Ask Orca — How the Marketing App Works

Ask, don't build#

You type what you want, in plain English, and the finished answer comes back ready to show.

• "Largest country exposures" → a finished pie chart.
• "Rating profile" → a finished breakdown table.
• "How much cash?" → the number, in context.

You write the question. Orca returns the finished thing. There is no step in between where someone formats a chart or tidies a table — the answer arrives done, already in the fund's colours. The question is yours; the finished, branded answer is Orca's. That single split — you ask, the engine answers — is what keeps the app small and the output consistent.

A dashboard is just six questions#

A whole client dashboard is six plain-English questions. Edit them, press Run, and six finished panels appear.

• Six questions on the left, six finished components on the right.
• Each panel is a chart, a table, or a headline figure — in brand colours.
• Change a question, re-run, and only that panel updates. You never touch a chart.

What looks like a bespoke dashboard is really six saved questions pointed at the engine. That is why a new dashboard takes minutes rather than a design sprint, and why it never quietly drifts out of date.

Orca does the thinking#

Ask Orca decides what kind of answer a question needs and sends it back as a ready-to-draw block. The page simply draws whatever arrives.

• Orca returns a typed block — chart, table, or fact sheet — and chooses which one fits.
• The page knows how to draw each kind of block, once.
• A new kind of question renders with no new work on the page.

This is the quiet strength of the design. The marketing app does not know what a yield curve is, or which part of the engine produced a number — it just draws the block it is handed. All of the intelligence sits on Orca's side of the line, so the app stays small and steady while the engine behind it keeps getting cleverer.

What happens between question and answer#

Four steps run end to end, every time you ask — and not one of them involves building a chart or a table by hand.

1. The data is fetched. Orca pulls the live numbers behind your question — holdings, prices, ratings, exposures — from the engine.
2. The model is asked to answer. A language model is handed the question and that data and asked to produce the answer — and to return it as a structure, not as prose. So it doesn't write "the biggest exposures are…"; it returns "this is a pie chart, and here are the slices," or "this is a table, and here are the rows," or "this is a single headline figure."
3. The app reads the structure and picks the matching component. The page looks at what kind of structure came back and drops it into the chart, table, or fact-sheet component that already exists for that kind. It never decides what the answer is — only which prebuilt component shows it.
4. It renders, on-brand. The finished component appears in the fund's colours, ready to use.

The components — chart, table, fact sheet — are built once and reused for every answer. The engine decides what the answer is; the page only picks which component shows it. So a brand-new kind of question tomorrow simply renders, with nothing new to build — which is why the app stays simple while the answers keep getting richer.

It already looks like you#

Every chart and table arrives in the fund's brand — not a generic template bolted on afterwards.

• The brand travels with the answer, decided centrally.
• The same app serves a different team simply by changing who is asking.
• Colour, type and tone stay consistent across every answer.

Because the look is resolved in one place and sent with each response, there is a single source of truth for how things appear. Presenting the same app to another mandate is a setting, not a rebuild.

It gets better on its own#

When the engine gains a new capability, it appears in the app automatically — nobody updates the page.

• New data and new analytics surface with no change to the front end.
• The app shipped today keeps improving as Orca does.
• No maintenance backlog, no "we will need a developer for that."

This is the compounding benefit. Most dashboards rot — they are snapshots that need re-cutting every time something changes. This one is a live window: the page is fixed and simple, the engine moves. The value grows without the upkeep growing with it.

From a screen to a client-ready page#

The same answers become a clean, branded document — the page you are reading was built exactly this way.

• Every answer is laid out to print, not just to scroll.
• A dashboard becomes a PDF a client can be sent, on-brand.
• One source of content, finished in several formats.

The discipline that makes the screen work — finished blocks, nothing scrolling, the brand baked in — is precisely what a good PDF needs. So the marketing screen and the client document are the same content rendered two ways. This page is the proof.

One engine, more than one door#

The engine answering these questions — Orca — is the same one behind Athena, the live cockpit the desk runs the portfolio on: holdings, compliance, P&L, and a daily value screen across roughly 11,000 bonds.

• Ask Orca is the simplest door onto that engine; Athena is the full one.
• The plain-English answer you get here is the same answer the desk works from.
• You don't need the full cockpit to get a real answer — you just ask Orca the quick way.

So this isn't a marketing toy bolted on the side. It's a light, direct window onto the same engine that prices and watches the whole fund.

There's a flip side worth naming. The automation marketing now relies on — the finished charts, the six-question dashboards, the client-ready PDFs — all comes out of this same engine. Take it away and that automation goes with it: back to building each chart by hand and waiting on an analyst. Athena isn't only the desk's cockpit; it's the backbone of marketing's automated output too.

Try it yourself#

One link, one question — the whole idea in ten seconds.

• Open ask.guinness.x-trillion.com and ask anything about the fund.
• Type a question, or press Run for the six-question dashboard.
• Every answer comes back finished and on-brand — built by you, in seconds.

Open Ask Orca→