Digital Innovation Agents

Appearance

Your first Business Analysis ​

Let's run a real business analysis. We take an idea, scope it, explore the problem, design a solution, and end with a Business Analysis document that hands off cleanly to Requirements Engineering.

The example idea ​

"I want to build a tool that helps distributed teams run better async retrospectives."

This is a typical starting point: a rough idea, no clear user, no hypothesis, no scope. Exactly where /business-analyse is most useful.

Methods you will see referenced below

The agent's main job is to notice gaps in your understanding and propose the right research or prototyping method to close them. The callouts through this tutorial link straight to the method cards under Discovery methods, Ideation methods, and Validation methods. You can read them now or wait until the tutorial walks you into the spot where one becomes relevant.

Step 1: Invoke the skill ​

In your AI coding tool (Claude Code, Cursor, Codex, OpenCode, or Gemini CLI):

/business-analyse

I want to build a tool that helps distributed teams run better async
retrospectives.

Step 2: Determine the project scope ​

The skill starts with a scope question:

Before we go into detail: what is your project purpose?

A) Simple Test / Feature
   -> Timeframe: hours to 1-2 days

B) Proof of Concept (PoC)
   -> Prove technical feasibility, 1-4 weeks

C) Minimum Viable Product (MVP)
   -> Functional product, 2-6 months

Pick B) PoC for this walkthrough. The skill now adapts the depth of the analysis: shortened Exploration, full Ideation, hypothesis-based Validation.

Step 3: Exploration phase ​

The skill asks targeted questions, one at a time. For a PoC, expect 8 to 12 questions covering:

  • Users and personas: who is affected? Who pays? Who blocks?
  • Needs: what functional, emotional, and social needs exist?
  • Insights: what did you learn from real user statements?
  • Trends and technology: what is happening in the space?
  • How-Might-We: the synthesis question that bridges problem and solution

Important: the skill co-creates artifacts with you. It proposes a draft persona and asks you to confirm or correct. It cites which user statement an insight comes from. Nothing is invented in the background.

If you cannot answer "who is the user"

Before drafting a persona, the agent will likely propose a field method so you can come back with real evidence:

  • Explorative interviews. 7 to 10 short conversations, 20 to 30 minutes each, to map the problem space when you do not yet know which segment matters.
  • Qualitative interview. One deep conversation, 60 to 90 minutes, when you already know who to talk to and need depth.
  • Stakeholder map when several departments or external parties are in the picture and you do not know who to interview first.
  • Extreme users when average-user interviews produce generic answers and the real driver has not surfaced yet.

If you have interview notes but no pattern

The agent will help you move from raw notes to a usable persona:

  • User motivation analysis to pull functional needs, emotional needs, and obstacles out of a pile of transcripts.
  • Persona synthesis cluster to turn clusters of insights into persona seeds.
  • Persona to finish each seed into a one-page reference everyone on the team can check decisions against.

If users describe an ideal workflow instead of the real one

Interviews alone will not get you past this. Two observational methods fit:

  • Fly on the wall. Silent observation in the real context. You learn what users do, not what they say they do.
  • Self-test. Walk the user's process yourself. Uncomfortable but fast.

If you cannot name the competitors or the market context

  • Market and trend analysis. One day of desk research, two hours of clustering, four to six themes with one potential field each.
  • User journey when the friction is really about the user's experience over time, not the market around it.

At the end of Exploration, you get a How-Might-We question like:

How might we help distributed product teams run retros that surface root causes, so action items actually ship?

Step 4: Ideation phase ​

The skill now shifts from understanding the problem to designing a solution. For a PoC, expect 8 to 10 questions:

  • Solution description and object model
  • Idea potential on three axes (Value, Transferability, Feasibility), scored 0 to 10
  • The Wow: which feature would you want the press to celebrate?
  • Critical hypotheses: what must be true for this to work?
  • Value proposition: the formal statement

If the solution space feels empty or keeps repeating itself

The agent will suggest a short ideation exercise. Which one depends on your team and the shape of the blockage:

  • Brainstorming. A 15 to 20 minute group session once you have a sharp HMW. Best when the team is willing to defer judgement and fast-talkers can be kept in check.
  • Brainwriting. The silent 6-3-5 variant, useful when previous brainstorms were dominated by one or two voices.
  • Inspiration cards when the team keeps reproducing the same three or four variants and you need a jolt from outside the domain.
  • Idea tower when a seed idea is promising but too thin to prototype. Additive only, no removals, until the concept is coherent enough to sketch.

If you cannot explain why users would switch

  • Jobs to be done. Name the functional job, the emotional job, and the social job for the target user, along with hiring and firing criteria. This is also the method the RE skill will rely on when it drafts user stories for the functional, emotional, and social layer.
  • Kill your company for the inverse lens. Pretend a startup is attacking you. The weaknesses they would target become the value proposition you need to defend.

If you have too many ideas and no shortlist

  • Idea clustering and selection. Cluster by theme first, then score the clusters against four criteria (User Value, Feasibility, Transferability, Risk), and pick the top three with evidence you can defend.

Example output of Idea Potential:

Value/Urgency:    8/10 (retros are a well-known pain point)
Transferability: 9/10 (applies to any distributed team)
Feasibility:     6/10 (async UX is tricky)

Step 5: Validation phase ​

For PoC scope, Validation is shortened to hypothesis prioritization and feasibility. You list the critical hypotheses, prioritize them, and define test methods.

If a critical hypothesis needs a test plan

The agent matches the hypothesis to the cheapest method that would actually falsify it:

  • Wireframes, storyboards, and paper prototypes. Low-fidelity sketches of the risky flow. Best as a first test when the flow itself is the hypothesis.
  • Wizard of Oz when the feature is expensive to build (AI, automation, complex backend) and you want to see if users would even use it.
  • Appearance prototype when the flow is clear and you are testing brand or visual trust.
  • Expert review when the question is "can this even exist" (regulation, technical feasibility, safety) rather than "do users want this".

If you are unsure about the business side

  • Business plan. A one-page Business Model Canvas with one sentence per cell and a number in every revenue and cost cell. Mark unsupported assumptions as the next hypotheses.
  • Value proposition quantification to compare two or three candidate value propositions against evidence instead of gut feeling.
  • Pre-mortem before committing resources. "It is six months from now and the project has failed. Write down why." Five minutes silent per person, then cluster the reasons and assign owners to preventive actions.

Step 6: Produce the documents ​

The skill now creates two artifacts:

  • _devprocess/analysis/BA-retrospectives.md: the full Business Analysis
  • _devprocess/analysis/EXPLORE-retrospectives.md: the Exploration Board

Both follow the templates in skills/business-analyse/templates/.

Step 7: Handoff ritual ​

/business-analyse ends with the mandatory 3-part Handoff Ritual:

  1. Artifact report: lists the produced files and the HMW question
  2. Handoff context: appends an entry to _devprocess/context/30_handoffs.md with scope, personas, critical hypotheses, and open assumptions
  3. Transition question: "BA complete. Shall I start /requirements-engineering now?"

If you say "yes", the next phase starts automatically. If you say "stop", the skill pauses and you can review the BA before moving on.

What's next ​