Leading Through Ambiguity When the Future Can't Be Forecast

1. When the Forecast Is Not the Problem

On large capital projects, stage-gates and investment packs still demand a single story: one demand path, one regulatory climate, one operating context, one preferred narrative that can be priced, scheduled and defended. That habit is not foolish. Boards, lenders and public sponsors need something determinate enough to approve. The difficulty appears when the wider field around the project is already moving faster than the pack can absorb. Regulatory regimes shift mid-shaping. Markets that once looked stable become contested. Technologies that seemed peripheral suddenly redefine what “done” means. Coalition parties disagree not only on numbers, but on what the situation itself is.

In those conditions the failure mode is often misdiagnosed. Teams double the Monte Carlo runs, refine the contingency, or polish another version of the same business case. Sometimes that work helps. Sometimes it does not — because the problem was never a bad number inside a shared frame. It was clinging to one frame while the contextual environment moved, or while rival parties held incompatible meanings of the same facts. Forecast quality is the wrong target when meaning itself is contested.

This essay treats that distinction as the entry point to leading under ambiguity: when more prediction cannot restore clarity, leaders need a disciplined way to hold several plausible futures long enough to reframe the present and test commitments before they harden into irreversible capital.

This comprehensive blog will explore how sponsors, project directors, PMC leads, joint-venture and steering members, and strategy cells can lead large capital projects when the future cannot be forecast with ordinary confidence. It will first separate ambiguity from uncertainty and related labels so the wrong tool class is not applied. It will then outline turbulent field conditions that weaken predict-and-control habits, and show how to sort issues that belong in day-to-day interfaces from those that belong in contextual scenario work. Building on ideas from the Oxford Scenario Planning Approach as developed in Strategic Reframing — credited in the References and used here as an idea source, not as reproduced method text — the essay will treat scenario work as a service to decision readiness: name who must learn and for what purpose, build a small set of plausible contextual futures, iterate reframing and reperception, and wind-tunnel options before locking a single narrative. Application to large capital projects is labelled transfer throughout: reasoned extension from strategic-scenario practice into project settings, not measured field evidence from IPA databases or randomised trials of scenario planning. Brief bridges point to companion ProTrain pieces on deciding the question first, smart teams and estimate hygiene, and governability under surprise, without repeating those toolkits.

2. Brief Bridges — What This Post Is Not

Three neighbours sit close enough to confuse. Keeping them distinct protects each method.

Decide the Question First. Framing Before Choosing owns and structures the live decision question — what is being decided, by whom, against which criteria — before options are ranked. That discipline remains essential when parties cannot even name the choice. This post starts one layer out: when the question is usable but the future contexts that give it meaning are contested or moving, alternate future-context frames are used to reperceive the present. The two stack; they do not replace each other. See Decide the Question First.

Smart Teams and estimate hygiene. Biased estimates, uniqueness bias and related hygiene problems typically operate inside a largely shared frame. Outside-view and reference-class tools belong after frames stabilize. Pouring those techniques into a room that still disagrees on what the situation means wastes analytic energy. Hand quantitative estimation to those toolkits once meaning contests quieten.

Governability. On large engineering projects, governability is a property of the coalition under surprise — whether parties can hold together and renegotiate when the unexpected arrives. Parties may exploit ambiguity for advantage, but governability is a coalition and institutional toolkit, not a scenario method. Scenario conversation can surface contested meanings; it does not substitute for coalition design.

Complexity vocabulary. Labels from project-complexity practice guides and complexity-science handbooks are adjacent language only. They help name that projects can be hard for behavioural, systemic and meaning-related reasons. They are not synonymous with Oxford-style scenario services, and this essay does not treat questionnaires or complex-adaptive-system models as the method body.

3. Name the Layer: Ambiguity Versus Uncertainty

Leaders need a triage before they buy another analytic product.

Ambiguity here means contested or unclear meaning: parties do not share a stable account of what is going on, which forces matter, or which story of the future the investment should be built for. The problem is not only missing data; it is rival interpretations of the same signals.

Uncertainty and risk, in the sense used for estimation and contingency, assume a usable frame. Inside that frame, numbers may be incomplete, distributions may be wide, and events may be hard to predict — yet the team largely agrees what variables matter and what “success” would look like. Classical risk processes — identify, analyse, respond — fit that layer.

Complexity appears in project vocabularies as a third adjacent label: situations that are hard to manage because of human behaviour, system behaviour and unclear expectations. Useful as a warning light; not a licence to collapse every hard problem into one method.

