The council seated, the issues a multi-model debate surfaces, where the models disagree, and the final readiness verdict. Here’s the shape of what you get — across different kinds of work.
Illustrative examples of structure and depth — not real customer work or outcomes.
What does Decidi produce?
Every council ends in a downloadable, sign-off-ready deliverable — an audit, a redline, a rewrite or a decision memo — with the issues ranked by severity, the points where the models disagreed named explicitly, a Final QA sign-off and concrete next steps. And when you need the finished file itself, “Build it out” produces it: a clean branded document, a real Word, Excel or PDF file, structured data, code, an image, or a short video — even a separate page for every row of a file you attach.
Auto-renews for a further 12 months unless cancelled 90 days before the term ends
Liability cap carves out “data loss” — leaving that exposure effectively uncapped
IP clause assigns improvements you make back to the vendor
Where the models disagreed
Two models read the indemnity as mutual; a third, on a closer reading, flagged it as one-sided. The moderator sided with the closer reading and recommended a redline.
Final verdict
GO-WITH-FIXES — sign only after three redlines (notice period, the liability carve-out, IP scope). Two clauses worth fighting for, one to accept.
Landing page review
A new product landing page, copy and screenshots provided.
The hero explains features before the outcome — a visitor doesn’t learn what they get in the first screen
No empty/loading state for the demo widget — it goes blank on slow connections
Primary CTA repeats four times but the pricing question it raises is never answered
Where the models disagreed
One model wanted the social-proof section moved up; another argued it would distract from the single call to action. The moderator kept one focused CTA and tightened the proof instead.
Final verdict
Three fixes before launch: lead with the outcome, add the missing widget states, answer the price objection inline. Strong visual system — leave it alone.
Pitch deck review
A seed-stage fundraising deck ahead of investor meetings.
The TAM is top-down and an investor will discount it immediately
Traction slide mixes vanity metrics with revenue — the real signal is buried
Unit economics don’t survive a second question about CAC payback
Where the models disagreed
The models split on whether to lead with traction or the wedge. The moderator recommended opening on the wedge for this audience, with traction as the proof beat.
Final verdict
Rework three slides: a bottom-up market, a cleaner traction story, and a defensible CAC payback. These are the questions partners will attack first.
P0 — the endpoint trusts the client-supplied content-type; an attacker can bypass the file-type check
P1 — no size limit, so a large upload can exhaust memory
P2 — errors are swallowed, so failures won’t surface in logs
Where the models disagreed
One model rated the missing size limit P0; the security engineer ranked the content-type trust higher as directly exploitable. The moderator ordered them by exploitability.
Final verdict
Block the merge until the content-type validation and size limit are added. Error handling is a fast follow. The core structure is sound.
Business decision review
Whether to hold prices or raise them 20% to fund growth.
The bull case assumes churn stays flat — the most fragile assumption in the model
No segment analysis: the increase may be fine for new customers but risky for the existing base
The downside scenario (a competitor holds price) isn’t costed
Where the models disagreed
The CFO favoured the raise for margin; the risk officer and Devil’s Advocate pushed back on churn exposure. The moderator recommended a segmented raise rather than a blanket one.
Final verdict
Raise for new customers only, grandfather the existing base, and watch churn for one cycle before extending. The blanket 20% is the riskier path.
Build it out
The finished file itself — in the format you need
Beyond the verdict, the council can produce the actual work — every item in full, ready to use.
A real build-out — structured data
Built by The Copywriter · captured live on the sample brief below
You hand it — messy notes
house blend, 14.99, our everyday medium roast, whole bean
Sunrise Decaf — 15.50, smooth swiss-water decaf, whole bean
ethiopia yirgacheffe 18, bright + floral single origin, whole bean
Cold Brew Kit - 24.00 - makes 8 cups, kit
espresso roast, 16, dark + bold, whole bean
Sampler Trio 21.00 three 4oz bags, kit
The council agrees the structure
Output every product as clean JSON — id, name, price_usd, category, one_line — consistent fields, every item, no placeholders.
Final QA signed off: “A fully populated JSON array containing all 6 products, each with consistent and complete fields matching the design — no placeholders, scaffolding, or missing entries.”
Branded document
A clean, on-brand HTML document — readable, printable, no raw markdown. The default for reports, memos and pages.
Word · Excel · PDF
Download any document deliverable as a real .docx, .xlsx or .pdf file — not just text.
Structured data
JSON or CSV — datasets, records and exports, every item complete and ready to import.
Code
Functions, SQL, scripts, config — runnable and copy-paste ready, in any language.
Attach a file with rows and build a separate finished item for every one — pages, metadata or records at scale.
Copy-paste blocks
A prompt, email or caption — the exact text to paste, with nothing to clean up first.
You pick the format (or let it choose), see the credit cost before you run it, and the work saves to your workspace to reopen any time.
In the app, each review arrives as a downloadable, on-brand document you can open, print and share — with the full debate behind it saved to your workspace to reopen any time.
Questions
Are these real customer examples?
No. These are illustrative examples of the structure and depth of a Decidi review — the council seated, the kinds of issues surfaced, where models disagree, and the final verdict. They are not real customer work or outcomes.
What does Decidi actually return?
A downloadable, sign-off-ready deliverable — an audit, a redline, a rewrite or a decision memo — with the issues ranked, the disagreement named, the Final QA sign-off, and the next steps. Not a chat transcript.
What formats can the deliverable be?
Beyond the verdict, “Build it out” produces the finished file in the format you need: a clean branded document, a real Word (.docx), Excel (.xlsx) or PDF file, structured data (JSON or CSV), code, an image, or a short video. Attach a file with rows and it can build a separate finished item for every row. You see the credit cost before you run it.
Can I run one of these on my own work?
Yes. Paste or upload your work, pick (or auto-seat) a council, and convene it. You’ll see the exact credit estimate before it runs, and the result is saved to your workspace.
Is the catalog example real?
Yes. It’s a live build-out run on the sample notes shown — not hand-written and not a customer’s data. Messy notes in, the complete JSON file out, checked and signed off by the Final QA audit. It’s the same “Build it out” you run on your own work.
