- L1 · leak31 hr/wk
Briefs are written by hand.
Median 2.6 hours per target brief; three analysts spend a combined 31 hours a week on it. 62% of the content is public information already sitting in their sourcing database.
time-tracked shadow week plus document timestamps, trailing 60 days
- L2 · leak8.4 hr/wk
Outreach moves at hand speed.
46 personalized first touches a week firm-wide, eleven minutes each, 3.2% reply. Partners write the ones that matter; analysts reuse old language for the rest, and it reads that way.
email log export, eight weeks, deduplicated
- L3 · leak27 late
The CRM is rotting.
39% of last quarter's external meetings never reached it. 27 follow-ups went out more than two weeks late. The firm's relationship history lives in two partners' heads.
calendar-to-CRM cross-check, one quarter
- C1 · ceiling1 mandate declined
Coverage caps the practice.
397 of 1,870 qualified targets contacted in twelve months. In May the firm declined a buy-side mandate for analyst bandwidth. By its own fee schedule, one more closed mandate is a six-figure success fee.
engagement log and fee schedule; probability of a close not modelled
- W · waste3 seats
Found money.
Three of five seats on a data subscription untouched since March. Two research databases overlap on the firm's actual use; flagged for the January renewal, not for today.
subscription inventory, verified in billing and login logs
One machine, seven lines: Claude and Claude Code on the firm’s own context, a voice vendor on the phone line, and the CRM and sourcing database they already pay for, finally wired in. One line is custom, and the market check is written on it. Cross off any line; only what they approve gets built or billed.
Setup
The firm's brain, written down: mandate criteria, sector theses, the house style for briefs and teasers, each partner's voice. Plus accounts, access, the shared channel, and the wiring into CRM, sourcing, and Google Workspace. Every other line runs on this. Done first, paid first.
Research desk
Claude Code on every analyst desk, working against that context: briefs, screens, and teaser drafts in the house style. Claude for the partners: memo drafts, committee prep, buyer lists. Nothing leaves without a human name on it. Answers L1.
Sourcing pipeline
The custom line, and the market check is written on it: nothing sold anywhere reads this firm's mandate, researches each target overnight, and writes the brief their way. So we build it: agents on a standard harness, against their own subscription and context, delivering ranked briefs to the shared channel every morning. Answers L1 and C1.
Outreach engine
First touches drafted from each target's own brief, in the signing partner's voice, queued in their mail client for approval. Nothing sends unapproved. Answers L2.
Callback line
A voice agent on the firm's number. Owners call back on evenings and weekends; it answers, qualifies, books the partner call, and logs it. Known contacts ring straight through.
Meeting to CRM
A recorder on every external call, feeding the CRM through its native integration; the CRM's own auto-capture switched on and tuned. The firm already pays for half of this; nobody ever configured it. Same-day follow-up drafts for approval. Answers L3.
Training
Working sessions for all eight. Partners on the research desk, analysts on the pipeline and Claude Code, the office manager on the whole board. Playbooks that survive turnover; a new-analyst runbook. Builds fail at the human layer; that is why this is its own line.
One build fee
All seven lines, approved before work begins. Setup at signing. Every other line billed only when it's live and working.
The keep, monthly
Eight people supported, by name. Month to month after ninety days. A hire adds a seat, a departure removes one.
Vendors and usage, at cost
Their accounts, their cards, vendor prices as printed. Model usage on a meter in their name. We add nothing.
One all-in number
Our fee, their subscriptions, expected usage, minus the seats they cancel. Stated before signing, reprinted monthly next to the measurements.
Valuation and structuring models.
Not building. Drafts and comps pulls, yes, through the research desk. The model itself is partner judgment with liability attached; a machine holding it would be a defect, not a feature.
“Have it pick which deals to pursue.”
Declined. Conviction is the product this firm sells. The pipeline ranks and briefs; it never decides. The day a machine picks your deals, you are the wrapper.
Full sale documents, end to end.
Not yet. The context needs two more closed deals' worth of house style before a draft is worth a partner's red pen. Revisit at the quarterly.
A website chatbot.
Declined. Deal flow arrives by phone and email: 214 inbound calls and emails against 4 site inquiries in the same eight weeks. The callback line stands at the door people actually use.
Ownership
Accounts, data, context, configurations, the pipeline's code, documentation: theirs. No software rent, no markup, ever. A handover document in their Drive from day one.
One page of accountability
Every system we're responsible for, named: what it does, what “working” means, and what we do when it breaks. The monthly report covers exactly that list.
Confidentiality
NDA before the look. Mandates, targets, and relationships never train anything, never leave their accounts, and never appear anywhere without written permission.
| Number | Source | Method |
|---|---|---|
| 31 hr/wk · 2.6 hr per brief | shadow week plus document timestamps | analyst time tracked across the week; medians from file create-to-send timestamps, trailing 60 days |
| 46 touches · 3.2% reply | email log export, eight weeks | first-touch messages deduplicated; replies matched by thread; internal and existing-relationship mail excluded |
| 39% unlogged · 27 late | calendar-to-CRM cross-check, one quarter | external meetings from partner and VP calendars matched against CRM activity records |
| 1,870 · 397 · 21% | sourcing saved searches vs CRM activity | universe per the firm's own mandate criteria; contacted means one logged first touch in twelve months |
| coverage model | modelled | briefing capacity at 30 targets a week against the universe, net of partner review time; a twelve-month range |
| shadow sessions | five working days, read-only, under NDA | two partners in deal reviews, three analysts on research days, the office manager on an operations day |
Next. A one-hour walkthrough of this page, at the firm’s office. Cross off anything they do not want; only what they approve gets built or billed. If the whole page is a no, they keep it with our thanks: the seats they cancel already paid for the meeting.
