Overview

The Future Exploration Framework (T18 in Ora’s territory map) handles forward-looking questions — what could happen, what should we prepare for, what would derail this if we tried it, what odds should we put on this resolving one way or another. It is the territory that sits between decision-making (which is about choosing now among options) and risk analysis (which is about structural failure modes of a system or design). Future Exploration is the place where the future itself is the object of work.

The framework runs in five modes. Consequences-and-sequel is the lightest mode — a forward causal-cascade tracing the immediate, second-, and third-order effects of a proposed action. It is the de Bono lateral-thinking move, structured as a branching cascade rather than as scenarios or probabilities. Probabilistic-forecasting is the Tetlock-style mode — a thorough probabilistic forecast pass producing a numeric probability or range, with locked resolution criteria, reference-class base rates, inside-view drivers adjusted toward the outside view, and named leading indicators. Scenario-planning is the Wack/Schwartz mode — driving forces inventoried and classified into predetermined elements vs. critical uncertainties, two critical uncertainties selected as axes of a 2×2 matrix, four scenarios constructed as internally consistent narratives. Pre-mortem-action is the Klein-style mode applied to an action plan — imagined failure narrated backward, failure modes inventoried, leading indicators per mode, pre-commitment mitigations, residual unmitigated risks named. Wicked-future is the molecular composition that runs scenario-planning, pre-mortem-action, and probabilistic-forecasting and synthesizes via three stages into probability-weighted scenarios with named failure pathways and divergence-points-to-monitor.

The framework’s load-bearing intellectual content is the stance distinction (neutral forecasting vs. adversarial pre-mortem vs. constructive backcasting), the depth ladder (light projection → probabilistic forecasting → scenario planning → wicked-future molecular), and the parsing decision (Decision D) that splits Pre-Mortem into two modes — pre-mortem-action (T18, on the action plan) and pre-mortem-fragility (T7, on the system or design) — that share the klein-pre-mortem lens but operate on different objects. The parse matters because the disambiguating question is not “are we in T18 or T7” but “is this about an action plan that could fail or about a system that could break under stress” — different objects of inquiry surface different failure modes.

The framework is honest about what it does not do. Backcasting (the constructive-future mode where you imagine the success and work backward to today) is currently deferred per CR-6 — Future Exploration’s wicked-future molecular flags its absence rather than silently substituting. The framework is also honest about the difference between forecasting (which requires resolvable, time-bounded, substantively important questions per Tetlock’s triage) and scenario planning (which is for genuinely uncertain strategic horizons where the value of planning is preparation for multiple plausible futures rather than prediction). Choosing the wrong mode for the wrong question shape is the dominant misuse.

The framework answers questions like: I’m about to propose this — what are the second-order consequences I’m not seeing? What probability should I put on this resolving by year-end? My organization is making a 5-year strategic plan and the environment is genuinely uncertain — how should we structure that? Before we sign off on this campaign, what would have to go wrong? I want all three — narrative scenarios, calibrated probabilities, and stress tests — and I want them composed.

Systemic context

Future Exploration is the forward-looking territory of Ora’s territory map (T18, super-cluster C — Decision, Future, and Risk). It sits adjacent to T3 (Decision-Making, where the question is choosing among options now), T7 (Risk and Failure, where the focus is structural failure modes), and T9 (Paradigm Examination, where the framing itself is in question). Cross-territory dispatch from T18 typically routes to T3 when scenarios have been generated and the question becomes which path to take, to T7 when failure-mode work shifts from the action plan to the underlying system, and to the territory’s own wicked-future molecular when scenarios, probabilities, and pre-mortems all need to be composed. The framework’s pre-mortem-action mode shares the klein-pre-mortem lens with T7’s pre-mortem-fragility — they are parsed siblings per Decision D rather than a single dual-citizen mode, and the parse is what allows routing to dispatch by object (plan vs. system) rather than by lens.

Ora’s output

Example question: I’m running an indie software project. I’m about to launch a paid tier. Walk me through the consequences I’m not seeing.

Ora’s response (running Future Exploration in consequences-and-sequel mode):

The question is forward-looking, focused on consequences of a proposed action, no probability framing requested, no scenario-set requested. The lightest forward-cascade mode (consequences-and-sequel) is the right route. I’ll trace immediate, second-order, and third-order effects across multiple branches.

Immediate effects (first-order — within weeks of launch).

