Now in private beta · Q3 2026 general release

See how stakeholders react
before you commit.

Arbiter pairs an adversarially-audited decision framework with a 100-agent market simulation. First it tells you what to do. Then it shows you what happens when you do it — how investors, customers, competitors, and the press actually respond.

Request access See a sample brief SOC 2 Type II in progress · AU & US data residency
Trusted by operators and advisors at
The problem

Decisions don't fail at the choice. They fail at the reaction.

The framework analysis is the easy part. The hard part is what happens after the email goes out — when investors recalibrate, customers churn, competitors price-match, and the team reads between the lines. Most founders find out in real time. Arbiter shows you first.

01 · Framing

The brief agrees with whoever wrote it.

Strategy decks rationalise. Founder updates omit. Advisors hedge. No one in the room is paid to argue the other side — until the round closes and the question gets re-litigated.

"Show me the case we'd lose."
02 · Reaction

You can't model people with a spreadsheet.

The ruling looks clean on paper. Then the lead investor cools on the strategy, two enterprise customers go quiet, and a competitor positions against you on LinkedIn. None of that was in the model.

"What were we actually optimising for?"
03 · Defence

"We decided this" isn't a document.

Six months later, when the call goes sideways, there's a Slack thread and a memory. No constraint scorecard, no sources, no simulation. Nothing to defend, nothing to learn from.

"How did we land here?"
How it works Part I · Arbitration

Five stages. One defensible ruling.

Every brief moves through the same pipeline — a constraint-driven framework that forces the analysis a tier-one strategy team would run. Each stage is a separate model with its own acceptance criteria.

01 · INTAKE

Constraint extraction

Parses your decision into hard constraints (cannot be violated), soft constraints (should optimise), decision criteria, risk tolerance, and time horizon.

01.5 · AUDIT

Adversarial bias review

A separate agent audits the constraints for preference leaks, post-hoc rationalisation, vagueness, and hidden objectives — then rewrites them neutrally.

02 · RESEARCH

Live evidence agent

Runs cited web research on pricing, competitors, and market signals. Five to ten findings, each scored for evidence strength against source credibility.

03 · ADVOCATES

Parallel argument

Independent agents argue FOR each option in turn — each addresses the constraints directly, surfaces its strongest evidence, and concedes its weaknesses.

04 · RULING

Arbitrator synthesis

A final arbitrator synthesises constraints, research, and advocate arguments into one recommendation with certainty score, risk register, and 30-day plan.

Part II · MiroFish simulation

Then we run 100 agents against your ruling. For 30 simulated days.

MiroFish builds a knowledge graph from your decision context, generates a stakeholder population grounded in real market signals, and lets them talk. Twelve rounds. No forced convergence. You see how your decision actually lands before it lands.

RULING Restructure decision A Lead investor +0.71 align B Board chair C Top 5 enterprise customers D Hiring pipeline E Series B prospects F Industry analysts G Tech press H Competitor · Aiko −0.58 align I Departing VP
Swarm verdict
Lead investor and board align quickly around the restructure narrative. Top customers stay supportive if the renegotiation is framed as roadmap investment. Aiko will counter-position publicly within 7 days; departing VP narrative is the most credible amplifier of risk. Get ahead of both before Round 5.
Who it's for

Built for the people who have to make the call.

Two audiences, one workflow. Arbiter is configured for the questions each role actually gets asked — not generic "AI for business."

For fractional CFOs

Twelve clients. One brain. Same standard of work.

Ingest each client's financials, retain context across briefs, and ship board-grade analysis in a fraction of the time. Co-branded output. Client-segregated workspaces.

  • 01
    Per-client workspacesIsolated context, financial documents, and brief history per engagement.
  • 02
    White-label briefsYour firm's logo, your tone, your sign-off. Arbiter stays in the engine room.
  • 03
    Variance & scenario packsMonthly board packs with consistent format, sensitivities pre-modeled.
  • 04
    Defensible methodologyEvery ruling carries source citations, the constraint scorecard, and the bias-audit trail.
Median time-to-board-pack 11 hrs90 min
For founders & operators

The brief your board would have written.

Pricing changes. Hiring freezes. Bridge vs. layoff. Channel bets. Submit the question, attach the financials, get a ruling you can actually defend — to your board, your co-founder, and yourself in six months.

