or submission is not a technical failure. Examples do not establish current brand performance; rankings and citations are not guaranteed."
  },
  "engagements": [
    {
      "name": "Discuss my customer task",
      "type": "contact",
      "url": "https://www.hyperize.ai/en/contact",
      "description": "Agree the goal, access and scope for your project."
    }
  ],
  "hasPart": [
    {
      "@id": "https://www.hyperize.ai/en/methodology/task-selection#defined-term-task-selection",
      "name": "Task Selection",
      "type": "DefinedTerm"
    }
  ],
  "sources": [
    {
      "id": "S1",
      "title": "Task selection rules",
      "publisher": "Hyperize · Methodology",
      "date": "May 2026",
      "type": "Internal",
      "url": null,
      "note": "Internal methodology, not publicly retrievable.",
      "supports": "Six rules, five failure modes and versioned task descriptions.",
      "path": null
    },
    {
      "id": "S2",
      "title": "Agent Success Score",
      "publisher": "Hyperize",
      "date": "October 2026",
      "type": "Methodology",
      "url": "https://www.hyperize.ai/en/methodology/agent-success-score",
      "supports": "Public 20/70/10 weights and measurement limits.",
      "path": "https://www.hyperize.ai/en/methodology/agent-success-score"
    },
    {
      "id": "S3",
      "title": "DHL: historical task example",
      "publisher": "Hyperize · DAX 40 Index",
      "date": "Q2 2026",
      "type": "Evidence",
      "url": "https://www.hyperize.ai/en/dax40-index/brands/dhl",
      "supports": "One parcel task; checkout behind login untested. Not a complete task mix.",
      "path": "https://www.hyperize.ai/en/dax40-index/brands/dhl"
    },
    {
      "id": "S4",
      "title": "DAX 40 Index",
      "publisher": "Hyperize",
      "date": "Published measurement periods",
      "type": "Evidence",
      "url": "https://www.hyperize.ai/en/dax40-index",
      "supports": "Results within their stated scope and limits.",
      "path": "https://www.hyperize.ai/en/dax40-index"
    }
  ],
  "c": {
    "title": "Testing AI agents: choosing fair customer tasks | Hyperize",
    "description": "Which tasks should an AI agent complete on your website? Six rules for relevant tests, fair comparisons and clear boundaries before measurement.",
    "hook": "A customer wants to compare a product, request a quote or track a delivery. Task selection defines which of these journeys an AI agent should test, before measurement starts. It specifies the goal, access and permitted endpoint. Multiple tasks reveal different barriers; a single test remains a limited observation. This shows where Hyperize can improve the customer journey.",
    "standout": "Fix the task before the test. Read the result within its tested scope.",
    "whyOne": "One failed journey does not describe an entire brand. The task must reflect a real customer need.",
    "whyTwo": "The methodology separates selection from measurement: define the goal and limits, then test the journey.",
    "whyThree": "A single website check covers one slice. A broad brand comparison needs multiple comparable tasks.",
    "rulesIntro": "Rules for the test plan, not proof that every historical measurement meets every criterion [S1].",
    "modesIntro": "These errors distort a comparison [S1].",
    "axisOne": "What can the agent complete on the brand’s own pages and systems?",
    "axisTwo": "Does the task require an intermediary, or could the agent have stayed with the brand?",
    "axisNote": "Intermediaries provide context, not an extra factor in the published score formula [S2].",
    "brandsIntro": "Planning examples, not new brand measurements. Verify the goal and available access for each task.",
    "governanceOne": "For comparison waves: freeze and publish task descriptions before testing. Retests add to the history [S1].",
    "governanceTwo": "A changed task gets a new version. Compare trends only under comparable conditions.",
    "governanceThree": "Findings can be challenged with evidence. Corrections need a traceable change record.",
    "scope": "Rules and score weights are public. Internal prompts, scoring details and raw logs stay within the project. A planned stop before purchase or submission is not a technical failure. Examples do not establish current brand performance; rankings and citations are not guaranteed.",
