Three Modes

Pick by stake. Not by curiosity.

Fast lookup. Full deliberation. Steered deliberation.

Pilot5 runs in three modes. The Expert is one smart-routed model answering a factual question fast. The A-Team is the full five-perspective panel deliberating automatically across two to four adaptive rounds. The Dream Team is the same five-perspective panel but with a human-in-the-loop pause after every round, up to six rounds.

The same panel architecture and audit guarantees apply across all three modes. What changes is depth, time, and your level of involvement. Cost scales with how much deliberation you actually consume — credits are reserved at start and finalised to actual token usage, with excess always refunded.

01 · 1 (smart-routed)

The Expert

One smart-routed model, seconds to answer, evidence-grounded.

Cost
0.4–0.7 credits
Time
≈ 8–18 seconds

A single benchmark-selected AI for the question type — pricing routing for cost, code routing for technical, strategy for analysis. Reuses the full Pilot5 knowledge stack (memories + documents + 250+ trusted sources + RAG). No multi-round deliberation, no Minority Report. Token-metered: you pay for what the model actually consumes, never more.

Best for: Quick factual lookups, calculations, definitional questions, anything where one good answer is enough.

02 · 5 (full panel)

The A-Team

Five perspectives, automatic deliberation, no interruptions.

Cost
3–5.5 credits
Time
≈ 6–9 minutes

The full five-perspective panel runs Divergence + anonymized Critique + Synthesis end-to-end without pausing. The adaptive orchestrator queues additional rounds (Devil's Advocate, Assumption Surfacing, Focused Research, Calibration) when a specific failure-mode signal fires — premature consensus, hidden assumptions, information gaps, or confidence drift. You walk away; the result lands as a chat message.

Best for: Strategic decisions, competitive analysis, investment evaluation, anything that needs cross-perspective stress-test but doesn't need you in the room.

03 · 5 + HITL

The Dream Team

Five perspectives plus you, with HITL pauses between rounds.

Cost
3.8–7 credits
Time
≈ 12–18 minutes

Same five-perspective panel as The A-Team — same R1 isolation, same anonymized critique, same adaptive-round catalog — but the orchestrator works harder and the loop stays open to you. Three differences from A-Team that earn the price gap: (1) HITL pauses after every round, deterministic, never threshold-gated — use them to add context, redirect angle, or pin a constraint. (2) Devil's Advocate fires earlier — at agreement ≥ 0.8 vs ≥ 0.9 for A-Team — so groupthink gets stress-tested sooner. (3) The orchestrator widens the calibration consensus window by 0.15 and guarantees at least one adaptive round before synthesis (Devil's Advocate, Calibration, or Assumption Surfacing) — you paid for deeper analysis, the system delivers it. Up to six rounds. Closes only when you sign off.

Best for: Board-level decisions, M&A diligence, regulatory strategy, anything where you need to steer the analysis as it develops.

What an HITL pause actually looks like

The Dream Team’s defining feature is a deterministic pause after every round. Not threshold-gated. Not “typically after R1.” Every round. Here’s the flow.

  1. 01
    Round completes. The orchestrator finishes the round, persists envelopes, computes agreement and confidence signals.
  2. 02
    Status flips to awaiting_feedback. The deliberation halts in place. No new persona calls fire. You see a chat-message summary of what the panel just produced and an invitation to steer.
  3. 03
    You write back. Free-text feedback (1–2000 chars) goes into the persisteduser_feedback field and is woven into the next round’s context. Add a constraint, redirect the angle, pin a fact the panel missed, or just tell the orchestrator to proceed.
  4. 04
    Resume. A compare-and-set on the row prevents two parallel submits from racing. The orchestrator picks up where it paused; the next round inherits your steering as part of the brief.

Up to six rounds means up to five steering moments before synthesis lands. The pauses are why The Dream Team beats The A-Team on decisions where the panel can’t infer your constraints from the question alone.

What lands when the panel converges

The synthesis output is a structured artifact, not a paragraph of prose. Every A-Team and Dream Team deliberation produces the same shape, archived in the audit trail and retrievable by ID.

Verdict
One of: GO · PIVOT · NO GO · INSUFFICIENT BASIS. The fourth is rare — the arbiter emits it when grounding is too thin to recommend either direction.
Confidence score
0–100, calibrated against post-critique evidence. Not raw model self-reporting.
Decision matrix
The trade-offs surfaced during deliberation, with their relative weights as the panel argued them.
Action plan
3–5 concrete next steps, each with action, owner, and deadline.
Information gaps
What the panel flagged as unknown but load-bearing. Often the most useful artifact — tells you where to do the homework you can’t outsource.
Falsification conditions
The conditions under which the verdict would flip. A way of marking the recommendation as testable rather than fixed.
Minority Report
When one perspective refuses to converge, that position is preserved separately — with its own verdict, confidence, reasoning, and falsifier. Not averaged. Not softened.
Testable predictions
Extracted from the synthesis with timelines (e.g. 90 days, Q3 2026), retrievable later via POST /v1/deliberations/{id}/outcome.
Provenance tags
Every claim carries [SOURCED] (traces to a verifiable reference) or [INFERRED] (analytical inference). Fabricated [SOURCED] tags are auto-downgraded to [INFERRED] during audit.

How to pick

  • Reversible decision? The Expert is usually enough. Save the panel for stakes that justify it.
  • Stakes you'd want a record of? The A-Team. Five preserved positions, calibrated confidence, Minority Report — defensible after the fact.
  • Stakes plus context the panel can't infer? The Dream Team. The HITL pauses are where you correct course before the synthesis bakes in.

Continue

  • Pricing comparison → — side-by-side credit cost, plan tiers, and what each plan includes.
  • How it Works → — the deliberation pipeline that runs inside The A-Team and The Dream Team.
  • The Panel → — the five perspectives that deliberate every A-Team and Dream Team question.
  • Audit Trail → — what every mode produces as a permanent governance record.