Coach persona and answer style — who the coach writes for and how replies are structured

The two dropdowns in your screenshot—Persona kouča and Odpovede kouča in Slovak, Coach persona and Coach responses in English—are per-project voice controls. They do not change your AI provider or model; they change who the coach writes for and how each reply is structured in both Chat and Agentic Strategy Pack runs.

Where you find these controls

You can set persona and answer style in three places—pick whichever is open while you work:

1. Agentic → Strategy run settings (tools menu on the Agentic panel) — labels Coach persona and Coach responses (your screenshot).
2. Chat → Composer settings (gear on the composer strip) — labels Reader persona and Coach responses.
3. Project details — same fields when editing offering focus, segment, and notes.

Wherever you edit them, the project keeps one persona and one coach response style—they stay in sync across Chat, Agentic, and Project details. While an Agentic run is in progress, those controls may be locked until the current phase finishes, so you do not change voice mid-run by accident.

On organisation (cloud) workspaces, everyone with access to the project shares the same saved persona and style. On a solo local workspace, the values stay with that project on your machine.

Coach persona (Reader persona)

Coach persona in Agentic is the same control as Reader persona in Chat: it tells the model which audience register to use—vocabulary depth, how much hand-holding, and whether to lead with decisions or teaching steps.

Built-in personas

Persona	Best for
General audience (Široké publikum)	Default—competent adult, jargon defined once, scannable structure.
Student / learner	Learning mode—shorter paragraphs, explicit prerequisites, stepwise explanations.
Manager / executive	Time-poor leaders—recommendation first, bullets, trade-offs, clear “so what”.
Senior practitioner	Dense, precise language; less hand-holding; assumptions still explicit when stakes are high.

Organisation custom personas

Admins can define extra personas under Company → Reader personas (name + coaching characteristics). In the dropdown they appear under Organisation. Use these when the whole company shares one voice—e.g. “Board pack neutral” or “Regulated industry cautious”.

What persona does not do

• It does not replace AI Skills (your personal instruction blocks in the composer).
• It does not switch specialist methodology—/market and /finance still use Strategy Pack prompts.
• It does not change Expert / Readable / Deliverable / Frameworks structure—that is the second dropdown.

Coach responses (answer style)

Coach responses controls how the coach structures each reply—pacing, narration, and whether output looks like coaching prose or finished artefacts. Specialist methodology (/market, /finance, and other slash commands), project context, and library material stay the same; only the reply shape changes.

Your screenshot shows Výstupy—the Slovak short label for Deliverable. English UI shows Deliverable.

Expert

• One focused move per turn—avoid questionnaire dumps.
• Assumptions, dependencies, and reasoning can appear in structured sections.
• Suited when you are exploring a domain in Chat and want the coach to teach frameworks while advancing.

Use when you want coaching narration alongside structured sections—not the default for new projects.

Readable

• TL;DR first, then at most a few clear sections with headings.
• Bullets over long prose; optional short Assumptions & trace block.
• Suited for busy founders reading on a phone—still correct, easier to scan.

Deliverable (Výstupy)

• Finished outputs only—tables, registries, diagrams, and structured choices when a decision is needed.
• No “Great question”, “I will…”, “let me know”, or “what I’ll do next” narration.
• Bold TL;DR up top; commentary compressed into the artefact or a tiny trace section.

Choose Deliverable when you want library-ready density—especially in Agentic runs where phases should produce assets, not coaching monologues between steps. It pairs well with General audience or Manager persona when the reader is a reviewer, not a student in a workshop.

Frameworks (Metodiky)

English short label: Frameworks; Slovak: Metodiky.

• Methodology canvases only—BMC, Value Proposition Canvas, PESTLE, ERRC, Porter, OKR tables, customer journey maps, and other Strategy Pack layouts rendered as interactive framework grids in the thread (Strategyzer-style sections, not plain markdown tables).
• Each canvas is followed by a References list tying items back to earlier artefacts (segments, drivers, prior canvases).
• No TL;DR, long coaching prose, or “what I’ll do next” narration—assumptions sit inside canvas items or references, not in separate essay blocks.

