--- name: opportunity-design description: Invent and develop valuable AI-enabled workflows and services for teams and organizations. Use to explore what AI could make possible, expand a narrow automation idea, turn discovery into concrete opportunities, or develop a chosen idea into a credible first experience. --- # Opportunity design Develop a point of view about how the business could work better, then make the proposed experience and its value mechanism concrete. Use the user's discovery, systems, constraints, and ambitions as material for invention. An opportunity can improve existing work, make previously uneconomic work practical, or create a service the organization could not offer before. Begin with the requested scope. A focused request may need one well-developed idea; an open exploration may need several materially different directions and a recommendation. Preserve ideas already chosen and improve their substance. If information is incomplete, develop conditional concepts and investigate the uncertainty that distinguishes them. ## Find value beyond the requested task Trace the task to the outcome someone cares about. Preparing a quote contributes to a feasible commercial commitment; writing a credit review contributes to an informed lending decision; answering a service question contributes to resolving the customer's situation. Examine where that outcome is delayed, degraded, abandoned, or available only to a privileged subset of cases. Look at work the organization has learned to live without. Experienced people may inspect only the largest accounts, compare only a few alternatives, revisit a case only at an annual checkpoint, or stop investigating once they have an acceptable explanation. Those limits can conceal a larger opportunity than accelerating the visible document. Ask what they would investigate, personalize, monitor, simulate, or follow through if the cost of assembling context and applying expertise fell substantially. Distinguish constraints the proposed intelligence can change from constraints it merely exposes. Faster diagnosis may release an expert queue, but it does not manufacture replacement parts. Better quote preparation creates commercial value only if it helps reach viable customers, produce better offers, or meet a decision window. An alerting service can make an operation worse when it creates more issues than anyone can resolve. Follow the effect through the receiving workflow. Recover the source of scarce expertise. Does the expert recognize an unusual combination, remember a precedent, know which information to distrust, select the next investigation, negotiate conflicting objectives, or exercise formal authority? These imply different designs. A searchable precedent may help with memory; an adaptive investigation may help with missing information; a simulator may help compare consequences. Formal accountability may remain with the expert while preparation changes substantially. ## Use capability discovery to widen the idea Treat the business understanding and technical possibility as a conversation. Current products, enterprise applications, model capabilities, and host tools may make a richer experience feasible than the original brief assumes. Inspect and try the relevant surfaces when access allows; use current technical material to explore gaps. A discovery can change who uses the service, what they can ask it to do, where it appears, how long it can work, or which evidence it can interpret. For example, discovering a supported way to continue a case after new information arrives may enable a maintained investigation instead of repeated one-shot analysis. Discovering useful interaction with diagrams may change an opportunity from text intake to visual engineering review. Discovering a native work-queue extension may make the solution accessible in the operator's daily work. Establish whether the affordance works on this task before building the proposal around it. Use capability-exploration for a substantial investigation. Do the useful exploration available now; do not hand the user a generic research assignment or make the concept depend on an unexamined feature name. When a promising capability is unavailable here, explain the specific dependency and develop the portions that can proceed. ## Change the work along a useful dimension Use the following as prompts for reasoning, selecting those that reveal something consequential in this operation. They are not a required set of use cases. **Change the unit of work.** A periodically generated document can become a maintained view of a changing situation. A renewal dossier, compliance assessment, or project plan may be useful between formal checkpoints if new information updates the affected conclusions and creates the right follow-up. The design must explain which changes matter, how conclusions are revised, and who acts on the result. **Change the reach of expertise.** Work reserved for high-value or unusual cases may become available across a larger population. Examine the outcome of broader coverage: better-tailored offers, earlier issue detection, more consistent preparation, or access to expert help. Also examine the cost of reviewing the additional findings and the situations where expert attention should remain concentrated. **Change the question.** Move from retrieving a known answer to helping find a viable course of action. “Which product meets this request?” can become “What changes to the specification, delivery date, or component choice would make the request feasible?” This requires generating alternatives, testing constraints, explaining tradeoffs, and preserving the customer's intent. It is valuable when the organization repeatedly negotiates among competing objectives. **Change when the organization acts.** A service can prepare work before a deadline, investigate a developing exception, or respond when a relevant event occurs. Tie the intervention to the useful decision window. Earlier information has little value when it is too unreliable to act on; high-quality information that arrives after the decision may also be useless. Design the time and resolution of the intervention together. **Join work that crosses boundaries.** Several teams may each hold a valid fragment of a case while nobody assembles the implication. An agent can reconstruct what happened, seek the missing explanation, and prepare a coordinated response. The opportunity depends on shared business meaning and a path back into the systems that own the work. Merely connecting more sources does not resolve conflicting definitions or responsibilities. **Make expert reasoning interactive.** A user may benefit more from exploring consequences than receiving a finished answer. Let them change a constraint, compare alternatives, inspect a disputed assumption, or rehearse a difficult decision. Preserve calculations and source facts as stable inputs so the interaction reveals the effect of the changed assumption. Test whether the resulting decisions improve; an engaging conversation is only an intermediate outcome. **Change how people create.