Wrong-tool failures are common. More Monte Carlo will not resolve a meaning contest among sponsors and partners. More scenario workshops will not repair a shared-frame estimating problem that needs outside-view discipline. The first leadership move under pressure is therefore diagnostic: is the room arguing about meaning, or about quantities inside an agreed story?

Figure 1: Tool-class triage — ambiguity (meaning and frame), uncertainty or risk (numbers in a shared frame), and complexity as an adjacent label. Wrong-tool arrows mark common misfires.

A practical test in a steering meeting is to listen for the kind of disagreement. If parties argue about whether a demand path should be P50 or P80 inside a shared story of who the customer is, the fight is quantitative. If they argue about whether the asset is a merchant plant, a regulated utility, a strategic reserve, or a stranded prototype — and each camp can cite the same market report — the fight is about meaning. Mixing the two produces a familiar pathology: ever finer spreadsheets that never settle the room, because the spreadsheet assumes the frame the room does not share.

When each fresh study is read as confirmation by every camp, leaders should stop buying studies and start structuring a conversation about alternate contexts.

4. Turbulent Fields and Conditions That Weaken Forecast Habits

Organisation research long distinguished environments by their causal texture — how parts of the wider field relate to each other, not only how an organisation bargains with its immediate counterparts. In the most demanding texture in that classic series, dynamics arise from the field itself: the ground is in motion, ground rules of interaction come into question, and adaptive capacity can be overwhelmed. Later scenario-planning work extends that lineage with related conditions often summarised under a short label in the Oxford literature: turbulence, deep (Knightian) uncertainty, novelty and ambiguity. The label is a diagnostic prompt, not a scorecard. There is no objective index that certifies a project as “in TUNA”; the judgment is reasoned from signals in the field.

Under those conditions, conventional predict-and-control planning weakens. Past evidence about demand, competition, regulation or technology transfer poorly to the next decade. Best-case and worst-case packs that still assume one latent future with optimistic and pessimistic bands are often disguised forecasts, not genuine alternative contexts. Probability language can become irrelevant or misleading when the frame itself is unstable. Plausibility — whether a future context hangs together as a coherent story the organisation can rehearse — becomes a more honest standard than false precision.

For large capital projects the practical signal is familiar: the investment narrative that cleared the last gate no longer organises the room; parties keep renegotiating “what we are really building for”; contextual shocks arrive faster than the transactional interfaces can absorb. That is when clinging harder to a single forecast usually deepens the problem.

Turbulence is not the same as a busy project. A schedule can be hectic inside a stable field, or orderly while the field that will price, permit or operate the asset reconstitutes underneath it. Novelty matters when prior capital analogues mislead: a first-of-a-kind configuration, a new offtake regime, or a technology that changes the unit of competition. Deep uncertainty matters when probabilities cannot be responsibly assigned because the set of relevant futures is itself unsettled. Ambiguity matters when competent people, reading the same brief, still cannot agree what success would mean.

These conditions license readiness to decide under more than one plausible world, not paralysis. Value is judged by purpose fit and frame quality — not by a hit-rate against a future nobody could have known.

5. Contextual Versus Transactional Environment

Not every sticky issue belongs in a scenario workshop. A useful split, drawn from the same strategic-scenario lineage, separates two layers of environment.

The transactional layer is the immediate field of counterparties and interfaces the organisation can influence by interacting — contractors, regulators in a known process, suppliers, lenders, operators, communities with defined forums. Problems here often need clearer RACI, better interface management, renegotiated commercial terms or execution moves.

The contextual layer is the broader field beyond direct bargain: macro regulation, technology regimes, geopolitical alignments, long-cycle market structures, climate and energy transitions, demographic and legitimacy shifts. Critical uncertainties that can reshape the transactional circle usually sit here. Scenario work redirects attention to future images of that contextual layer so the present transactional situation can be reframed — sometimes called upframing — rather than endlessly optimising inside yesterday’s story.

Sorting prevents “scenario theatre”: elaborate futures packs about a delayed permit that actually needed a RACI fix, or Monte Carlo on a demand story while the meaning of the asset class itself is contested. A short working rule: if the dispute is primarily “who does what with whom under known rules,” stay transactional; if the dispute is “which rules and which world are we designing for,” escalate to contextual scenario conversation.

