Blackford AI WorkshopVoice profile + Juniper Square / Headless GPX ideas
Draft-only, human-approved
Private equity AI working session

Build the first useful assistant for Blackford-style IR work.

A live page for sitting down with a Blackford Capital contact: learn his email voice, create one reusable drafting skill, then pressure-test Juniper Square / Headless GPX use cases for investor relations and fund operations.

Prepared for Tucker Carlile. Public research suggests the contact may be Nhat Vu, Investor Relations Analyst at Blackford Capital. If the name is different, the workflow still holds.

01Use this in the room

The two-session agenda

Do not start with “AI can do anything.” Start with one personal win, then one Blackford-specific systems conversation.

1

Connect Outlook / email context

Use sent emails only at first. Learn voice, tone, greeting style, level of detail, and common follow-up patterns.

2

Create the voice profile

Generate a small voice document: how he writes, what he never says, how formal he is with LPs, colleagues, founders, and internal updates.

3

Run one email drafting skill live

Pick one real thread. Draft only. Human approves. Capture the correction so the skill gets better.

4

Shift to Juniper Square

Ask what data lives in Juniper: LP records, fundraising pipeline, interactions, documents, reports, investor-center activity.

5

Pick one safe Juniper skill

Start read-only: tearsheet, LP briefing, fundraising pipeline summary, investor follow-up draft, or data hygiene report.

Today’s win condition

He leaves with one voice profile, one email skill he can reuse in Claude/ChatGPT, and a short list of Juniper Square use cases worth testing next.

02Voice first

Build the Blackford email voice skill

This copies the Brandon/Jim pattern: connect email, learn voice from sent mail, draft replies only, then save the prompt as a reusable skill.

Immediate valueDrafts in his voice

LP follow-ups, meeting recaps, diligence asks, internal reminders, and scheduling replies.

SafetyNever send automatically

Everything is a draft. No commitments, numbers, terms, or forward-looking claims without human review.

MemoryCapture edits

When he changes a draft, save the rule: too formal, too long, wrong greeting, missing context, wrong level of certainty.

PE fitSegment by audience

Different tone for LPs, internal team, founders/sellers, advisors, and portfolio-company leaders.

You are helping create my private email voice profile for investor-relations and private-equity work.

Your job is to learn how I actually write from examples, then create a reusable voice guide.

Inputs:
- 10-20 recent sent emails or drafts
- The audience for each email if known: LP/investor, internal team, founder/seller, advisor, portfolio company, other

Steps:
1. Separate real writing from one-word replies, forwards, newsletters, auto-responses, and throwaway messages.
2. Identify patterns by audience: greeting, sentence length, formality, level of detail, sign-off, common phrases, and how direct I am.
3. Extract what makes my writing sound like me.
4. Identify things I do not want: generic AI language, over-polish, fake warmth, unsupported claims, or commitments I did not make.
5. Ask me what feels wrong before saving the profile.

Output:
- My voice in one paragraph
- Audience-specific rules
- Do / do not list
- Example phrases to reuse
- Phrases to avoid
- Rules for drafting follow-up emails

Important:
- Draft only. Never send.
- Do not invent facts, dates, numbers, commitments, or fund/company claims.
- If a fact is missing, mark it [VERIFY].
- Keep sensitive investor and fund information private.
03Juniper Square / Headless GPX

The real systems wedge

Juniper Square describes Headless GPX as an AI operating layer for private markets, available to MCP-compatible clients like Claude, Copilot, ChatGPT, Gemini, and custom tools. Treat this as the structured-data layer, not just another CRM.

Plain-English framing

If Outlook teaches the assistant how he communicates, Juniper Square teaches it who the investors are, where each relationship stands, what documents exist, and what the next fund/LP action should be.

Important access note

Headless GPX appears to be an official Juniper Square MCP/API product, but not a public self-serve package with open install docs. The right next step is to ask Juniper to enable Headless GPX or developer API access for the account, then test read-only workflows first.

Investor tearsheet

Summarize one LP: relationship history, commitments, recent activity, docs viewed, last touch, open questions, next best action.

Fundraising pipeline report

Show stages, stale prospects, next actions, missing materials, and likely blockers before the next IR meeting.

Personalized LP email draft

Draft a note grounded in LP profile, prior interactions, fund materials, and approved language.

Data hygiene report

Flag missing tags, stale contact data, inconsistent notes, missing follow-ups, or investor records without owner/action.

Investor-center check

Confirm current reports/notices are named, categorized, and available to the right audience.

Capital-activity assistant

Draft internal explanations for capital calls, distributions, and reporting events. Human/legal review required.

Security boundary

Start read-only. Do not let a general chat agent write to investor records, send LP emails, change fund documents, or expose confidential data until permissions, audit logs, and approval gates are explicit.

04Blackford-specific map

Where AI likely fits Blackford

Blackford is a Grand Rapids-based lower-middle-market private equity firm focused on majority-control investments in founder/family-owned companies, especially manufacturing, distribution, and services. Public materials point toward practical needs in LP communication, deal flow, portfolio updates, and operational reporting.

LP reporting assistant

Draft quarterly updates, KPI commentary, portfolio narratives, and variance explanations from source docs.