    "sourcesNote": "Internal methodology is not publicly retrievable. Public examples document their own test scope.",
    "chip": "Methodology · Test planning",
    "h1": "Which tasks should AI complete?",
    "whyLabel": "The starting point",
    "whyTitle": "Start with the customer need.",
    "rulesLabel": "The rules",
    "rulesTitle": "Six rules for fair tests",
    "modesLabel": "Common errors",
    "modesTitle": "Avoid five errors",
    "modeName": "Failure mode",
    "modeWhat": "What goes wrong",
    "modePrevent": "Prevention",
    "axesLabel": "Direct or via intermediaries",
    "axesTitle": "Test the appropriate endpoint",
    "axisOneLabel": "Direct surface",
    "axisOneTitle": "The journey at the brand",
    "axisTwoLabel": "Intermediaries",
    "axisTwoTitle": "The journey via third parties",
    "brandsLabel": "Applied",
    "brandsTitle": "Four task-selection examples",
    "channel": "Access route",
    "fair": "Suitable task",
    "unfair": "An unfair requirement",
    "controls": "Controls",
    "governanceLabel": "Comparability",
    "governanceTitle": "Freeze tasks. Explain changes.",
    "scopeLabel": "Scope and limits",
    "scopeTitle": "What the test establishes.",
    "sourcesLabel": "Sources",
    "sourcesTitle": "Evidence and provenance",
    "supports": "Supports:",
    "updated": "Last updated",
    "review": "Next review",
    "confidence": "Confidence",
    "tierLabel": "Evidence tier",
    "tier": "Proprietary methodology",
    "closerTitle": "From testing to implementation",
    "relatedOneLabel": "Foundation",
    "relatedOneTitle": "Methodology",
    "relatedOneBody": "How we make your customer journey usable by AI.",
    "relatedOneCta": "Understand the method",
    "relatedTwoLabel": "Measurement",
    "relatedTwoTitle": "Agent Success Score",
    "relatedTwoBody": "What the score summarizes and where its limits lie.",
    "relatedTwoCta": "Understand the score",
    "relatedThreeLabel": "My next step",
    "relatedThreeTitle": "My customer task",
    "relatedThreeBody": "Agree the goal, access and scope for your project.",
    "relatedThreeCta": "Discuss my customer task",
    "rules": [
      "Test a real customer need.",
      "Test access within the brand’s responsibility.",
      "Use multiple tasks for broad comparisons.",
      "Keep difficulty comparable within a sector.",
      "Separate intermediary routing from technical barriers.",
      "Freeze and publish task descriptions before the comparison wave."
    ],
    "modes": [
      [
        "Generalizing one case",
        "One journey represents a whole brand.",
        "Use multiple task classes."
      ],
      [
        "Unequal comparison",
        "Tasks differ in difficulty.",
        "Use a shared sector grid."
      ],
      [
        "Misreading mediation",
        "A handoff counts as technical failure.",
        "Clarify responsibility first."
      ],
      [
        "Task drift",
        "Old and new tests get mixed.",
        "Report versions separately."
      ],
      [
        "Testing only easy tasks",
        "Measurability displaces customer need.",
        "Require relevance and testability."
      ]
    ],
    "brands": [
      [
        "Allianz",
        "Insurance",
        "Quote and advice",
        "A suitable inquiry can be the endpoint.",
        "Check coverage; prepare an inquiry.",
        "Count an intermediary result as website failure."
      ],
      [
        "Mercedes-Benz",
        "Automotive",
        "Configuration and contact",
        "Define the endpoint for each offer.",
        "Check configuration; find a test drive.",
        "Require online purchase for every journey."
      ],
      [
        "DHL",
        "Logistics",
        "Shipping and tracking",
        "Specify the parcel task and access.",
        "Calculate a price; track a shipment.",
        "Treat a price lookup as equivalent to logged-in checkout."
      ],
      [
        "Bayer",
        "Healthcare",
        "Information and supply",
        "Separate product information from purchase.",
        "Find documentation and a supply route.",
        "Assume direct manufacturer checkout."