Conversion branch. Some subset of current free users converts to paid; some don’t. Expected pattern from the SaaS conversion literature: low-single-digit percent of free users convert to paid in the first month. The user-segment most likely to convert is the heavy-active segment (already getting value); the segment least likely is the low-engagement segment (getting little value, won’t pay).

Friction branch. Free users who decide not to pay encounter a friction wall they didn’t have before. Some accept it; some leave; some become vocal about the change. Expected pattern: the vocal subset is small but disproportionately visible (forums, social media, support channels).

Revenue-mechanics branch. Payment processing, tax collection, refund handling, subscription lifecycle management all activate immediately and cannot be deferred. Each carries an operational cost that did not exist before launch.

Second-order effects (within months).

Product-priority branch. Paid users have standing to demand features. Your roadmap reorients around their requests; the unpaid majority’s preferences become quieter signals. If you’re not deliberate, the product drifts toward what the paid segment wants — which may or may not be what makes it valuable in the broader market.

Brand-perception branch. Positioning shifts from “indie tool” to “indie business.” Some users read this positively (sustainability); others negatively (commercial capture). The negative reaction is louder.

Competitive-response branch. Adjacent free tools become more attractive to departed users; competitors notice the paid tier and reposition.

Audience-cui-bono branch. Users who valued the project specifically because it was principled-free read the launch as a betrayal of an implicit contract. Strength of reaction depends on whether prior communication suggested the free tier was permanent.

Third-order effects (within a year).

Identity-of-the-project branch. The paid tier becomes the load-bearing source of operational sustainability; roadmap decisions are filtered through “does this serve the paying tier.” The project that existed before the paid tier is gone, even if you did not intend to replace it.

Founder-attention branch. Time allocation shifts from product work to operations work — billing, refunds, tax compliance, customer-success touches. The fraction of your week available for the work that originally produced the project’s value goes down.

Successor-product branch. Other indie developers watch your launch as a case study. You become a reference case in your community whether it succeeds or fails publicly.

Branches the action’s frame doesn’t naturally surface.

Three parties are affected by the launch but are not visible from the launch’s own framing. (1) Contributors who never converted to paid users but were emotionally invested in the project’s success. They lose standing; the project they helped support is now organized around a constituency they’re not part of. (2) Users in regions where your pricing is high relative to local purchasing power. They are effectively excluded by the price point; the project’s accessibility footprint narrows. (3) Future-self. The you of 18 months from now operates a paid product with operational obligations to paying customers; the freedom to pivot the project, abandon it, or radically change its direction is reduced.

What this isn’t. This is a forward-cascade trace, not a probability estimate (no numeric forecasts on any branch) and not a scenario set (no narratives about how the future unfolds in alternative paths). If you want probabilities on specific branches resolving, the next mode is probabilistic-forecasting. If you want narrative alternative futures (paid tier succeeds-modestly, paid tier succeeds-wildly, paid tier fails-quietly, paid tier fails-publicly), the next mode is scenario-planning. If you want all three composed, the next mode is the wicked-future molecular. The cascade above is the lightest pass — it surfaces effects without committing to probabilities or alternative trajectories.

That is what Future Exploration produces in consequences-and-sequel mode: a structured cascade with branching, second- and third-order reach, scan for parties not visible from the action’s frame, and explicit naming of what the mode is and is not. The user can act on the cascade directly, or escalate to a heavier mode if more depth is warranted.

Commercial AI comparison

Comparison content auto-populates when the comparison-refresh framework runs against this question. Drafters do not author this section.

Brief comparison commentary

Auto-populates with the comparison content above.

How to use this framework

You can run the Future Exploration pattern with any AI of your choice. The composition is single-pass for any of the five modes.

The prompt:

[Paste the framework specification]

Run Future Exploration on this question.

Question: [Forward-looking question.]

Mode (optional): [Consequences-and-sequel / probabilistic-forecasting / scenario-planning / pre-mortem-action / wicked-future. If not specified, the framework infers from question shape.]

Action or plan or system in scope: [Whatever the future-question is about.]

The AI runs the within-territory disambiguation first if the mode wasn’t specified — Q1 (stance) for “likely consequences” / “probabilities” / “alternative stories” / “what could go wrong” — and routes to the appropriate mode. If the question warrants composition (multiple modes integrated), the framework escalates to wicked-future molecular. The output is mode-shaped: a cascade for consequences-and-sequel, a probability with range for probabilistic-forecasting, a 2×2 scenario set for scenario-planning, an imagined-failure narrative with mitigations for pre-mortem-action, an integrated package for wicked-future.