Investor Q&A copilot

Answer LP questions from approved fund docs, past updates, CRM notes, and investor-center materials.

Deal intake triage

Summarize CIMs, teasers, and broker emails against Blackford criteria: sector, revenue, EBITDA, ownership, cash flow, and add-on fit.

Portfolio KPI narrator

Turn monthly financials/operating notes into management-ready commentary: drivers, risks, margins, customers, working capital.

Add-on acquisition screener

Rank targets by geography, service line, strategic rationale, and likely fit with platform companies.

Board prep assistant

Create pre-read summaries, agenda drafts, decision logs, follow-up lists, and cross-portfolio pattern notes.

05Why this is timely

Private markets are moving from chat to workflow

The best live demos are not “ask AI anything.” They are narrow analyst/IR workflows with source material, approval gates, and clear output.

Deal sourcing and screening

AI is being used to screen CIMs, banker materials, company data, and predefined criteria for faster go/no-go triage.

Diligence and data-room review

Summarize files, flag missing disclosures, detect inconsistencies, and draft diligence questions from sanitized materials.

LP updates and fundraising

Draft first-pass LP updates, DDQ responses, bespoke investor responses, and follow-up emails from approved source docs.

CRM hygiene and meeting prep

Prepare LP/founder meeting briefs, identify stale relationships, capture next actions, and improve pipeline discipline.

Portfolio KPI monitoring

Turn KPI sheets and operating notes into trend summaries, board questions, and follow-up actions.

Finance-tool ecosystem

Claude financial-services agents, Excel/PowerPoint/Word add-ins, and CRM agents are converging around the exact PE workflow layer.

Avoid live unless sanitized

Do not use real LP names, commitments, side letters, fund terms, live data rooms, or automated investor email sending in a first meeting. Use mock records or approved public/sample material for demos.

06Copy-ready prompts

Run these live

These are designed for the room: easy to paste, easy to discuss, and useful even before a full integration exists.

1. Email drafting skill

You are my private email drafting assistant for private-equity / investor-relations work.

Use my voice profile and the full email thread.

Classify the email:
- LP / investor
- internal team
- founder / seller
- advisor / banker / broker
- portfolio company
- scheduling / admin
- other

Then produce:
1. One recommended draft response in my voice.
2. A shorter version if appropriate.
3. Any facts I need to verify before sending.
4. Any follow-up task that should be tracked.

Rules:
- Draft only. Never send.
- Do not invent facts, dates, fund terms, performance data, commitments, or promises.
- Use [VERIFY] for missing facts.
- Keep confidential data out of the response unless it was already in the thread and is appropriate for the recipient.
- If the email is sensitive, recommend human review instead of drafting.

2. Juniper Square use-case discovery

You are helping me identify safe, high-value AI workflows around Juniper Square / Headless GPX.

First, interview me before recommending anything.

Ask about:
- What data we store in Juniper Square
- Which workflows are most painful: LP reporting, fundraising pipeline, investor follow-up, document management, investor onboarding, capital activity, CRM hygiene
- Which actions are read-only vs. write/send/change
- What permission boundaries and approval steps are required
- What a successful first 30-minute demo would look like

Then recommend:
1. Three read-only workflows we can safely test first.
2. One email-drafting workflow that stays human-approved.
3. One CRM hygiene workflow.
4. One workflow we should avoid until permissions/audit logs are clear.
5. The exact first prompt to run against sample or demo data.

3. Investor tearsheet / meeting prep

Create an investor tearsheet for this LP/prospect.

Use only the source material I provide or connected, permissioned records.

Output:
- Relationship snapshot
- Last touch and next best action
- Current fundraising / fund status
- Documents or reports they may care about
- Open questions or concerns
- Personalized talking points
- Suggested follow-up email draft
- Items to verify before sending

Rules:
- Separate verified facts from inference.
- Do not invent fund performance, terms, or commitments.
- Mark missing data as [VERIFY].
- Draft only. Human approves all outreach.

4. Deal intake triage

Review this inbound deal material against Blackford-style criteria.

Use the source document only.

Output:
- Company summary
- Sector
- Revenue / EBITDA if stated
- Ownership / founder-family fit if stated
- Why it may fit manufacturing, distribution, services, or add-on strategy
- Red flags
- Missing diligence questions
- Recommended next step: pass, ask for more info, or discuss internally

Rules:
- Do not infer numbers that are not stated.
- Separate source facts from analysis.
- Keep the summary concise enough to forward internally after review.
Workshop sentence

“The first win is not replacing judgment. The first win is making every investor email, meeting prep, and CRM note start from better context.”

07Folder / skill model

Simple workspace shape

This mirrors the Brandon/Jim pattern: one folder, a reference layer, and reusable skills.

Blackford AI Operating Center/ ├── reference/ │ ├── voice-profile.md │ ├── approved-language.md │ ├── audience-rules.md │ ├── sensitive-data-rules.md │ └── juniper-square-notes.md ├── skills/ │ ├── email-drafter.md │ ├── investor-tearsheet.md │ ├── fundraising-pipeline-brief.md │ ├── deal-intake-triage.md │ └── crm-hygiene-check.md └── runs/ └── YYYY-MM-DD-workshop-notes.md