On a large capital project the split can be walked issue by issue. Interface lag between engineering and procurement is usually transactional. A sudden rewrite of interconnection rules that changes the economic unit of the plant is contextual. A contractor claim about productivity rates is often transactional quantity once the scope meaning is shared; a dispute about whether the owner ever intended a phased modular programme versus a single big-bang completion can be an ambiguity about purpose and context. Sponsor politics inside a known governance charter stays close to transactional process; a shift in the legitimacy of the asset class in the host society is contextual. Mis-sorting is expensive either way: scenario workshops that should have been RACI workshops waste calendar and credibility; RACI workshops that should have been contextual conversations produce crisp charts for the wrong world.

Figure 2: Contextual versus transactional split — example capital-project issues sorted so scenario work is not applied to RACI problems, and interface fixes are not asked to resolve field-level meaning contests.

6. Scenario Planning as a Service to Decision Readiness

At this point the essay turns from diagnosis to method stance. The ideas that follow are taken from the Oxford Scenario Planning Approach as set out in Strategic Reframing and related practice notes; they are paraphrased for capital-project leadership and are not a reproduction of that method’s branded wording, section architecture or glossary.

Scenario planning in this tradition is strategic learning under conditions where prediction is a weak guide. Its aim is not to declare which future will occur. Its aim is to improve the quality of the pre-decision frame and the options considered inside it. Scenarios function as a service to that learning — part of a wider intervention — rather than as a deliverable pack that can be filed after the workshop.

Several practical premises follow for owners and PMC teams.

Name the learner and the purpose first. Before producing decks, name who must learn — the individual or coherent group with power to start or stop the work — and what they need to be able to do afterward: approve, reshape, delay, partner differently, or reopen a gate question. Insufficient attention to learner and purpose is a major cause of disappointment with scenario exercises. Time spent clarifying intent and usability often matters more than time spent polishing slides.

Treat the work as social process as well as analysis. Scenarios create structured space for disagreement. Rival framings can be aired without forcing premature consensus. That “safe enough” conversational space is a design feature of the intervention, not a soft substitute for judgment. It should not be confused with a full theory of psychological safety; the point here is narrower: without room to contest meaning, the exercise collapses into theatre around the sponsor’s preferred story.

Prefer a small set of plausible scenarios over an encyclopaedia of futures. Two to four alternative contextual frames are usually enough to challenge perception. Orphan packs that never touch live options fail the usability test that practice notes and public teasers of classic scenario writing both warn about: managers will not know what to do with beautifully written futures that never connect to commitments.

Judge success by decision readiness, not hit-rate. Under deep uncertainty, contribution is assessed against explicit purposes and the quality of the pre-decision frame — not by whether the team “predicted” the future that later arrived.

Keep competitive and collaborative moves both in view. Turbulent fields often punish a purely adversarial stance and a purely harmonious one alike. Scenario conversation can surface where partners, regulators and communities may need to be treated as co-shapers of the field rather than as fixed counterparties — without romanticising collaboration where interests truly conflict.

Do not confuse method fidelity with slide volume. There is no single mandatory toolset that fits every learner and purpose. Facilitators and internal strategy cells should navigate choices — interview depth, workshop design, how many scenarios, how immersive the downframe — against the named purpose. A thin intervention that changes how a steering committee sees its options can outperform a lavish pack that never leaves the strategy shelf.

For owner organisations this stance has an uncomfortable implication: scenario work that cannot name its learner often defaults to “inform the board,” which is too vague to guide design. Prefer concrete purposes: reopen the offtake assumption before FID; test whether the alliance commercial model survives three regulatory worlds; decide whether to keep a modular option alive through the next gate. Purposes that sound like “build awareness” rarely create decision readiness.

Figure 3: Scenario planning as a service loop — learner and purpose, contextual critical uncertainties, a small scenario set, and decision readiness. The loop serves learning, not forecast crowning.

7. Small Scenario Set and Critical Uncertainties

Once learner and purpose are named, the working material is a handful of critical uncertainties in the contextual environment — factors that are both highly uncertain and highly consequential for the transactional options under review. Those uncertainties are combined into a small set of plausible scenarios: each scenario a coherent alternate context that could transform how interfaces, contracts, technologies and stakeholder roles behave.

Practice notes from published Oxford-linked work emphasise preparation and iteration: spend serious time aligning on who learns and why; expect more than one pass; do not confuse a best/worst envelope with distinct contexts. Secondary cases in that literature illustrate corporate and institutional settings; they are not construction outcome proof and are not treated here as field evidence for megaprojects.

What to reject is as important as what to build. Reject futures packs that never interrogate the live option set. Reject a single “official future” with decorative upside and downside. Reject the habit of producing scenarios after the investment narrative is already locked — at that point the pack is usually rationalisation. The test question after each scenario is built: which present options look robust, fragile, or newly inventable under this context?

