Skip to content

Reports, PowerPoint masters, and design hand-offs

Sonaloop can deliver research in two editable directions:

  • branded PDF and PowerPoint reports for stakeholders; and
  • a provider-neutral MCP hand-off for Figma, canvas, code, and document tools.

The two paths are independent. A PowerPoint master belongs to the workspace; a design destination such as Figma remains connected to the AI host, not to Sonaloop.

Upload a PowerPoint master

Workspace owners open Workspace → Templates and upload a .pptx file. The master must use a 16:9 widescreen canvas.

Sonaloop keeps the reusable presentation structure:

  • theme and theme fonts;
  • slide masters and named layouts;
  • layout backgrounds, fixed artwork, and placeholders; and
  • page size and document properties.

Concrete slides from the uploaded file are removed immediately. The stored template has zero content slides. Uploads with macros, embedded objects, ActiveX, or external links are rejected. Sonaloop inspects theme colors and fonts, placeholder geometry, and the available slide purposes. The settings page shows a simple Ready or Limited result plus the purposes the template supports; technical details stay in the advanced deck settings.

The same page accepts an optional PowerPoint style reference. Upload a typical, well-designed presentation when the master alone does not explain rules such as “yellow is used for markers, never text.” Sonaloop analyzes color usage, explicit type sizes, images, charts, tables, and visual density before removing every concrete example slide. It stores the derived rules plus a sanitized zero-slide package—not the example content. The master continues to own geometry; the style reference owns semantic usage guidance.

New PowerPoint exports choose the closest company layout for cover, agenda, section, content, comparison, personas, stimulus screens, image, and closing slides. Layout names are recognized in common languages; opaque agency names fall back to the portable structure of their title, body, picture, and multi-column placeholders. Report text and images populate native placeholders where available, while editable shapes and charts are fitted to the layout's content grid. PowerPoint-native text fitting prevents long localized copy from spilling out. Empty content and picture placeholders are removed, so the deck opens in edit mode without insertion prompts; narrow branded footer fields receive only a compact label that cannot collide with fixed artwork.

The master owns the exported presentation's backgrounds, fixed artwork, fonts, and colors. Sonaloop derives chart and card colors from its theme and layout artwork; it does not layer the separate workspace web palette on top. It also does not add its own canvas, logo, footer, or forced font. Every export runs a content-free fidelity check for unchanged master parts, slide bounds, layout usage, placeholder use, font overrides, and off-theme colors. Bright accent colors remain available for rules, markers, fills, and charts. If an accent does not meet readable text contrast—or the style reference establishes that it is never a text color—generated text uses the master's ink color while non-text artwork is preserved.

Build the presentation story

The master supplies the brand, but it does not write the presentation. Once a report is complete, Sonaloop gathers its method, findings, personas, screenshots, evidence, and limitations. The connected AI host uses those inputs and the method's presentation profile to create a reusable deck plan.

The cover first tells a cold reader what artifact or journey was tested, for whom, and which decision the test informs. The short main story then moves from conclusion to visual evidence to the concrete decision or next action. Pacing defaults to one slide per 60–90 seconds, so a ten-minute readout is normally six to eight core slides. Related proof belongs in one dense visual frame instead of a sequence of sparse finding slides. Depending on the method, this can include:

  • a decision dashboard that combines the decision, strength/denominator, and reasons;
  • the actual stimuli side by side;
  • the participating personas on one overview slide;
  • a before-and-after preference or confidence shift;
  • the current screen beside an editable replacement-copy/UI proposal;
  • quotes, charts, risks, and recommended next steps.

Individual response matrices, method notes, limitations, and a source index move to a short appendix. Detailed persona profiles are added only when they help the decision. Sonaloop does not create filler agendas for short decks or finish with an empty thank-you slide.

Headlines are written to stand alone outside the research team: normally three to eight words, never more than 10 words or 56 characters. Unexplained pronouns, internal process labels, and bare variant letters are avoided until the compared stimuli have been introduced.

Every slide has native PowerPoint speaker notes. They can hold the key takeaway, full talk track, source references, limitations, backup details, transition to the next slide, and suggested timing. The slide itself can therefore stay visual and concise without losing the researcher's reasoning.

This makes the deck a stakeholder projection rather than an engine trace. It does not copy report sections one-for-one. Full prose, source identities, citations, and audit detail remain in Sonaloop, while evidence references keep the presentation traceable.

The stakeholder PDF renders this same reviewed story rather than a shortened list of internal research phases. Compared stimuli stay together, personas appear as compact profiles, and the synthetic-persona interpretation note stays at the end. Reports created before a delivery story exists still show their available input screens and cohort before the authored report sections.

Replacing a master affects only new exports. Existing files do not change.

Customer-facing result states

The normal Cloud customer view describes what the user can do, not the internal health of an autonomous agent run. Once a durable report, synthesis, prototype, outcome, or deliverable exists, the job reads Result ready. A paused or expired internal run does not turn a useful result into an error.

Global run-attention widgets, run chips, journals, setup plans, and engine-health language stay out of the customer surface. Owners and operators retain that technical view for audit and recovery. A real missing user choice remains visible as Input required.

Continue in Figma or another design tool

The read-only get_design_handoff MCP tool returns a bounded sonaloop.design_handoff.v1 bundle for one project. It includes:

  • the research goal and selected reports;
  • the reviewed delivery story, when available;
  • compact persona context and lazy portrait access;
  • findings and persona voices with evidence references;
  • open questions, behavioral predictions, and decisions;
  • existing concepts, prototypes, flow screens, visual asset handles, and usability results;
  • evidence-linked editable revision proposals;
  • the active workspace's brand, colors, typography, layout, and chart tokens; and
  • a generic destination workflow.

The bundle does not contain local filesystem paths, run commands, destination credentials, or a hidden Figma integration. Visual sources include exact Sonaloop view_asset calls so the host can fetch only the screens it needs.

Connect Sonaloop and the destination MCP separately in the same AI host. Then ask:

Get the design hand-off for this Sonaloop project. Use the connected design tool to create or update the requested artifact. Preserve the workspace tokens, keep evidence references in the rationale, and leave unresolved questions visible.

The host reads the Sonaloop hand-off, fetches required screens, and writes through the destination MCP. Figma can therefore grant access to one file or library rather than to an entire account. The same workflow works with other design, canvas, code, or document tools.

If the destination produces a shareable interactive URL, the host can register it back in Sonaloop with register_remote_prototype and run the normal persona/usability validation loop. Sonaloop stores the URL as metadata and does not need the destination token.

Evidence boundary

A screenshot proves product state, not observed behavior. The hand-off preserves claim posture and source refs so a destination agent cannot honestly promote a hypothesis into a measured fact. Recorded usability sessions remain the behavioral evidence.