Markets are moving faster, and the numbers back it up. On 5 August 2024 the VIX posted its largest one-day jump on record, spiking roughly 180% intraday to nearly 66. In April 2025 it took only five trading days for the index to peak near 60, a move that in earlier cycles would have taken weeks, and on 4 April the options market printed a single-day volume record above 100 million contracts. The structure of trading has compressed too: zero-days-to-expiry options accounted for about 59% of SPX volume in 2025, roughly double their share in 2022. A growing portion of risk now appears and disappears inside a single session.
Most risk infrastructure was not designed for this. If risk lands the next morning, if repricing a book means asking a quant to rerun a grid, if the dashboards can only answer the questions someone thought to previously build reports for, then the number is stale before anyone acts on it. The firms we see pulling ahead have stopped competing primarily on pricing model sophistication and started competing on something more mundane: how quickly the people making decisions can price, aggregate, slice, and stress the live book themselves.
Three seats, same problem
A buy-side PM wants exposure and P&L attribution when a position moves, not at end of day, because that is when the sizing and rebalancing decisions actually happen. A sell-side trader is working an even tighter loop, where pre-trade pricing, hedging, and intraday limit checks all depend on numbers that reflect the current market and the current book. And quants on both sides are being asked to produce richer analytics, non-linear payoffs, VaR, expected shortfall, multi-dimensional XVA, while too much of their week disappears into data ETL and reconciliation rather than modeling. Different seats, but the underlying requirement is the same: analytics that keep pace with the decision.
Built for on-demand, not overnight
The reason traditional batch architectures run overnight is that they reload and recompute everything whenever the data changes. Furthermore, a single logical computation is often the result of several systems running their own batches sequentially, all catering to the slowest common denominator. An engine designed for real-time analytics works differently. It updates incrementally, keeps track of dependencies between computations, recomputing only the parts of a calculation actually touched by a change in trade, market, or reference data. Positions and risk factors stream in continuously, and even complex measures, from a single sensitivity up to VaR or PFE, stay current against the live book. Pricing and risk stop being scheduled events. They become queries, answered in seconds.
Aggregation is where the value leaks
Firms invest heavily in their risk and pricing libraries, then treat what happens to the output as an afterthought. Sensitivities, scenarios, and P&L vectors get generated, and the work of aggregating and presenting them falls to spreadsheets, overnight reports, or whatever the reporting stack happens to support. In our experience this is where much of the investment quietly loses its value. Analytics that arrive too late to inform a decision might as well not exist. Analytics that can only be summarized by a small pre-defined set of aggregation axes invariably prove to be insufficient in times of crisis, where fast ad-hoc exploratory work is required.
A flexible aggregation engine takes everything the pricing library produces (risk, scenarios, sensitivities, P&L decomposition, VaR and expected shortfall vectors, capital charges) and lets users interrogate it along whatever dimension the question requires. Slice VaR by desk, then re-pivot by risk factor. Drill a firm-wide sensitivity down to the handful of trades driving it. Decompose P&L by strategy, then book, then underlying. And because heterogeneous sources are consolidated into one consistent view, the front office works from a single version of the book instead of reconciling exports from three systems that almost agree.
Put that behind self-service dashboards and the request queue shrinks. PMs, traders, and risk managers build their own views and follow their own line of questioning, and quants get to stop moonlighting as glorified report generation bureau. Put that into a platform that cleanly separates aggregation logic (should this be a sum weighted by notional or shares?) and presentation (should this be a table or a chart?) and you have a winning proposition.
What-if analysis
The most valuable questions on a desk are conditional ones. What if I add this hedge? Scale this position? Shock this factor? Get hit on these quotes? A native what-if capability branches from the live book, applies the changes incrementally, and makes every measure available on the scenario branch side by side with the master. No duplicated dataset, no batch job, no waiting until tomorrow to find out.
Where AI fits
The more interesting development is agentic AI: agents that don’t wait to be asked. An agent watching the book can monitor exposures against limits continuously, notice when a hedge has drifted, compute the cheapest hedge and execute it. Another can track running orders against the market, flagging fills that deviate from expected cost or venues behaving oddly intraday. A third can watch the market itself, connecting a move in vol or a sector sell-off to news events and to the positions actually exposed to it, with the attribution already worked out down to the trade. The pattern is the same in each case: the agent runs the loop (query the live book, compare against a threshold or an expectation, explain the difference) on its own, and escalates to a human when something warrants judgment. None of this replaces the trader or the quant. It changes what they spend attention on: less scanning for problems, more deciding what to do about the ones that matter.
The loop is the edge
Sophisticated models are table stakes now. What separates firms is how fast they can run the full loop (price, aggregate, visualize, simulate, explain) and whether the people making decisions can run it themselves, on live data. Volatility that arrives in days rather than weeks punishes anyone waiting on an overnight batch. The firms that have shortened the loop are simply reacting to what is actually happening, while everyone else reacts to yesterday.
Follow us on LinkedIn for the latest news and updates.
