Four domains,
one question at a time.

The decisions our clients face do not respect disciplinary borders. A sovereign issuance window is a macro question that is also a political-coalition question that is also a flows question. A bank's recapitalisation is a company question that is also a sovereign question that is also a regulatory question. Delphic Strategies is built around that observation.

The four domains.

We work across four lenses, deliberately. Most boutiques pick one. The synthesis is the product.

How a mandate runs.

An engagement starts with the client's actual question — a position under stress, a counterparty under review, a thesis being challenged, a board paper that needs an honest second opinion. From there we map which of the four domains the question genuinely depends on, and scope the work accordingly. Some mandates are a single dossier. Others become standing coverage, updated against tracked catalysts.

Reading at scale.

The published surface of a cross-domain question is now larger than any single analyst can read. Primary documents, secondary commentary, market and trade data, filings, transcripts, shipping and satellite signal — the volume is the problem, and pretending otherwise is what produces house views built on a curated subset of the evidence.

Frontier AI systems are operationally integrated into the work for this reason. They do what scale rewards: continuous monitoring across the full source set, pattern reading across more variables than working memory will hold, first-pass synthesis at a breadth a small team could not cover. They do not do what scale punishes: framing the question, judging the synthesis, weighting what matters, owning the call.

The discipline is to be precise about which work is delegated to the instrument and which is not. Reports carry the principal's name because they carry the principal's judgement. The instrument extends the principal's reach; it does not replace the principal's responsibility for the conclusion.

Where we use scenarios.

For situations with a small number of mutually exclusive endings — a sanctions decision, a leadership transition, a kinetic inflection point — we use named scenarios with explicit probability weights and tracked catalyst calendars. The Iran–Hormuz framework is the most visible current example: five plain-English tracks (DEAL, STRIKE, SIEGE, SALVO, PROXY), updated against a published catalyst calendar, currently in its eightieth day.

The scenario framework is paired with a parallel global oil supply and demand analysis — capacity, inventories, flows and routes, OPEC+ posture, sanctions and SPR — that tests each scenario branch against the fundamentals the underlying market can actually absorb. The geopolitics tells us which futures are plausible; the fundamentals tell us what each one would cost. Neither document stands alone for this kind of question.

For situations where the relevant uncertainty is continuous rather than discrete — a yield path, a deposit-base trajectory, a margin structure — we use conditional sensitivity instead. The discipline is the same: state the assumptions explicitly, quantify what would need to change, and let the client own the call.

What we are not.

We are not a newsletter, a subscription product, or a generalist consultancy. We do not write what the consensus already believes. We do not produce house views designed to be agreeable across a wide readership. The work is for clients with a specific question, who would rather be helped to think than told what to think.

To discuss an engagement, see contact.