Mentat Desk
Decision intelligence for strategic teams

Make high-stakes choices you can defend

Mentat Desk structures every decision as a ledger of evidence, competing simulations, and a reasoning trace you can audit. The AI Mentat surfaces hidden assumptions, generates adversarial scenarios, and retrieves patterns from every past decision your team has made.

Open the app See how it works

Free plan available · No credit card required

Live from this instance

Active projects

3 workspaces

Strategic directions

7 arcs

Open decisions

12 questions

Data source

LIVE

AI features

4 active

Simulation modes

4 types

Active projects

3 workspaces

Strategic directions

7 arcs

Open decisions

12 questions

Data source

LIVE

AI features

4 active

Simulation modes

4 types

Why Mentat Desk

Structured reasoning beats
gut feel — every time

From noise to evidence

Every metric, research finding, constraint, and assumption gets classified and linked. Before you model anything, your Intel ledger tells you exactly what you know and what you're betting on.

Scenarios with explicit confidence

Competing hypotheses become concrete forward projections with confidence-adjusted expected values. The AI Mentat computes the gap — you make the call.

Decisions that survive scrutiny

Publish a signed brief with full reasoning trace. Board members, investors, and team leads can follow the logic — not just read the conclusion.

Step 1 — Capture and classify evidence

Structure raw signals before you reason

Paste in data points, research notes, customer calls, and constraints. Ask Mentat drafts signal templates tuned to the current key choice — every template is editable before you save.

  • Mentat generates Intel templates for any key choice in seconds
  • Instant Fact / Assumption / Constraint classification
  • Five Intel types: metric, event, research, operations, note
  • Document upload: extract Intel from PDF, DOCX, XLSX, transcripts
  • Every signal links back to its source document
Intel Ledger — GTM Pivot Q3 2026 5 classified
Metric

Mid-market NRR: 110% vs Enterprise NRR: 95%

Fact
Event

Lost $150K ACV deal — competitor had enterprise integrations

Risk
Research

Gartner: 75% of enterprise buyers require SOC2

Constraint
Operations

3 AEs (mid-market trained), 1 SE (entry-level)

Constraint
Metric

Burn rate: 14 months runway at current pace

Fact
Mentat classified 5 of 5 signals automatically
Simulation Comparison — 18-month horizon 3 branches
SimulationConfidenceARREV

Enterprise-first

45%

$2.5M

$1.1M EV

Hybrid motion

Recommended
70%

$2.2M

$1.54M EV

Mid-market only

85%

$2.1M

$1.79M EV

Mentat insight: Hybrid motion has 25pp higher confidence; expected value $1.54M vs $1.13M. Recommend Hybrid.

Step 2 — Model simulations

Compare outcomes with explicit confidence

Baseline + unlimited what-if branches. Each simulation carries a hypothesis statement, confidence level, and a chain of reasoning you can challenge step by step.

  • One-click 'Suggest scenarios' from current evidence
  • Confidence-adjusted expected value per scenario
  • Delta reasoning: see exactly what changed between branches
  • Scenario assumptions are linked directly to Intel items

Step 3 — Audit reasoning traces

Every conclusion is backed by an auditable chain

Instead of taking a recommendation on faith, step through the exact inference chain. Each step is tagged as fact, assumption, calculation, or model-derived — so you can challenge the logic before you commit.

  • Step-by-step reasoning chain with source attribution
  • Assumption flagging with risk labels
  • Update any assumption, re-run inference in seconds
  • Side-by-side trace comparison across scenarios
Reasoning Trace — Enterprise-first Baseline AUDITABLE
01

Hiring model

verifiedCRM data

2 AEs × 4 deals/qtr × $40K ACV = +$1.28M ARR

02

Retention model

assumptionFlagged

Enterprise NRR 95% + feature investment → 105%

03

Concentration model

calculatedSignal math

Enterprise 60% of customers → 75% of ARR; concentration → 35%

04

Confidence synthesis

derivedWeighted avg

Hiring 50% risky + feature 30% + market 20% → 45%

Conclusion: ARR $2.5M at 45% confidence

Shared Brief — board-ready · no login required PUBLISHED

Decision Question

Should we shift GTM focus to enterprise accounts?

