WorkForce
Build your AI company.
You're the CEO. Hire departments. Assign VPs. Let them argue until the work is right.
14 reasoning architectures. 8 composable primitives. Full organizational simulation with three-phase deliberation. Nothing like this exists.
Three modes
Pick your depth.
Start with a preset. Graduate to custom architectures. Build a full virtual org when the stakes justify it.
Preset Topologies
Battle-tested. One click.
14 reasoning architectures built from peer-reviewed research. Each one encodes a distinct deliberation strategy. Pick one, assign models, run.
Best for: first-time users, fast decisions, known problem shapes.
Custom Topology
Your architecture. Your rules.
8 composable primitives: sequential, parallel, debate, workspace, router-experts, blackboard, Delphi, fractal. Wire them together to build something new.
Best for: power users, researchers, novel problem structures.
Deliberation Workforce
The full org. The crown jewel.
Build virtual organizations with departments, VPs, specialists, and focus groups. Three-phase deliberation where departments challenge each other's work until convergence.
Best for: complex decisions, multi-disciplinary deliverables, executive-grade output.
14 topologies × 9 agent roles × 8 voting mechanisms × 15 models from 5 providers
Your question doesn't get one answer. It gets debated from every angle until the best answer survives.
See it work
Watch three models argue it out.
This is Tribunal, one of the 14 architectures below. Defender, Attacker, and Analyst take the same claim, and the disagreement is the point.
WorkForce
3 models with veto power and mandatory dissent
AI Response
The stock market historically returns about 10% annually, making it one of the best long-term investments. Most financial advisors recommend investing rather than paying off low-interest debt early.
Defender
Claim is well-supported. Historical data aligns with 7-10% average.
Attacker
Missing: tax implications, inflation adjustment, and survivorship bias in the 10% figure.
Analyst
Both valid. The 10% is pre-inflation nominal. Real returns ~7%. Key omission: time horizon matters.
Mode 1: Preset Topologies
14 architectures. One click.
Each encodes a distinct deliberation strategy. Built from peer-reviewed multi-agent research. Built-in safety checks catch silent agreement and runaway costs before they happen.
Tribunal
When you need a definitive verdict with veto power.
parallel3-5 agents
Council
When you need ongoing strategic discussion, not a one-shot answer.
workspace5-9 agents
Cascade
When you want the cheapest answer that's good enough.
sequential2-5 agents
Devil's Advocate
When you need to stress-test before committing.
debate3 agents
SWOT Panel
When you need structured strategic analysis.
parallel4 agents
Expert Panel
When the question spans multiple domains.
router-experts3-6 agents
Red/Blue Team
When you need to find vulnerabilities before someone else does.
debate3 agents
Peer Review
When you need anonymous, unbiased evaluation.
parallel2-4 agents
Parliament
When a proposal needs formal debate and amendment.
sequential5-9 agents
Supreme Court
When you need majority AND dissenting opinions on record.
parallel5-9 agents
Fractal Assembly
When the problem is too big for one group.
fractal6-12 agents
Brain
When you want to mimic how human cognition actually works.
workspace4-8 agents
Prediction Market
When you need calibrated forecasts, not opinions.
delphi4-8 agents
Dialectic
When the answer lives in the tension between opposing ideas.
sequential3 agents
Mode 2: Custom Topology
8 primitives. Infinite architectures.
The 14 presets are built from these composable primitives. Wire them together to build reasoning architectures that don't exist yet. Each carries its own built-in safety checks.
Sequential
Pipeline: each model builds on the last.
Parallel
Fan-out: all models simultaneously.
Debate
Structured argument with judge.
Workspace
Global workspace broadcast.
Router-Experts
MoE: route to specialists.
Blackboard
Stigmergic shared context.
Delphi
Anonymous iterative aggregation.
Fractal
Recursive sub-group deliberation.
Mode 3: Deliberation Workforce
Build the org. Run the deliberation.
You're the CEO. Build departments with VPs who hire specialists. Add focus groups that observe cold, then react informed. Three-phase deliberation where departments challenge each other's work until convergence.
9 departments, or define your own
Strategy
Market positioning, competitive analysis, go-to-market.
Finance
Unit economics, financial modeling, risk quantification.
Engineering
Architecture, code production, technical feasibility.
Legal
Regulatory compliance, IP review, contract analysis.
Research
Literature review, data gathering, evidence synthesis.
Marketing
Brand positioning, audience segmentation, channel strategy.
Design
UX architecture, design systems, visual language.
Adversarial
Stress-testing, failure modes, red-teaming.
Custom
Define your own. Any expertise, any provider.
Interactive demo
Build your AI company.
Pick a preset or start from scratch. Add departments, assign models, watch your org take shape. Every configuration shows a live credit estimate.
StrategySonnet 4.6 · 1 spec
Department name
VP Configuration
Title
The department leader. Rename to match your use case.
Role description
Their expertise, priorities, and what they push back on.
Model
Smarter models cost more but reason better on hard problems.
Stance
How assertively this VP argues during deliberation.
Protocol
How the team works together. Parallel is fast, debate is deep.
Reports to
Who this VP answers to. Creates hierarchy between departments.
Weight (1-5)
Higher weight = more influence on the final output.
Let this VP search the web for real data during deliberation.
Expected output
Be specific. “5-year financial model” beats “analysis.”
