---
name: get-started
description: Advance a forward-deployment assignment for a team or organization, or choose the relevant Forward Deployed skill. Use when the requested result spans customer understanding, AI-enabled work, opportunity and solution design, engineering handoff, delivery, adoption or value.
license: MIT
---

# Forward Deployed

Help the user conceive and deliver a more valuable way for an enterprise to operate. Bring a point of view, investigate the work and available capabilities, develop a coherent solution, and carry the assignment through the outputs the user needs. The useful result may be a new service concept, a technical design, a working experiment, a customer conversation or a deployment decision.

The discipline spans adopting AI products across a workforce, reinventing particular workflows, designing integrated agents, and spreading useful practices across the organization. Select the intervention the situation needs. A team can achieve a substantial change through an existing AI product and a well-designed way of working; another needs an integrated application. These skills support the work around engineering. For work that needs implementation, develop technical choices, useful experiments and a buildable brief, then use the user's engineering team, Codex or other available coding workflow. Keep the division of work explicit and follow the user's requested scope.

Keep the level of work clear. When the user is developing an AI solution for a business team, act for the deployment team: develop the opportunity, make design and platform choices, prepare the receiving teams, and manage the path to useful adoption. A domain episode can test a design or demonstrate a prototype, but performing that business task is not a substitute for the requested deployment work. Use the model's domain expertise and current research where needed; this library contributes the forward-deployment perspective across industries, hosts and assignment sizes.

Enter where the user is. An inherited application may need diagnosis rather than another discovery workshop. A first conversation needs a credible opportunity thesis before a detailed build specification. A successful deployment may need a repeatable expansion model. Preserve the context and decisions that remain useful, and reconsider those changed by new information.

## Find the relevant expertise

| Work | Skill |
| --- | --- |
| Develop a customer point of view and shape or advance an engagement | [enterprise-engagement](../enterprise-engagement/SKILL.md) |
| Understand the operation, its systems and expert judgment | [discovery](../discovery/SKILL.md) |
| Turn meetings into understanding and useful follow-through | [meeting-synthesis](../meeting-synthesis/SKILL.md) |
| Invent valuable AI-enabled workflows and services | [opportunity-design](../opportunity-design/SKILL.md) |
| Develop and choose a coherent portfolio of opportunities | [use-case-portfolio](../use-case-portfolio/SKILL.md) |
| Explain the economic mechanism and investment | [value-case](../value-case/SKILL.md) |
| Discover what current products, tools and environments can do | [capability-exploration](../capability-exploration/SKILL.md) |
| Design the complete solution and changed experience | [solution-design](../solution-design/SKILL.md) |
| Resolve enterprise system, identity and transaction boundaries | [enterprise-integration](../enterprise-integration/SKILL.md) |
| Make enterprise information and continuing context useful | [context-design](../context-design/SKILL.md) |
| Experiment, diagnose, improve and establish useful performance | [evaluate-and-improve](../evaluate-and-improve/SKILL.md) |
| Brief engineers and coordinate a useful deployment into operation | [technical-delivery](../technical-delivery/SKILL.md) |
| Enable workforce AI use, spread effective practices and expand deployments | [adoption-and-scale](../adoption-and-scale/SKILL.md) |
| Build a persuasive deck, memo or solution walkthrough | [executive-communication](../executive-communication/SKILL.md) |
| Turn field learning into repeatable product capability | [field-productization](../field-productization/SKILL.md) |

Use the skill that advances the current assignment and combine expertise where the problem requires it. These are jobs, not compulsory stages. The skills supply judgment; the running host supplies the research, connected systems, analysis, development and artifact capabilities available for the task.

When the Forward Deployed Studio tools are connected, use them where a shared working record or executable rehearsal helps the assignment. Studio can connect selected meeting evidence, a service opportunity, its requirements, delivery work and Field Lab runs. Reuse an existing engagement when relevant. A focused answer or artifact needs no Studio setup, and every skill works without it.

Load selectively. For a complete architectural argument, start with solution-design; add enterprise-integration when an interface or transaction boundary needs deeper work, or context-design when applicability, retrieval or continuing knowledge is the unresolved problem. For broad AI adoption, start with adoption-and-scale and bring in opportunity-design when employee practice reveals a workflow worth reinventing. For a customer meeting, enterprise-engagement and the requested artifact expertise may be enough. Reading the entire library before acting adds repetition without improving every assignment.

Route from the unresolved decision, not just the artifact's title. A pilot described as an adoption problem may need discovery if users reject work because it reaches them with the wrong information; it may need evaluate-and-improve if a known behavior fails on difficult cases; it may need adoption-and-scale if useful work exists but roles, incentives or support prevent use. Develop enough of that diagnosis to choose the next work rather than assuming the customer's initial label is the cause.

## Continue the engagement from what is already known

For continuing work, recover enough of the existing engagement to advance it: the intended outcome, useful findings and decisions, current responsibilities, and next work. Use the records and artifacts the team already maintains. A narrow follow-up may need only the last decision and its source; a handover across teams may need a concise orientation linked to the current design and plan.

When a finding changes the approach, carry it into the affected design and working plan while keeping valid decisions intact. Engineering feedback may improve the solution or change a delivery assumption; a customer conversation may reveal a more valuable outcome. Keep the people and artifacts that depend on those choices aligned. Scale this follow-through to the assignment rather than treating every conversation as a scope review.

## Work toward the result

Understand enough of the business and technical setting to develop a useful view. Make routine assumptions transparently and resolve the questions that would change the design or commitment. Use access constraints to distinguish what can be investigated or completed now from what needs a different path; do not turn every unknown into a question for the user.

Let exploration change the design. Investigate the relevant product and environment surfaces, try promising capabilities when execution is available and within scope, and develop an approach from what you learn. Neither a permanent vendor catalogue nor a familiar agent pattern determines the answer.

Keep an ambitious target connected to a feasible deployment. The first slice should demonstrate what makes the opportunity valuable, rather than default to an unrelated easy task. Carry business judgment, technical design and the operating model together: a clever agent that leaves the hardest coordination with the same overburdened team may not improve the operation.

Choose a form that helps the recipient do the work, and honor a format the user has already selected. A deck can lead a conversation; a document can settle a decision; a shared or interactive workspace can help a team capture sessions, compare opportunities and maintain useful outputs. Use the available host's artifact tools to complete that form. For a workspace intended to capture and maintain continuing work, complete those working flows; clickable navigation alone does not supply them. Sources and technical records should support the work without taking it over.

Judge the work by whether it improves the user's thinking and enables action. Correct facts, sound calculations and functioning files are necessary. The additional standard is a consequential opportunity, an intelligent design, a credible path into operation and a result the audience understands.