How many scenarios? Practice in this lineage commonly lands between two and four. Fewer than two collapses into a residual official future. More than four tends to exceed what a steering body can rehearse with care, and invites the “impenetrable pack” failure mode noted in public teasers of classic scenario writing: managers admire the prose and then cannot act. Quality of contrast matters more than quantity. Scenarios should differ in the contextual logic that drives them, not merely in the intensity of a single variable.

Critical uncertainties should be chosen for consequence to the live decision, not for intellectual completeness. A long list of interesting macro trends is not a scenario set. Two or three uncertainties that would force different capital, partnering or sequencing choices are enough to start. Iterate: first-round scenarios often reveal that the real uncertainty was mis-specified, which is a successful learning outcome, not a process failure.

8. Reframing, Reperception and Wind-Tunneling Options

Impact does not come from owning a scenario document. Impact comes from completing a learning cycle: using contrasting future frames to reframe how the present is seen, then reperceiving options, roles and commitments in that new light. One pass is often insufficient; several iterations may be needed before the room experiences a genuine shift in what seems obvious.

In practice the cycle moves in two directions. Upframing lifts attention into higher-order contextual stories. Downframing returns to immersive “what if” rehearsal of near commitments — contracting strategy, capacity sequencing, partnering posture, regulatory engagement, operating model — under each scenario. Roles can flip across contexts: a counterparty that looks like a supplier in one world may look like a partner or a blocker in another. Wind-tunneling asks, option by option: how does this commitment fare under each plausible context? Which assumptions must be true for it to remain wise? What early indicators would show the wrong world is arriving?

Classic scenario writing has long warned that futures which never touch decision makers’ working mental models leave little trace. Public teasers of that lineage stress usability; fuller metaphors remain outside what this essay claims from paywalled sources. The operational standard here is simpler: if after the workshops the gate pack is unchanged in its frame and options, the cycle did not complete.

The intended outcome is not a predicted future. It is a better pre-decision frame and a set of options that have been pressure-tested for robustness, flexibility or staged commitment under ambiguity.

Wind-tunneling is deliberately modest. It does not claim to optimise a portfolio across probability-weighted worlds when probabilities cannot be known. It asks whether a commitment is a bet on one world, a hedge across several, or a staged move that preserves learning. Many capital commitments are irreversible long before the contextual fog lifts; the leadership task is to see that irreversibility clearly, not to pretend a single NPV path removed it. Options that look elegant in the official future and catastrophic in a plausible neighbour deserve either redesign, staging, or an explicit risk acceptance by the true learner — not silent hope.

Teams should also watch for false closure. A workshop that ends with “we aligned on Scenario B as most likely” has often slipped back into forecasting. The discipline is to keep the set alive long enough to change how options are seen, then to decide with eyes open about which assumptions the organisation is willing to bet on — and which early indicators would force a revisit.

Figure 4: Reframing–reperception cycle with wind-tunneling — alternate futures reframe the present; options are rehearsed under each context; iterate until decision readiness improves.

9. Owner and PMC Application on Large Capital Projects

What follows is transfer: application of strategic-scenario ideas to large capital projects. It is not IPA-measured outcome evidence, not a randomised evaluation of scenario planning on projects, and not a rename of FEL gates or reference-class forecasting as “scenario method.”

Front-end shaping under regulatory–market turbulence. When FEL calendars and investment narratives are being locked, insert a scenario conversation before a single official future hardens. Use it when contextual drivers — policy, offtake regimes, technology pathways, legitimacy — could invalidate the transactional plan. Do not relabel Merrow-style FEL stage discipline as Oxford scenario practice; stack them. Scenario work challenges the frame; FEL calendars still govern readiness to spend.

Joint ventures and alliances with contested framings. When partners disagree on what the asset is for, or on which world the venture is built for, a small scenario set can surface incompatible assumptions without forcing premature compromise. Governability work still designs the coalition; scenario work clarifies the meaning contest the coalition must metabolise.

Commissioning and operations rehearsals. Alternate contexts change staffing, offtake, maintenance logistics and regulatory oversight. Rehearsing commissioning under more than one plausible operating world is transfer of the wind-tunnel idea into late-project readiness — again labelled as such, not as field trial proof.

Claims and change. Separate ambiguous meaning (“what did we agree this scope was for?”) from uncertain quantities (“how many units, at what rate?”). Scenario language helps when the dispute is really about context and purpose; quantum and delay analysis belong when the frame is shared and the fight is numerical.

