Azrivo / How it works
How Azrivo works
Azrivo is a web app for working a decision through with a team of AI personas — not people. Ask one specialist for a focused answer, compare several side by side in a panel, or set specialists debating a real decision — openings, rebuttals, and a neutral synthesis that names the crux and lands a decision memo with trade-offs, failure modes, and decision boundaries. Plan Mode is its own tool: hand it any goal — a strategy, a campaign, a launch, even a trip — and it researches, plans, and builds roadmaps, documents, spreadsheets, diagrams, and execution artifacts you approve. You keep the final call.
Most AI hands you an answer. Azrivo works the problem.
Three steps from a hard question to a call you can defend.
- 01 Type the decision Ask one specialist, compare a panel of takes, set specialists debating to a verdict, or give Plan mode a goal. Azrivo assembles the right specialists for your question — no setup.
- 02 They argue it out Openings, rebuttals, concessions. Real opposition on demand — with live research and working arithmetic where the question needs it — not a tidy list of “it depends.”
- 03 You get the call — and the work A debate lands a neutral verdict and a decision memo — the call, with its reasoning. Need the work done too? Plan is its own mode: hand it any goal — a strategy, a campaign, even a trip — and it researches, plans, and builds the documents, spreadsheets, and roadmaps once you approve.
Rather watch first? Replay a real run — no account needed.
One room. Dial the depth to the decision
Debate and Plan are the signature workflows. Need something lighter? @ask builds one specialist for a focused answer; @panel builds a few and compares their takes — the same room, dialed down. Every cast is built for your question; pick the depth the decision deserves.
-
Debate
Set specialists against each other; a neutral synthesis names the crux.
@debate … -
Plan
One line in, a working plan and real artifacts out.
@plan … -
Panel
Describe who you want — a cast is built for it, side by side (lighter).
@panel … -
Ask
A specialist built for your question, one answer (lighter).
@ask …
Most AI answers. Azrivo argues first.
Set specialists against each other and Azrivo runs a real debate — openings, rebuttals, and a neutral synthesis that names the crux and calls it, instead of a tidy list of “it depends.”
Each specialist is an AI persona with a defined lens, not a real person — so you get the clash on demand, without the scheduling cost of a human panel. The verdict is a sharp starting point to stress-test, not a ruling to obey.
A serious debate can surface
- competing specialist positions
- direct rebuttals
- concessions & revised views
- the unresolved crux
- a final recommendation
- a decision boundary — what would change the answer
- the key trade-off
- failure modes
- next steps
- the strongest case for the other side
The output — a decision memo, not a pros/cons list
- Recommendation. the call, stated plainly
- Key trade-off. what you give up to get it
- Decision boundary. the condition that would flip the answer
- Failure modes. where this goes wrong
- Next steps. the first concrete moves
- The other side. the strongest case against
One line in. A working plan out.
@plan turns a one-line request into an executed plan — it asks the questions that change the answer, shows the alternatives it rejected and why, waits for your sign-off, then does the work. You approve the plan and keep the judgment; the team produces the drafts to check.
- 01 Clarify. When the request leaves a real fork, it asks the few questions that would change the plan — before any work starts.
- 02 Weigh. One recommended path, plus the alternatives it rejected — each with a stated reason it lost.
- 03 Approve. The proposed plan lands for your sign-off. Nothing runs until you say go.
- 04 Deliver. The team executes and hands back real deliverables — a structured first draft to verify.
Plan Mode creates real artifacts, not just advice.
research briefs · strategy documents · scoring models · spreadsheets · roadmaps · governance frameworks · implementation checklists · diagrams · slide decks · illustrations · working HTML prototypes · export to PDF / DOCX.
@plan Should we open a second location for our cafe this year?
01 Clarify — when the request leaves a real fork, no work yet
- Is the real goal more total revenue, or proving the second-site model before you scale?
- What’s the rough monthly budget you’d commit to a second site?
- Do you already have a neighborhood in mind, or should the plan help you choose one?
02 Weigh — one recommended path, then the roads not taken
- Recommended — Open one nearby, model it first. A single site close enough to share staff and supply runs — prove the unit economics, then decide on more.
- Set aside — Open two at once. Doubles the cash at risk before you’ve proven a single new site.
- Set aside — Add catering / wholesale instead. A different business with its own operations — solve “should we expand” first.
- Set aside — Wait a year, raise prices. Protects cash but answers a different question than the one you asked.
03 Approve — the gate
Plan proposed — 6 steps across 3 tracks. Nothing runs until you approve. Approve, or tell me what to change.
04 Deliver — a first draft to verify, not to ship blind
- Break-even spreadsheet. Rent, staffing, and volume as inputs — so the numbers are yours, not ours.
- Location scorecard. Charts weighing neighborhoods on foot traffic, rent, and competition.
- One-page decision memo. The recommendation, the risks, and what would change the call.
- A slide to present. Walk a co-owner or lender through it — exportable to PDF or DOCX.
Every figure here is an AI-generated draft that models your inputs — it doesn’t assert market facts, and it can be wrong. Check the numbers before you commit.
Specialists can use tools when the work needs it
The specialists do more than talk. Depending on the task, they can research, calculate, verify, structure, and produce artifacts.
- Live research. Search current sources and cite them where it matters.
- Fact checking. Verify claims instead of relying only on model memory.
- Arithmetic & modeling. Calculate totals, compare options, and score trade-offs.
- Documents. Produce structured memos, briefs, PDFs, and DOCX files.
- Spreadsheets. Build scoring models, tables, and lightweight calculators.
- Visual artifacts. Generate charts, diagrams, roadmaps, and illustrations to support documents.
- Web artifacts. Produce working HTML prototypes and self-contained pages.
First-draft artifacts and working prototypes — reviewable deliverables, not production-ready guarantees. You keep judgment and verify anything high-stakes.
Bring it whatever you’re actually deciding
Work calls, personal ones, half-formed ideas. Azrivo remembers the context you’ve already given it, so the next conversation doesn’t start from nothing.
- Debate. “Should I take the offer that pays more but moves me off the tools?”
- Panel. “Have a marketer, a designer and a skeptic each review my landing page.”
- Plan. “Turn my idea for an online course into a launch plan I could actually follow.”
- Debate. “We can hire one senior or two juniors for the same budget — which breaks first?”
- Ask. “Help me understand why I keep putting off the one task that matters most.”
- Ask. “Pick up the project we were working on last week and tell me what to do next.”
Every one of these is a first message. Type yours on the home page.
See what Azrivo produces
Not prompts. Not generic advice. Real Azrivo runs — the prompt, the specialists, and the artifacts they produced. Debate decides. Plan executes. Artifacts prove it.
- Build vs. buy: adjacent market expansion — Debate run. A real @debate — a Risk Analyst, Financial Modeler, and Strategist argue acquire-vs-build, then a neutral synthesis lands a decision memo with a clear boundary. View example
- Automation roadmap — Plan run. A real @plan run turning a vague objective into a full automation operating system — intake framework, weighted scoring model, AI-readiness gate, prioritization matrix, phased rollout roadmap, and a governance loop. View example
- MarginPilot landing page — Artifact run. A real @plan run on a fictional product brief — it researched the competitive landscape, wrote the conversion architecture, generated its supporting visuals, wrote the copy, and shipped a self-contained HTML landing page. View example
Want to test your own decision? Try Azrivo free.
Bring it the decision you’re actually sitting on.
Type it on the home page and a team of AI specialists gets to work on it — ask, compare, argue it to a verdict, or hand Plan mode the goal. You make the final call.