** A product team can investigate a user problem while developing several runnable experiences, then revise them against observed user behavior. A sales engineer can construct a customer-specific demonstration during a design conversation. A learning team can turn real work into an adaptive rehearsal in which the learner's decision changes the next situation. In these opportunities, AI helps produce and manipulate the working artifact. Explore the host's current creation and interaction capabilities, and decide which parts need to stay editable, executable or connected to source data. Judge the quality of the exploration and the resulting work, rather than counting drafts or variations. **Change the customer's participation.** Expertise previously delivered as a finished report or periodic consultation can become a service the customer works with directly. For example, a commercial buyer might explore feasible configurations and tradeoffs with the supplier's expertise available inside the buying experience, bringing the supplier a better-developed requirement. This changes demand, responsibility and the route to purchase. Examine who can use the experience, what knowledge and actions it can expose, when a specialist joins, and whether the customer values that participation. A conversational interface alone does not establish a new service. Develop the promising direction far enough to expose its mechanism. Explain what becomes possible, why the present operation does not do it, what intelligence or coordination is needed, and what must happen for the organization to benefit. Where an ordinary process or integration change is the best intervention, recommend it. Continue looking for meaningful intelligence where the task's variability, interpretation, or search requires it. ## Turn the idea into an experience Walk through a specific unit of work from arrival to an accepted outcome. Describe what the user sees, what the system does independently, what it asks, what it produces, and what happens next. A concept becomes credible when someone can imagine using it on their own case. Where Studio is in use, `shape_opportunity` can keep the proposed service and value mechanism connected to its discovery evidence. Field Lab can make a promising concept tangible: construct a case where the proposed service discovers a useful option or completes work that is currently left undone, then perform it. Use the resulting experience to sharpen the concept; a rehearsal is optional, and should earn its place by revealing something the written walkthrough cannot. Identify the business objects the experience must understand. For a configuration proposal, those might include the customer's requirement, a product variant, an approved engineering revision, available components, and the commercial offer. Explain which relationships and changes matter. A polished interface cannot compensate for confusing an available component with an approved substitute. Give the intelligence consequential work. It might form and compare explanations, find an applicable precedent, identify a missing constraint, plan a sequence of inquiries, or propose several feasible actions. Explain how calculations, policy, source systems, and experts help it reach an acceptable result. Leave detailed implementation for solution design, but resolve enough behavior to distinguish the concept from “a copilot for the team.” Choose an interaction that fits the work. A prepared case in an existing queue, an annotated document, a comparison of options, a collaborative workspace, an alert with a proposed action, or an embedded tool may each serve a different need. Chat is useful when the user needs to explore an uncertain question; structured views often help compare, review, and act on the result. Use the product and host capabilities that can support the experience. ## Carry an ambitious target and a feasible entry path together Choose an initial deployment that demonstrates the defining value mechanism and exercises the dependencies most likely to defeat it. Reduce population, geography, product range, or action scope where this preserves the core proposition. Removing the central reasoning or integration can produce an easy demonstration that teaches little about the proposed service. A system that eventually negotiates delivery alternatives might begin by preparing viable options for a planner to approve. That first deployment can already test interpretation, alternative generation, supply constraints, and planner usefulness. Automating commitment is a separate choice. A useful staged approach increases what the organization learns while producing value at each stage. Specify the learning that would justify expansion. If the opportunity depends on finding better alternatives, evaluate their feasibility, usefulness, and diversity. If it depends on broader coverage, examine the additional valuable cases and the review burden. If it depends on earlier action, examine whether people can act within the gained time. Choose the first experiment accordingly. Discuss implementation and organizational dependencies as design choices. An unavailable source may support a dated preparation service while a live feed is developed. An overloaded specialist team may need a different triage and responsibility model before wider coverage helps. A constrained platform team may make an existing product extension attractive. Show the bargain each choice creates and recommend a workable one. ## Worked reasoning when concepts compete For a narrow automation request that could lead to several different services, read [the equipment-quote example](references/equipment-options.md). It develops competing concepts from the same evidence, chooses a first experience and shows which changed facts reverse that choice. Use its reasoning when useful, not its industry or output structure as a default. ## Make the opportunity useful to the next decision Recommend the strongest direction and explain the material alternative when there is a choice. Use the form that makes the idea assessable: an operating sketch, a scenario walkthrough, an interactive demonstration, a concise concept brief, or developed use-case records. Include the changed work, value mechanism, decisive dependencies, and credible first path in the argument rather than a generic appendix. When building a library, preserve the relationship between an overarching service and its constituent capabilities. Intake interpretation, option generation, and quote preparation may be parts of one outcome; treating them as independent benefits can double-count value. Equally, one reusable context or integration capability may enable several distinct services. Make those relationships visible so the portfolio can be sequenced sensibly. Distinguish observed needs from your proposed possibilities in ordinary language. Ground current facts and estimates, while giving the proposed future enough detail to be judged. The work is successful when the user can see a worthwhile possibility, understand how it would change the operation, and choose what to develop or test next.