Morgan Stanley

Resist listing asset classes. Open by naming what you are actually allocating, which is not bonds but three separable risks: duration, credit spread and currency. Say which of the three you want to own, which you are not being paid enough to take, and why. Then build the book: a core of government duration sized to your view on the policy path, inflation linked bonds as the hedge against the one thing that genuinely breaks a bond portfolio, investment grade where the financing need from data centre and grid capex is enormous and the question is whether the spread compensates you for that supply, high yield and private credit only where the documentation earns it, and emerging markets differentiated country by country rather than bought as a block. Put geopolitics in the term premium and in the inflation tail rather than in a sentence about uncertainty: defence and fiscal spending mean more issuance and steeper curves, and energy supply is the live route to an inflation surprise. Put megatrends in the supply and demand for paper: ageing pension funds want long duration, the energy transition and AI infrastructure want to borrow. Finish with what would change your mind and why you would hold liquidity back, because in a geopolitical shock the dislocation is the opportunity.

Your first sentence decides the answer: say whether the surprise is demand led or a supply and cost shock, because they propagate differently. Demand lifts growth and inflation together and the central bank simply tightens. A supply shock, energy or tariff driven, raises prices while squeezing real incomes, so the bank faces a trade off and the market has to guess which half of the mandate wins. Then walk the chain instead of listing markets: prices into real incomes and consumption, inflation into expectations, expectations into wage setting, all of it into the reaction function, and the reaction function into the policy path. On rates, the front end moves first and hardest. A credible central bank gives you a bear flattener. Doubt about credibility, or fear of fiscal dominance, gives you a bear steepener instead, with breakevens wider and term premium rising, and that distinction is the whole answer. On FX, the textbook response is a stronger currency on wider rate differentials, but that only holds if real rates rise. If the market reads it as bad inflation, or the country imports its energy, the currency falls on the terms of trade instead. On regimes: in a low inflation regime bonds hedge equities, in a high inflation one the correlation flips positive and the diversification you were relying on disappears. Anchor the whole thing with two live numbers, the latest print and where the market prices the terminal rate.

Two questions bolted together, and the marks sit in the second. Keep the commodity view short and mechanical. For energy: spare capacity and supply discipline against demand, with the geopolitical risk premium and the shape of the forward curve as the tells. For precious metals: gold is a real rate story that official sector buying has repeatedly overridden, and silver carries an industrial leg through solar. Two sentences each is plenty. Spend the rest translating, because that is what is being tested. A conviction trade needs five things: the instrument, the direction, the thesis in one line, the horizon, and what would kill it. Make the three independent rather than three versions of the same bet. The natural routes are inflation, through breakevens or linkers if you think energy pushes prices up; the curve, through a steepener or flattener expressing what the central bank does about it; and relative value, long a commodity exporter's local debt or credit against an importer, or energy credit against the broad index. Say roughly how you would size them, and mention carry, because a trade that bleeds while you wait is a different trade. Markets interviewers are listening for whether you can turn a view into risk, which is the step most candidates never take.

It asks for a plan of action, so answer as a sequence rather than an essay on whether private credit is a bubble. First, triage your own book: exposure by borrower, sector and sponsor, but above all by structure, because that is where these go wrong. Where is the collateral, who verifies it exists, is it pledged anywhere else, what lien are you in, what do the covenants actually require, how much of your income is PIK rather than cash, and how are your marks set when nothing trades. Second, diagnose. Be able to say in one sentence what happened in each name they put to you, and which failure it was: collateral and verification breaking down, leverage stacked at both the asset and the fund level, documents loosened by years of competing for deals, valuations that never move, or a liquidity mismatch in a semi liquid vehicle facing redemptions. Third, change the underwriting: third party verification and borrowing base audits, representations against double pledging, information rights, independent valuation, less PIK tolerance, less single sponsor concentration. Fourth, and this is what separates the answer, say where you would now deploy. Stress creates entry points: discounts to net asset value in listed vehicles, secondaries, and rescue financing on terms and documents nobody could have demanded two years ago. A purely defensive answer is half an answer.

The failure mode is "I use it to summarise articles", which tells them nothing. Give one thing you actually did, and make it something that would have been impractical by hand: pulling and parsing a set of filings or transcripts, building a screen, automating a data pull you were doing manually every week, tagging earnings calls, or writing the code for a simple backtest. Then four beats: what the task was before, what you built, what it produced, and what changed that you can put a number on, time saved or coverage you could not otherwise have reached. Then the beat that matters most for a research seat: how you checked it. Say plainly that you never let it produce a figure you had not verified against the source, because a hallucinated number in a research note is not a small mistake. Show you know the shape of the tool, strong on structure, summarising and code, weak on precise figures and on anything after its training data, and if you have pointed it at real market data, say what broke. Morgan Stanley runs its own internal tooling, so the signal they want is that you would use this as an analyst rather than as a student. Expect a follow up, so never describe something you cannot walk through.