For best results:

  1. Be honest about whether the question is forecasting or scenario-shaped. Forecasting needs resolvable, time-bounded, important questions; scenarios are for strategic horizons under genuine uncertainty. Asking for a probability on “will my company succeed long-term” is the wrong shape for forecasting (unresolvable and unbounded); asking for scenarios on a campaign that launches next month is overpowered (scenario-planning is for multi-year horizons).
  2. Provide the action or system explicitly. The pre-mortem mode in particular needs the object — is the failure question about an action plan unfolding in time (T18 pre-mortem-action) or about a system or design with structural fragilities (T7 pre-mortem-fragility)? Naming the object up front saves a routing step.
  3. Don’t suppress the second-order branches. When consequences-and-sequel surfaces effects on parties not visible from the action’s frame, those are the load-bearing additions the mode produces. The first-order effects you mostly already know; the second- and third-order effects on parties outside your frame are why you ran the mode.

The framework is deliberately tool-agnostic. The mode structure, the pre-mortem parse, the Tetlock forecast discipline, and the Wack scenario method are conceptual disciplines that survive the lift to any environment.

Other examples

  • probabilistic-forecasting on a calibrated forecast. A user asks for a probability that a specific bill passes the U.S. Senate by year-end. Resolution criteria locked (passes Senate by Dec 31, 2026); reference class identified (similar bills with similar party-line support over the past decade — base rate 30%); inside-view drivers named (current vote count, minority-party leadership stance, schedule constraints); outside-view adjustment direction stated (toward base rate from inside-view optimism); probability range 25–35% reported; leading indicators named (committee markup completion, scheduled floor vote announcement). Demonstrates the mode’s discipline against the inside-view-only failure mode that produces miscalibrated single-point forecasts.
  • scenario-planning on a 5-year strategic horizon. A small organization is planning its next 5 years under uncertainty about regulatory environment and funding-source diversification. Driving forces inventoried via STEEP; predetermined elements separated from critical uncertainties; two critical uncertainties (regulatory tightening yes/no; funding base broadens yes/no) selected as axes; four internally consistent scenarios constructed (tight-and-broad, tight-and-narrow, loose-and-broad, loose-and-narrow); per-scenario implications and leading indicators named. Demonstrates the discipline against single-trajectory-dressed-up-as-scenarios — each quadrant is genuinely different in dynamics and implications.
  • wicked-future molecular on a major launch. A team is planning a product launch with high stakes. The composition runs scenario-planning (full), pre-mortem-action (full), and probabilistic-forecasting (full), then synthesizes: probability bands overlaid on scenarios respecting scenario-internal consistency; pre-mortem failure modes mapped to specific scenarios (not all failure modes appear in all scenarios); divergence points where scenarios branch identified concretely with leading indicators; the absence of backcasting flagged explicitly per CR-6 rather than substituted. Demonstrates the molecular composition’s discipline — three full components plus three synthesis stages plus explicit gap-flagging.

Citations

The framework draws on three source traditions. Forecasting comes from Tetlock’s Superforecasting (2015) and the Good Judgment Project research that established calibrated probabilistic forecasting as a learnable skill — outside-view base rates first, inside-view adjustments second, frequent updating, calibration tracking via Brier scores. The triage discipline distinguishes useful forecasting from forecasting theatre. Scenario planning comes from Pierre Wack’s work at Shell in the 1970s, popularized by Schwartz’s The Art of the Long View (1991) and developed methodologically by van der Heijden (1996) — driving forces, predetermined elements, critical uncertainties, 2×2 matrix construction, internal-consistency checks. The pre-mortem comes from Klein’s Harvard Business Review article (2007) and the prospective-hindsight psychology underlying it.

The forward-cascade tradition (consequences-and-sequel) draws on de Bono’s lateral-thinking moves and Stone’s Policy Paradox (2012). The Decision D parsing principle (pre-mortem-action vs. pre-mortem-fragility as parsed siblings sharing one lens) is internal to Ora and resolved the dual-citizenship question — parsing rather than dual-citizenship preserves routing precision while honoring the shared underlying lens. The framework was compiled 2026-05-01 from the territory map’s T18 entry; v1.0 with PFF-conforming structure throughout.

Downloads

  • Framework specification (PDF) — link to ora-ai.org canonical artifact when published
  • Framework specification (plain text) — link to ora-ai.org canonical artifact when published
  • Full white paper (PDF) — link when published