Stack, do not merge. After frames stabilize, hand estimation to outside-view and reference-class hygiene (Smart Teams lane). Keep Decide the Question First for ownership of the live decision question. Keep FEL/shaping calendars for capital realism. Contested-context scenario work sits upstream of those tools when meaning and field turbulence are the binding constraint.

Transfer also implies limits. Scenario services will not repair an owner team that lacks the authority or attention to act on what it learns. They will not substitute for technical due diligence, geotech truth-telling, or commercial competence. They will not make a weakly governable coalition suddenly cohesive. Used well, they reduce the chance that a beautifully estimated project is optimised for a world that was never the only plausible one. Used poorly — as theatre after FID, or as a branding exercise that never names a learner — they consume goodwill and teach the organisation that “strategy workshops” are ornamental.

PMC and owner strategy cells can operationalise transfer with light artefacts: a one-page learner/purpose note before any futures work; an issue-sort table (transactional vs contextual) attached to the risk register rather than buried in a side deck; a wind-tunnel matrix that lists live options against the small scenario set with red/amber/green robustness notes and explicit assumptions; a gate question that asks not only “is the estimate ready?” but “under which contexts is this option still wise?” None of these artefacts is a proprietary Oxford deliverable; they are ordinary project hygiene informed by the idea that frames and futures come before crowning a forecast.

Figure 5: Stack map — contested-context scenarios, owned decision question, estimate hygiene, and capital realism. Map the layers; do not merge the methods.

10. Close — Decision Readiness Under Ambiguity

A short protocol for owner and PMC leaders:

1. Signal. Ask whether turbulence, deep uncertainty, novelty or ambiguity in the field is overwhelming a single business-case story.
2. Learner and purpose. Name who must learn and what decision readiness means before any workshop.
3. Sort. Separate transactional interface problems from contextual critical uncertainties.
4. Small set. Build two to four plausible contextual scenarios — not a disguised forecast envelope.
5. Cycle. Complete at least one reframing–reperception loop; iterate if the gate narrative is unchanged.
6. Wind-tunnel. For each live option, ask how it fares under each scenario and what would falsify its assumptions.
7. Then estimate. If numbers dominate inside a now-shared frame, apply Smart Teams / outside-view hygiene. If the decision question itself is still unnamed, return to Decide the Question First. If surprise will still arrive after commitment, invest in governability of the coalition.

Under ambiguity, leadership is less about crowning a forecast and more about keeping the organisation able to learn before capital hardens. Scenario planning, used as a service to that learning and credited to the Oxford tradition that shaped these ideas, is one disciplined way to do so. Companion ProTrain pieces develop framing ownership, estimate hygiene and coalition governability in their own right. Forthcoming pieces in the Decision Making and Leading People lanes — including further work on leading in complexity and front-end loading — will extend adjacent practice without collapsing those methods into this one.

The invitation is clear: refuse the comfort of a single story when the field no longer supports one, and invest scarce calendar in learning that changes options — not in packs that decorate a decision already made.

References

1. Emery, F. E., & Trist, E. L. (1965). The causal texture of organizational environments. Human Relations, 18, 21–32. Public archive PDF: http://www.moderntimesworkplace.com/archives/ericsess/sessvol3/GEMTRCAUp53.pdf
2. Project Management Institute. (2014). Navigating Complexity: A Practice Guide. Project Management Institute. ISBN 978-1-62825-036-7. (Adjacent complexity vocabulary only.)
3. Ramírez, R., Churchhouse, S., Palermo, A., & Hoffmann, J. (2017). Using scenario planning to reshape strategy. MIT Sloan Management Review. Open accepted manuscript (Oxford ORA): https://ora.ox.ac.uk/objects/uuid:916e9b13-163c-4a2a-989d-2804e907f4a8
4. Ramírez, R., & Wilkinson, A. (2016). Strategic Reframing: The Oxford Scenario Planning Approach. Oxford University Press. ISBN 978-0-19-874569-3. (Primary idea source for the scenario-planning stance in this essay; wording and branded method architecture are not reproduced.)
5. Wack, P. (1985, September). Scenarios: Uncharted waters ahead. Harvard Business Review. (Public teaser only; full article paywalled — no claims beyond publicly readable material.)
6. Wack, P. (1985, November). Scenarios: Shooting the rapids. Harvard Business Review. (Public teaser only; full article paywalled — usability warning drawn from publicly readable opening only.)