Q is the AI built into every TraiQ install. It pairs a large language model with TraiQ's own calculation engines — so it reasons like a treasury analyst, while every figure it quotes comes from the same engine that drives your dashboards. Bank book or corporate group, Q knows the vocabulary, the methodology and your limits.
Plain English, your own shorthand, follow-ups and all. Q works out what you mean, runs the numbers through TraiQ's engines, and answers with the evidence — the deals behind the figure, the assumptions behind the scenario.
Language models are brilliant at understanding a question and explaining an answer — and unreliable at arithmetic on a balance sheet. So Q never does the maths in its head. It interprets, TraiQ's deterministic engines compute, and Q explains.
In plain English and your own shorthand — "govvies", "the semis book", "WACoD". Q maps treasury jargon — and your own house glossary — to the right part of the book.
If the question could mean two materially different things, Q asks first — in one round, with the readings laid out — rather than guessing.
Q calls TraiQ's own tools — the same calculations behind every dashboard tile — so its numbers tie to the screen by construction.
An answer with the deals behind it, the assumptions stated, and every figure marked by where it came from.
Q has read access to the whole book and the platform's full analytical toolkit — on a bank book and a corporate group alike.
Positions, exposures, maturities and counterparties — filtered, grouped and totalled, with the deal IDs to back every figure.
"Why is my LCR 221%?" "Why did net leverage jump?" — the drivers behind any ratio, and what moved it since any earlier date.
Every limit and covenant worst-first — red, amber, green — with value, trigger and headroom. "Am I breaching anything?" answered in one line.
Rate and FX shocks, balance-sheet rotations, a new deal tested against your ratios, limits and covenants — run on a copy, never the book itself.
IRRBB, liquidity, FX and reverse stress for banks; covenant, cash trough, funding cost and hedge ratio for corporates.
When no standard tool fits, Q composes the platform's own engines into a new calculation — run in an isolated sandbox, with the working shown.
One name across every instrument — credit exposure, funding provided, limit headroom and expected loss, in a single view.
Hand Q figures from your TMS, ledger or a draft return and get back only the breaks beyond tolerance, traced to the deals.
Around the conversation sits a full workspace — where Q's analysis becomes something your team can keep, re-run and put in front of a committee. Q proposes; your team reviews and saves.
Upload risk appetite statements, treasury policies and reference documents. Q reads them alongside the live book, so it can set the numbers against what your policies actually say.
A funding review, a hedge proposal, a stress study — its files, web sources and outputs kept together, so Q picks up where you left off.
Design a new simulation with Q on a structured specification. Q only writes the calculation after your team signs the specification off.
Describe the report you need; Q assembles it from the platform's own charts and tables. Nothing Q invents makes it in — every section must resolve to real data.
ALCO and board decks laid out by Q from a brief, exported as native PowerPoint with editable charts. Verdicts and recommendations stay human-written.
Save the questions you ask every morning or every month — the daily read, the market watch, the pre-ALCO check — and have them run on schedule, ready when you are.
Briefs, reports, decks and models — each one built with Q, checked by your people, and ready to use again tomorrow.
Standing briefs run on schedule and arrive as the few lines that matter — limits, liquidity, the names to watch.
Saved report packs re-run on today's numbers and export as PDF, PowerPoint or Excel.
Q lays out the ALCO or board deck from a brief — native charts your team can still edit, speaker notes included.
Q drafts the specification with you, marks every line it assumed, and writes nothing until a person signs it off.
Q is designed to behave like a careful colleague, not a confident chatbot. The guardrails are part of the product, not a policy document.
Figures that came from TraiQ's engines are marked as engine-traced. Anything Q worked out itself is labelled separately — so a derived number can never pass for a calculated one.
Every what-if, projection and scenario ends with a section headed Assumptions. Nothing is buried in the prose.
Q will point out an unused limit or a costly funding line when you ask — framed as a computed observation, never as advice.
Q's access to your deals, limits and configuration is read-only. What-ifs run on a copy, and anything Q drafts lands unsaved until a person saves it.
Web research is off by default, and your deployment sets its scope. On TraiQ's own search, every query is screened for deal IDs, amounts and counterparty names — and approved by a person — before it leaves.
A simulation, report, deck or brief Q helped build stays marked unverified until a second senior user approves it.
Press cross-check and Q re-derives the answer independently, so you can see whether it holds before you rely on it.
Every Q turn is logged on your own server — model, tools used, cost, and a fingerprint of each result — without storing the conversation itself.
Q isn't tied to one AI vendor — seven providers are supported today. Choose the model and the hosting that fits your mandate, and change it later without changing the platform. Your key, your account, your terms: TraiQ never proxies, stores or sees your book.
Anthropic's Claude or OpenAI's models, under your own key and account — for maximum capability.
Claude and other models through AWS Bedrock in Sydney, or open models from an Australian provider in Australian data centres.
An open model on your own servers, fully air-gapped if you need it — no external calls at all.
Q Bridge connects the AI agents your organisation already runs — Claude, OpenAI's Codex and other agents built on the Model Context Protocol — to the same tools and engines Q uses. Reconcile the TMS overnight, draft the ALCO note in your own workspace, or pull a finished deck into the agent your team already has open.
Read-only, keyed per analyst, and every figure stamped with its source.
See Q work on the demo book — or bring us a question from your own treasury and we'll walk through how Q would answer it.