  • 01
    Decision log per companyEvery ruling, every assumption, archived and searchable. A real paper trail.
  • 02
    Adversarial framingBuilt-in steelman of the option you didn't pick. Surfaces blind spots before the round does.
  • 03
    30-day implementation planNot just "what to do." A milestone schedule, owner per task, and triggers to reopen the decision.
  • 04
    Investor-ready PDFExport a brief you can send into a data room without a redaction pass.
Decisions logged per company Avg. 34 in first year
The output

A document, not a chat log. Twelve pages, signed.

Every Arbiter brief is structured the same way: a one-line ruling, the constraints it was scored against, the evidence base, the model, the counter-case, and a thirty-day plan. Designed to be read by a board chair on a phone or a partner on a Sunday.

§ 01Question presented & constraint scorecardpp. 1–2
§ 02Ruling, confidence, & material assumptionsp. 3
§ 03Evidence base & source manifestpp. 4–6
§ 04Financial model & sensitivity tablepp. 7–8
§ 05Adversarial review & steelmanpp. 9–10
§ 0630-day implementation plan & triggerspp. 11–12
Brief #2604-AX · 02 May 2026
Bridge financing vs. operational restructure
Prepared for: Hexa Labs Board · Confidential
Arbiter

§ 02 — Ruling

Arbitrated ruling · 84% confidence
"Defer the bridge. Execute a 14% headcount reduction and renegotiate the Sept '26 commercial agreement. Re-open this decision on the earlier of: ARR exceeds $7.2M or burn exceeds $480k/mo."

§ 04 — Sensitivity (next 12 months)

Scenario
Runway
ARR exit
Δ ownership
Bridge — flat
14 mo
$8.1M
−11.2%
Restructure
16 mo
$7.4M
0.0%
Hybrid
15 mo
$7.9M
−4.8%

§ 05 — Adversarial review

Steelmanned case for bridge: pricing power decays if competitors raise in Q3; restructure signals weakness to enterprise pipeline. Audited — partially mitigated by §06 sequencing.

Pricing

Pay per brief. Or by the engagement.

Three briefs a month are free, forever. Beyond that, buy a single brief, a five-pack, or a monthly retainer. No token math, no per-seat tax, no surprise overage.

Single brief
$99/ brief
One decision. One pipeline run. One full MiroFish simulation.
  • Full 5-stage arbitration
  • 100-agent MiroFish simulation
  • Stakeholder network export
  • Financial PDF ingestion
  • Investor-ready PDF export
  • Email support
Buy a brief
Best value
5-pack
$399/ five briefs
For a strategy cycle, a fundraise, or a quarter of board meetings.
  • Everything in Single brief
  • Five briefs · $80 each
  • Credits never expire
  • Shareable read-only brief links
  • Decision log & search
  • Priority email support
Buy the 5-pack
Monthly retainer
$499/ month
For founders and operators making decisions every week.
  • Unlimited briefs
  • Unlimited MiroFish simulations
  • Multi-company workspace
  • White-label PDF export
  • Slack-shared briefs
  • Dedicated reviewer on first 5 briefs
Start retainer
Frequently asked

What CFOs ask before they sign.

Is this just GPT with a prompt? +
No. Arbiter is a pipeline — six discrete stages, each with its own model, prompt, and acceptance criteria. The adversarial audit is a separate agent that can veto and rewrite the framing before the ruling is drafted. We benchmark every release against expert-written briefs from the same prompts.
What happens to my clients' financials? +
Uploaded documents are encrypted at rest, scoped to the workspace they were uploaded to, and never used for model training — ours or anyone else's. On the Firm tier you can elect AU-only or US-only data residency, and configure retention windows down to 24 hours.
How accurate is the financial modeling? +
Every model is shown with its assumptions, sources, and sensitivity bands — and labeled as Arbiter-generated. The brief is decision-support, not a substitute for management accounts. We expose the working file so you can audit and override any cell.
What does "ruling" actually mean — is it really a recommendation? +
Yes, and explicitly so. Arbiter is opinionated by design. Every brief commits to one option, scores its confidence, names the assumptions that would flip the ruling, and writes the triggers that would re-open it. You can override the ruling; you can't get a brief without one.
Who built this? +
Arbiter is built by operators with prior roles at top-tier strategy consultancies and venture-backed CFO seats. We work closely with a council of practicing fractional CFOs and Series A–C founders who pressure-test every release against live engagements.
Request access

Private beta. By introduction or invite.

We're onboarding a small cohort of founders, operators, and fractional CFOs before general release. Leave your email and we'll review within two business days.

No drip sequence · No card · Reviewed by a human