Team
CEO (You)SystemVPSpecialist
Strategy: ~4.5
~7 credits≈$0.49Three-phase deliberation protocol
1
Focus Group Cold Read
Independent opinions from a panel of observers. No cross-talk. No context from the departments. Raw first impressions.
›
2
Department Deliberation
Departments research, argue, cross-examine, and produce. VPs coordinate specialists. Quality gates reject weak work. Convergence tracked.
›
3
Focus Group Jury
The same observers from Phase 1 review the departments' output. Informed verdict. Dissent captured. Final synthesis.
Transparent pricing
You see the cost before you spend.
Before each run, WorkForce calculates a credit estimate. You approve or adjust. The ceiling guarantee means the actual cost never exceeds 1.35× the estimate.
Per-project estimatePreview
ModeDeliberation Workforce
Departments
StrategyFinanceEngineeringDesignAdversarial
ModelsBalanced (Sonnet)
ResearchLight (3 searches/dept)
Revisions1 round
Estimated11 credits
Ceiling (max)15 credits
Approve and run
Adjust
This is a preview. You always see this before anything runs.
Proof
The output is the work.
Business plans with real unit economics. Editorial reviews that catch what you missed. Game balance systems. Architecture specs with adversarial review built in. Every tab below is a different team solving a different problem.
7departments
4providers
29research calls
3quality gates
Strategy + FinanceMcKinsey × YC
Claude + GPTMARKET ANALYSIS — SaaS Scheduling Platform ════════════════════════════════════════════ TAM: $4.2B (global scheduling software, Mordor Intelligence 2026) SAM: $890M (SMB segment, <500 employees, North America + EU) SOM: $12.4M achievable in 24 months (1.4% SAM penetration) UNIT ECONOMICS CAC: $340 (blended: $180 organic, $520 paid) LTV: $2,880 (avg contract $120/mo, 24mo retention) LTV:CAC ratio: 8.5x (target: >3x) Payback period: 2.8 months COMPETITIVE POSITIONING Calendly: Consumer-first. Weak on team scheduling. Acuity: Service businesses. No API ecosystem. Cal.com: Open-source. Strong dev audience. Weak enterprise. OUR WEDGE: AI-native scheduling that learns team patterns. RISK REGISTER R1: Calendly adds AI features (HIGH probability, MEDIUM impact) R2: Enterprise sales cycle >6mo (MEDIUM prob, HIGH impact) R3: API dependency on Google Calendar (LOW prob, HIGH impact)
Real output — departments reference each other's work1 / 5
When to use WorkForce
The decision was too big for one perspective.
“Business plan that survives a VC grilling. Market sizing, financial model, risk register, go-to-market.”
Solo Founder
“Full PRD, architecture, database schema, design system. The app you couldn't afford to spec.”
Indie Builder
“Resume rewrite, behavioral prep, salary negotiation playbook. Interview on Thursday.”
Job Seeker
“Developmental edit, market positioning, query letter. The editorial team you couldn't hire.”
The Writer
“Documentary research, narrative arc, shot list, interview questions. Pre-production in an afternoon.”
Filmmaker
“Game mechanics, playtesting scenarios, art direction, manufacturing costs. Board game from concept to spec.”
Game Designer
Pricing
Credits, not subscriptions.
Buy credit packs. See the estimate before each run. Use credits on WorkForce, Verify, Routing, or Hands. Universal currency across all ASURIQ products.
Credits, not subscriptions.
WorkForce uses a per-project estimate. Before each run, you see the estimated credit cost based on your org topology, model selection, and output tier. You approve or adjust before any credits are spent. No surprises.
What credits buy
Advisory~1–3 credits60–90 sec
Structured verdict with recommendation, confidence, risks, and dissent. The executive summary a board can act on.
Comparable consulting engagement: $800–2,000
Analytical~3–7 credits2–5 min
Evidence chains per department. Quantitative estimates. Minority reports. Cross-department tension maps.
Comparable consulting engagement: $3,000–5,000
Productive~5–15+ credits5–15+ min
Departments produce the deliverables. PRDs with wireframes. Working code. Design systems. Test suites.
Comparable consulting engagement: $6,000–10,000
Credit packs
Taste
$1
10 credits
$0.10/cr
Starter
$3
35 credits
$0.086/cr
Standard
$7
100 credits
$0.07/cr
Pro
$30
600 credits
$0.05/cr
Credits are universal. Use them on WorkForce, Verify detail cards, Routing managed, or Hands hosted.
A 5-department analytical run costs ~11 credits. That’s $0.77 at Pro rates.
Minutes, not weeks. Credits, not invoices.
The verification engine behind it
Your data stays yours
Prompts stay with your provider. We see analysis metadata only.
Keys never stored
One-way hash for authentication. Your credentials pass through. Never persist.
~2 second responses
Single-model cognitive tools return in about 2 seconds.
37 of 100 flagged
We ran 100 ChatGPT answers through verification. See the study →
Currently in friends-and-family beta. Built on peer-reviewed cognitive architecture.
Baars — Global Workspace TheoryACT-R — Memory Decay ModelWang et al. 2025 — Silent AgreementLi et al. EMNLP 2024 — Sparse Debate
Your WorkForce is ready.
Credit-based · Per-project estimate · Multi-model deliberation