Recommended: Hybrid Motion

70%

Hybrid motion balances enterprise growth with mid-market retention. Expected value outperforms enterprise-first by $440K.

Key assumptions

4 tracked

Identified risks

3 flagged

Contingency plan

Defined

Probability statement

72% · 5 fields

Live link · expires 2026-11-30 · read-only

Step 4 — Publish and share

Recommendation briefs stakeholders actually read

Package your key choice into a polished brief with a signed, expiring share link. Board members can explore the full reasoning trace without logging in.

  • One-click publish with signed token link
  • No login required for stakeholders
  • Configurable expiry — revoke at any time
  • Full reasoning trace visible in read-only mode

AI Mentat — four modes

Not a chatbot.
A co-reasoner.

The AI Mentat engages at four distinct points in your decision process — each mode produces structured output that directly affects your Verdict.

Phase 1

Probability Statement

Reads the scenario's reasoning trace and linked assumptions to generate a structured 3-part summary: a one-sentence conclusion, a confidence derivation, and the single reversal assumption that would flip the outcome.

  • Conclusion — one sentence, falsifiable
  • Confidence derivation — how the figure was computed
  • Reversal assumption — the bet that, if wrong, flips everything
  • Mentat advisory note — what to watch before publishing
Mentat Probability Statement — Hybrid Motion
Mentat

70%

confidence

The hybrid motion will grow ARR to $2.2M within 18 months with mid-market NRR sustained above 108%.

Derived from 4 high-confidence evidence signals; enterprise ramp assumption remains the primary risk ceiling.

Reversal Assumption

If the enterprise AE fails to ramp within 120 days, ARR drops $300K and the Hybrid scenario collapses to Mid-market parity.

Mentat advises: Watch AE ramp velocity as your primary leading indicator. If the AE closes one deal in 90 days, upgrade to Enterprise-first Simulation.

Phase 2

Adversarial Simulation

The Devil's Advocate mode identifies the 2–3 highest-volatility assumptions in any scenario and simultaneously inverts them to their worst-case values. The gap between source and adversarial is your true downside exposure.

  • Targets highest-volatility assumptions automatically
  • Worst-case projection saved as a new What-If branch
  • Mentat explains which bets were flipped and why
  • Gap between source and adversarial = downside risk
Hybrid Motion — Devil's Advocate Adversarial

Inverted Assumptions

AE ramp: 90 days → 180 days (historical enterprise failure rate: 60%)

Mid-market NRR: 110% → 98% (attentional dilution from enterprise pivot)

Enterprise NRR: 95% → 88% (premature entry without SOC2)

Adversarial ARR forecast

$1.4M at 22% conf

Mentat advises: The $800K gap is your true downside exposure. Add a contingency gate: "If AE has no closed deal by day 90, revert to mid-market focus."

Phase 3

Blank Slate Simulation

Reads only the decision title and objective — deliberately ignoring all Intel. Generates 3 first-principles hypotheses an outside advisor would consider, then scores each for novelty against your existing scenarios.

  • Bypasses existing evidence to prevent confirmation bias
  • 3 distinct first-principles hypotheses per run
  • Novelty score: how different each is from your existing simulations
  • Novel scenarios saved and labelled for follow-up
Blank Slate — First Principles Simulation Blank Slate

Generated without reading your Intel. The Mentat reasoned from title and objective only — no assumptions inherited.

Partnership-led growth

Novel · 94%

White-label partnerships with two system integrators generates $800K ARR without direct AE hires.