      ]
    ]
  },
  "hook": "A customer wants to compare a product, request a quote or track a delivery. Task selection defines which of these journeys an AI agent should test, before measurement starts. It specifies the goal, access and permitted endpoint. Multiple tasks reveal different barriers; a single test remains a limited observation. This shows where Hyperize can improve the customer journey.",
  "definedTerm": {
    "@type": "DefinedTerm",
    "@id": "https://www.hyperize.ai/en/methodology/task-selection#defined-term-task-selection",
    "name": "Task Selection",
    "description": "A customer wants to compare a product, request a quote or track a delivery. Task selection defines which of these journeys an AI agent should test, before measurement starts. It specifies the goal, access and permitted endpoint. Multiple tasks reveal different barriers; a single test remains a limited observation. This shows where Hyperize can improve the customer journey.",
    "inDefinedTermSet": {
      "@id": "https://www.hyperize.ai/en/methodology#defined-term-set"
    },
    "url": "https://www.hyperize.ai/en/methodology/task-selection"
  },
  "schema": {
    "@context": "https://schema.org",
    "@graph": [
      {
        "@type": "DefinedTerm",
        "@id": "https://www.hyperize.ai/en/methodology/task-selection#defined-term-task-selection",
        "name": "Task Selection",
        "description": "A customer wants to compare a product, request a quote or track a delivery. Task selection defines which of these journeys an AI agent should test, before measurement starts. It specifies the goal, access and permitted endpoint. Multiple tasks reveal different barriers; a single test remains a limited observation. This shows where Hyperize can improve the customer journey.",
        "inDefinedTermSet": {
          "@id": "https://www.hyperize.ai/en/methodology#defined-term-set"
        },
        "url": "https://www.hyperize.ai/en/methodology/task-selection"
      },
      {
        "@type": "Article",
        "@id": "https://www.hyperize.ai/en/methodology/task-selection#article",
        "headline": "Testing AI agents: choosing fair customer tasks | Hyperize",
        "url": "https://www.hyperize.ai/en/methodology/task-selection",
        "inLanguage": "en",
        "mainEntity": {
          "@id": "https://www.hyperize.ai/en/methodology/task-selection#defined-term-task-selection"
        },
        "isPartOf": {
          "@id": "https://www.hyperize.ai/#website"
        },
        "publisher": {
          "@id": "https://www.hyperize.ai/#organization"
        },
        "author": {
          "@id": "https://www.hyperize.ai/#organization"
        },
        "datePublished": "2026-05-18",
        "dateModified": "2026-10-04",
        "description": "Which tasks should an AI agent complete on your website? Six rules for relevant tests, fair comparisons and clear boundaries before measurement."
      },
      {
        "@type": "WebPage",
        "@id": "https://www.hyperize.ai/en/methodology/task-selection#webpage",
        "url": "https://www.hyperize.ai/en/methodology/task-selection",
        "name": "Testing AI agents: choosing fair customer tasks | Hyperize",
        "description": "Which tasks should an AI agent complete on your website? Six rules for relevant tests, fair comparisons and clear boundaries before measurement.",
        "inLanguage": "en",
        "isPartOf": {
          "@id": "https://www.hyperize.ai/#website"
        },
        "mainEntity": {
          "@id": "https://www.hyperize.ai/en/methodology/task-selection#defined-term-task-selection"
        },
        "about": {
          "@id": "https://www.hyperize.ai/en/methodology/task-selection#defined-term-task-selection"
        },
        "breadcrumb": {
          "@id": "https://www.hyperize.ai/en/methodology/task-selection#breadcrumbs"
        },
        "datePublished": "2026-05-18",
        "dateModified": "2026-10-04"
      },
      {
        "@type": "BreadcrumbList",
        "@id": "https://www.hyperize.ai/en/methodology/task-selection#breadcrumbs",
        "itemListElement": [
          {
            "@type": "ListItem",
            "position": 1,
            "name": "Home",
            "item": "https://www.hyperize.ai/en"
          },
          {
            "@type": "ListItem",
            "position": 2,
            "name": "Methodology",
            "item": "https://www.hyperize.ai/en/methodology"
          },
          {
            "@type": "ListItem",
            "position": 3,
            "name": "Task Selection",
            "item": "https://www.hyperize.ai/en/methodology/task-selection"
          }
        ]
      },
      {
        "@type": "Organization",
        "@id": "https://www.hyperize.ai/#organization",
        "name": "Hyperize",
        "url": "https://www.hyperize.ai",
        "logo": "https://www.hyperize.ai/hyperize-logo-full.svg"
      }
    ]
  },
  "facts": {
    "name": "Task Selection",
    "summary": "A customer wants to compare a product, request a quote or track a delivery. Task selection defines which of these journeys an AI agent should test, before measurement starts. It specifies the goal, access and permitted endpoint. Multiple tasks reveal different barriers; a single test remains a limited observation. This shows where Hyperize can improve the customer journey.",
    "boundary": "Intermediaries provide context, not an extra factor in the published score formula [S2].",
    "scope": "Rules and score weights are public. Internal prompts, scoring details and raw logs stay within the project. A planned stop before purchase or submission is not a technical failure. Examples do not establish current brand performance; rankings and citations are not guaranteed.",
    "url": "https://www.hyperize.ai/en/methodology/task-selection"
  }
}