Reachy - Authoring Agent

1. Reading Questionnaire

2. Building Survey

3. Question Review (full)

4. Live: Mid-Flight

Build a survey from this questionnaire. Coffee brand tracker for Q3 2026.

📄 Coffee_Brand_Tracker_Q3_2026.docx

Got it. I'll read through the questionnaire, pull out every question with its routing instructions, and then build the survey. You'll get to review everything before I generate anything.

✓

Reading Your Document

0.8s

Parsed document structure, headings, and question tables

Sections4 sections found

Content312 paragraphs, 6 option tables

✓

Mapping Survey Structure

3.2s

Identified all questions, section order, and study objectives

Questions28 found

SectionsScreener → Brand Awareness → Satisfaction → Demographics

Study info3 objectives, 2 quota groups, 10min target

✓

Extracting Question Details

6.1s

Reading each question's text, type, options, and routing instructions

Q1-Q10 · Screener + Brand Awareness10 ✓

Q11-Q20 · Satisfaction Battery10 ✓

Q21-Q28 · Demographics + Close8 ✓

Quality Check

0.1s

All questions validated against your document. 2 items need your input.

✓ 28/28 complete✓ No duplicates✓ All matched

⚠ Q14 — question type unclear

⚠ Q21 — routing not found in document

Done reading your questionnaire. 28 questions across 4 sections. 2 items need your input:

Q14 — this has 6 brand attributes on a 1-7 scale. Should it be a grid (all on one screen) or separate scales?
Q21 — I couldn't find the exact routing instruction in your document, so I inferred it. Please verify.

Opening the review panel.

📝Authoring — Question ReviewMCP App↗

Questions

Sections

Needs Input

Ready

Full review in tab 3 →

Q14 should be slider_grid. Q21 routing is correct, approve it.

Updated Q14 to grid format. Q21 routing approved. Your questionnaire is locked in. Now building the survey.

✓

Loading Question Templates

5ms

Selected the right question templates for your study

Standard questions (single choice, multi-select, open text)

Scales & grids (rating, NPS, attribute grids)

AI-powered questions (follow-up probes)

Not neededmedia, conjoint, MaxDiff, visual semiotics

✓

Building Survey Questions

8.4s

Created all 28 survey cards from your questionnaire

Breakdown8 single choice, 4 multi-select, 3 attribute grids, 4 rating scales, 1 NPS, 3 open-ended, 3 AI probes, 2 transition cards

✓

Building Survey Flow & Routing

12.3s

Created the survey logic, skip patterns, and respondent routing from your instructions

Routing24 routing rules traced back to your document

Flow32 steps (28 questions, 2 branch points, 2 logic blocks)

⚠ 2 routing rules I couldn't trace back to your document

✓

Quality Assurance

4.8s

Ran 4 checks to make sure nothing was missed or misinterpreted

Objective Coverage — all research goals addressed✓

Survey Flow — every path reaches an endpoint✓

AI Probes — follow-up objectives are clear and relevant✓ 3 valid

Routing Accuracy — skip logic matches your document✓ 24/24

✓

Test Run

1.2s

Simulated 3 respondent paths through the survey

Complete path (main flow)22 questions → survey complete ✓

Premium segment path18 questions → survey complete ✓

Screen-out path (fails age check)3 questions → screened out ✓

Your survey is ready for review. 28 questions, 3 AI follow-up probes, 2 screeners. All quality checks passed and test runs completed successfully.

2 routing rules couldn't be traced back to your document — I'll flag those for you to verify.

Opening the review panel.

→ Survey Review MCP App opens here (separate mockup)

Example: Auto-correction in action

✗

Quality Assurance

4.2s

Found 2 issues — auto-correcting

Objective Coverage✓

Survey Flow — Q17 can never be reached✗

AI Probes✓

Routing Accuracy — Q9 missing an answer option in skip logic✗

↻ Auto-correcting — attempt 2 of 3

●

Fixing & Rebuilding

Re-reading your original document to correct the issues

Issue 1Q17: your doc says "If Q16 = Yes" but I routed from Q15 — fixing

Issue 2Q9: your doc says "Very positive OR Positive" but I only handled one — fixing

Here's everything I found in your questionnaire. Each question shows the type, text, and routing instructions exactly as written in your document. Flagged items are highlighted for your review.

📝Authoring — Question ReviewMCP App↗

Questions

Sections

Needs Input

Ready

Screener (Q1-Q4)

Q1	single_choice	What is your age? "Screen out if under 18 or over 65." Screener §1	✓
Q2	single_choice	How often do you drink coffee? "Screen out if 'Never'." Screener §2	✓

Brand Awareness (Q5-Q10)

Q5	multi_select	Which coffee brands are you aware of? Select all. "Show all 12 brands. Pipe selected to Q6." Brand Awareness §5	✓
Q7	nps	How likely to recommend [brand]? "Ask for top 3 brands from Q6. Loop." Brand Awareness §7	✓

Satisfaction Battery (Q11-Q20) 1 flagged

Q14	rating_scale?	Rate each brand attribute on a scale of 1 to 7. "Grid: Trust, Quality, Value, Innovation, Sustainability, Style. Scale 1-7." ⚠ 6 attributes on same scale → slider_grid or separate rating_scale? Satisfaction §14
Q15	ai_probe	Tell us more about your experience with [top brand]. "AI follow-up: probe for moments, emotions." Satisfaction §15	✓

Demographics (Q21-Q28) 1 flagged

Q21	single_choice	What is your household income range? "If income > $100k, route to Q22a (premium). Otherwise Q22b." ⚠ Routing inferred — no source text [INFERRED] Demographics §21
Q28	end_card	Thank you for completing this survey!	✓
+ 22 more questions (all validated ✓)

26 validated · 2 need attention

Q14 should be a grid — all 6 attributes on one screen. Q21 routing looks right to me, approve it.

Done. Q14 set to attribute grid with all 6 items. Q21 routing approved. Everything looks good — building the survey now.