AI that understands what moved a market, not just what printed.
A price is the easy part. What a desk needs is the mechanism behind the move, whether it touches the contract they actually hold, and what has failed on this setup before. Recapi builds that understanding, from raw source to reviewed trade.
WHERE RECAPI CAN HELP
Six areas of work where Recapi has already closed the gap.
Each is a desk discipline with its own evidence standard. These are the ones where we built the understanding and watched the work change.
01 · MARKET INTELLIGENCE
Recapi read the move and named the mechanism behind it
Filings, disclosures, exchange data and index prints were read as one body of evidence rather than four feeds, so a move came back with a cause, a source for every claim and a clear statement of whether it touched the contract the desk actually holds.
02 · RESEARCH & EVIDENCE
Every claim carried the document it came from
Primary filings did the work. Commentary informed, but never carried a claim on its own.
03 · SIGNAL CONSTRUCTION
A signal fired only when its conditions held
Assumptions written down, and a condition that has failed before kept as a condition.
04 · RISK & EXPOSURE
Sizing accounted for what really shares a driver
Three positions on one input are one bet, and the size was cut accordingly.
05 · TIMING & PASS-THROUGH
The lag was modelled instead of assumed away
Contract resets and delivery terms set when a cost actually lands.
06 · POST-TRADE REVIEW
What the thesis got wrong became a condition on the next one
A closed position was reviewed against the argument that opened it, not only against the P&L. Where the thesis was right, the reasoning was reinforced; where it was wrong, the miss was written back into the signal so the same setup cannot fire the same way twice.
HOW WE WORK WITH YOUR BUSINESS
We start with your workflow, not with our product.
We sit with the analysts, traders and risk owners who hold the standards nobody writes down: what counts as confirmation, which contract matters, and which setups have burned this desk before. Only then do we decide what intelligence is worth building.
Nothing gets built until we can explain how your work actually happens.
01
Understand how it works
Sit with the desk as it actually decides, not as the process document says.
02
Find where intelligence is missing
Name the calls that rest on judgement no system holds.
03
Build the understanding
Turn filings, data and desk standards into structure AI can reason over.
04
Deploy into the workflow
Into the blotter and the morning read, not a portal beside them.
05
Keep improving it
Every review writes a condition back into the signal.
WHAT AI CAN ACTUALLY TAKE ON
Four things that only work once it understands the desk.
Read the filing and say what actually changed
A move explained by mechanism, with the primary source behind each line and a clear note on what stays unresolved.
Build a signal that states its own assumptions
Conditions, evidence and the setups that have failed before, all held together so a signal cannot quietly fire on half of them.
Size it against exposure you already carry
Positions that move on one driver are treated as one bet, so the suggested size reflects real concentration rather than a count of tickers.
Explain the result against the argument that opened it
Attribution that separates the part of the thesis that held from the part that got lucky, and writes both back into the signal.
WHAT WE'VE ALREADY BUILT
This is already running on live research.
We went deep into commodity and industrials research and built intelligence that reads primary sources, names the mechanism behind a move, states what is still unresolved, and argues its confidence instead of asserting it.
It is not a demo and not a template. It is what happens when Recapi learns a domain properly.
Mechanism-first
A cause with sources, not a headline with a number.
Confidence argued
What raises it and what lowers it are both shown.
PRODUCT GLIMPSES
Three places the understanding shows up.
Screens from the system in production. Data shown is illustrative.
ONE WORKFLOW IS JUST THE BEGINNING
Commodity research is one domain we went deep into. It is not the boundary.
The same approach applies wherever a decision rests on evidence, exposure and what has failed before. These are the next places it fits.
COMMODITIES
Already built
Source to signal to review, running on live research today.
CREDIT
Covenants read in context
Obligations tracked against the filings that create them.
EQUITY RESEARCH
Theses that keep their evidence
A note that can still be audited six months later.
MACRO
Mechanisms, not narratives
Policy changes traced to the contracts they actually touch.
TREASURY
Hedges tied to real exposure
Cover sized on what moves together, not on line items.
COMPLIANCE
Decisions that explain themselves
Every position carries the reasoning that opened it.
What we built here is one example of what becomes possible when Recapi enters a complex domain and learns how the work actually happens.