This is the default for new projects. Use it when stakeholders should see and edit professional framework layouts in Chat or Agentic—not mixed coaching essays. Pair with Manager or Senior persona when the reader already knows the methodology names.

Frameworks vs Deliverable: Deliverable still allows a bold TL;DR plus mixed assets (tables, registries, diagrams, structured choices). Frameworks is stricter: canvases and references only. Pick Frameworks for /market, /brain, and BMC/VPC-style work; pick Deliverable when you need a narrative TL;DR plus heterogeneous pack sections in one reply.

Side-by-side

Style	Structure	Coaching narration	Typical use
Expert	Sections + reasoning trace	Yes—paced, one question	Learning, intake, debate in Chat
Readable	TL;DR + headings + bullets	Light meta	Daily coaching, mobile reading
Deliverable	TL;DR + artefact body	Stripped out	Pack outputs, exec review, mixed assets
Frameworks	Interactive framework canvases + references	None—canvases only	BMC, VPC, PESTLE, journey maps, Agentic methodology phases

Changes apply from the next message after save—not retroactively on old threads.

How the two settings work together

Think of persona as for whom and answer style as in what packaging:

Example combinations:

• Manager + Frameworks — BMC, VPC, and phase canvases with references; default voice for new Strategy Pack work.
• Manager + Deliverable — board-ready tables and decisions when you need a TL;DR plus mixed assets, not canvas-only layout.
• Student + Expert — teaching cadence with explicit reasoning while learning /brain or /intake.
• General + Readable — friendly coaching for mixed teams in Chat.
• Senior + Frameworks — dense methodology output without coaching filler when the reader knows the frameworks.

Chat vs Agentic

Surface	Label for persona	Shared settings?
Chat	Reader persona	Yes—same project persona and coach responses
Agentic	Coach persona	Yes—Agentic may add stricter formatting for phase outputs

There is no separate Agentic-only personality—voice comes from these project settings plus your run brief, phase list, and library context. Settings → provider/model still govern which AI provider and model run each phase.

For the difference between conversational Chat and orchestrated Agentic, see Fundamental AI terminology in Sculptor.

What stays separate

Keep these distinct so you configure the right layer:

Control	Layer
Provider + model	Settings (which AI service and model)
Persona + answer style	Project (audience + reply shape)
AI Skills	Browser-local instructions (tone, taboos) — AI Skills in Sculptor
Slash commands	Which Strategy Pack specialist runs
MCP	Optional tools per turn or run

Skills and persona can both mention “write for executives”—if they conflict, prefer Strategy Pack process for phase methodology and use skills for personal or team taboos; persona sets the baseline register for everyone on the project.

Practical setup guide

1. Open Project details once per venture: set General audience unless the primary reader is clearly a student cohort or exec team.
2. Leave Frameworks as Coach responses for methodology-heavy work (default for new projects)—canvases render in the thread with references.
3. In Chat, try Readable for day-to-day coaching without canvases; switch to Expert when you want more reasoning visible.
4. Before a long Agentic run that mixes prose packs and canvases, try Deliverable + Manager; stay on Frameworks when every phase should output methodology canvases only.
5. If your org maintains custom personas, align the team on one organisation persona for customer-facing packs and document it in your enablement wiki.
6. Send a short test message after changing either dropdown—the hint in the UI applies from the next reply.

Limits to know today

• Not per-conversation overrides—only per project (use a different project or update project settings if two threads need different voices).
• Not synced via AI Skills—skills are personal to your browser; persona and coach responses follow the project (and your team’s cloud project on organisation workspaces).
• Does not disable structured choices—Deliverable and Frameworks still allow pick-one options when a single decision is required.
• Deep context toggle (Chat tools menu) is independent—it controls how much product/requirements context is included, not persona or coach responses.

Related reading

• Fundamental AI terminology in Sculptor — provider, model, prompt, skills, MCP, agents.
• From specialist chat to a library-ready Strategy Pack — when to use Chat vs Agentic after you set voice.
• AI Skills in Sculptor — personal instruction blocks on top of project persona.

These two dropdowns are small in the UI but shape every Strategy Pack reply your stakeholders read—set them once per project, then iterate when exports feel too chatty, too prose-heavy, or missing rendered canvases.