Enterprise · Investment banking
Archivist IB
Deal signals, pitch books and diligence for investment banks, with the source next to every number. Run inside your own walls.
Use cases
How we make the line go up
1) Morning list
Which company will need you this week.
- You set the watch list and the signals that matter
- Filings on KAP, SEC and SEDAR+ scanned every night
- Each morning: company, signal, reason and source
- Public data only, so it runs before any mandate
2) Pitch book in an hour
Type the name. The draft arrives in your template.
- Your template and deal history, connected once
- Sector, peers, precedent deals and filing history pulled
- A source beside every number
- The banker edits and finishes
3) IPO pricing report review
Upload the pricing report. Get a benchmarked draft.
- Assumptions and multiples checked against past IPOs
- Item by item: what departs from peers, what is weak
- Every point sourced; your analyst writes and signs
- Public data only
4) Deal confidentiality
Deal data never leaves the bank.
- Outside models see code names, never company names
- At close, one click wipes NDA data and certifies it
- Only the deal team sees a deal, even in search results
- Every access and deletion on a record no one can alter
5) A workflow for every desk
One archive, a use for each desk.
- ECM: foreign funds in the stock, prospectus drafts
- DCM: a refinancing calendar from issuers' maturities
- M&A: a buyer's past deals, the data room clause by clause
- Closing: every version of the SPA and schedules, one list
6) Post IPO subscription
Resell it to the companies you take public.
- A workspace in your brand for every company you list
- Their team drafts filings and investor answers, sourced
- A monthly fee instead of a one time mandate
- The company stays with you for the next deal
Technicals
Your CFO and your compliance team, both at yes
1) The right context
Models need the right context, not more of it.
- Retrieval picks relevant, current, sourced passages first
- Faster, sharper answers from less input
- 99%+ recall@100 against exact nearest neighbour search
- When it is unsure, it says so
2) Deployment
Client data stays inside your boundary.
- Your servers, your cloud or compliant endpoints
- Can run with no outside network access
- Local, private or subscription models, one interface
- One core, a package per desk
- GDPR, GLBA, PIPEDA and KVKK aligned
3) The vault
The best outside model, without your names leaving.
- Names and numbers become code names inside your walls
- The outside model works without knowing who is who
- Real names return on your screen, never outside
- Every outbound call logged; unused when local is enough
4) How it works
An open question becomes an answer you can audit.
- Ask by topic and goal, not keywords
- Scope by company, period, document type and market
- Structured, full text, semantic and graph search combined
- Every statement opens its passage on the same screen
5) Documents first
The hard part is the documents, not the model.
- Signatures, scans and OCR handled in production
- One PDF stitched from eight or more formats, read whole
- Format drift across exchanges absorbed
- A fraction of what cloud conversion vendors charge
6) Access control
What a person can see, their agents see. No more.
- Row level security, decided next to the document
- Permission can narrow through an agent, never widen
- SSO, SCIM and OIDC from your identity provider
- Every access, export and deletion signed and kept