62%

PLG enterprise trial

Novel · 71%

Product-led growth with a 30-day enterprise trial converts 18% of trials to $35K ACV deals.

55%

Vertical specialisation

Partial · 38%

Healthcare vertical with HIPAA compliance unlocks a 2x ACV premium in a segment competitors ignore.

48%

Phase 4

The Vault

Searches all completed, outcome-indexed decisions across every project using semantic similarity. Synthesises them into calibration warnings — which assumptions your organisation historically over-estimates, and by how much.

  • Semantic search across all past decisions (pgvector)
  • Identifies which assumption categories caused calibration errors
  • Suggested confidence adjustment based on historical errors
  • Growing library: every completed decision improves future guidance
The Vault — 4 similar past decisions found Vault

In 3 of 4 similar enterprise hiring decisions, teams over-estimated AE ramp speed by 35% on average. Two of those decisions ended with ARR missing by 20%+.

Assumption Warnings

AE ramp timeline

over-estimated ~35% (3 decisions)

Use 120-day ramp default, not 90-day

Enterprise NRR uplift

over-estimated ~12% (2 decisions)

Reduce NRR uplift assumption by 8pp

Suggested confidence adjustment

−15pp

Vault advises: Reduce Hybrid scenario confidence from 70% to 55% before publishing a Verdict. Historical pattern is consistently over-optimistic on ramp speed.

The full journey

From raw question to signed brief

01

Capture Intel

Classify every signal as fact, assumption, or constraint.

02

Form hypotheses

Competing explanations with hidden assumption extraction.

03

Build simulations

Baseline + what-ifs with confidence-adjusted expected value.

04

Run AI Mentat

Probability Statement, Devil's Advocate, Blank Slate, Vault.

05

Publish Verdict

Signed brief. Frozen assumption ledger. Board-ready.

06

Record outcome

Forecast variance fed into Memory for next decision.

Full documentation

User Guide & Workflows

Step-by-step walkthroughs from first Question to published Verdict, with real-world examples for every workflow.

Read the docs
The Story Behind the Software

Welcome to the House of Mentats

In Frank Herbert's Dune, Mentats are the ultimate decision-makers — humans trained in structured reasoning, computation, and strategic foresight. They don't guess. They think.

Mentat Desk brings that discipline to your organisation. Every decision you capture becomes part of an evolving library that makes your team smarter, faster, and more decisive with every choice.

Think clearly

Signal classification and assumption tracking eliminate hidden bets and confirmation bias.

Reason systematically

Auditable step-by-step traces let you challenge logic before committing, not after.

Learn perpetually

Outcome reviews feed The Vault, improving calibration guidance for every future decision.

Step 6 — Learn and improve

Your forecasts get better with every decision

After execution, run an outcome review. Mentat computes forecast accuracy, surfaces which assumptions were wrong, and stores the lessons in your decision memory. The Vault uses these outcomes to calibrate guidance for similar future decisions.

  • Actual vs forecast comparison with variance breakdown
  • Which assumptions held — which were wrong
  • Decision Memory retrieval on future similar questions
  • The Vault grows richer with every completed outcome

Outcome Review — Enterprise GTM Pivot

MetricForecastActualDelta
ARR forecast$2.2M$2.15M−2.3%
Enterprise NRR95%97%+2pp
Mid-market NRR110%112%+2pp
AE ramp time90 days120 days+33%

Vault update: AE ramp assumption updated to 120-day default. Will surface on next enterprise hiring decision.

Pricing

Simple, transparent plans

Start free. Scale when you need. No lock-in.

Cost estimator

Team seats12
Operations / month1,600

Estimated monthly

$0

$0 / year

Best for teams proving strategic rigour.

  • Share links included
  • Full reasoning trace export
  • API-first integration
Mentat Desk

Start your first decision today

Free plan. No credit card. One decision is all it takes to see the difference structured reasoning makes.

Open Mentat Desk Read the docs Browse public decisions