Prospect Research

⚠️ DATA RULE: Never trust pre-training knowledge for current company state. Always pull live data. Pre-training is stale. Internal knowledge first → web research second → synthesize third.

⚠️ DATE RULE: Check currentDate from context before creating any file. All files named YYYY-MM-DD-{topic}.md.

STEP 0 — Confirm Account and Goal

Before any research:

1. Confirm the account name (exact spelling — this becomes the folder path in kebab-case)
2. Is this initial research or a refresh?
3. Check if accounts/{account-name}/ already exists
4. Check if account-brief.md already exists

If no folder exists → create the structure now:

accounts/{account-name}/
├── account-brief.md        ← Create skeleton from _TEMPLATE
├── research/
├── discovery/
├── meetings/
└── emails/

STEP 1 — Internal Knowledge First

Check what already exists before going to the web:

CRM (if MCP configured):

• Search for existing relationship, open opportunities, past interactions
• Pull any contact records for the account

Local files:

• accounts/{account-name}/ — prior research files, call notes, meeting briefs
• knowledge/competitive/ — relevant battlecards
• knowledge/personas/ — ICP persona guides

What to look for:

• Existing CRM relationship / opportunity status
• Known contacts + roles
• Previous engagement history
• Relevant prior research findings

STEP 2 — Web Research

Detect mode first: Check if mcpbrightdata* tools are available.

FULL MODE (Bright Data available)

Run in parallel via mcpbrightdatasearchenginebatch:

Query 1: "{Company} tech stack {{SOLUTION_CATEGORY}} 2025 2026"
Query 2: "{Company} engineering hiring jobs 2026"
Query 3: "{Company} news funding announcement 2026"
Query 4: "{Company} CTO VP Engineering Chief Architect"
Query 5: "{Company} {{SOLUTION_CATEGORY}} challenge problem pain"
Query 6: "{Company} AI machine learning 2026"
Query 7: "{{COMPANY_NAME}} {Company industry} case study customer"

--- Deep Signal Layer (always run) ---
Query 8:  "{Company} engineering blog AI machine learning 2026"
Query 9:  "{Company} AI conference talk keynote 2025 2026"
Query 10: "{Company} hiring AI engineer vector LLM RAG embedding 2026"
Query 11: "{Company} {CTO/VP name} blog post LinkedIn article 2026"

Then scrape for depth:

• Company website → About, Engineering, Careers pages
• Recent blog posts (prioritize AI/engineering content)
• Press releases from last 90 days
• LinkedIn company page (if accessible)

LITE MODE (no Bright Data — WebSearch fallback)

1. "{Company} tech stack 2026"
2. "{Company} CTO VP Engineering LinkedIn"
3. "{Company} news funding 2026"
4. "{Company} {{SOLUTION_CATEGORY}} use case"
5. "{Company} engineering blog AI"

Note in output: "⚠️ Lite mode — search results only, no page scraping. Lower signal quality."

Key signals to find:

• Confirmed tech stack (databases, cloud provider, languages, tools)
• Engineering headcount and growth trajectory
• AI/ML initiatives — what are they building specifically?
• Recent funding, product launches, leadership changes
• Engineering leadership names, titles, any recent changes
• People publishing publicly about AI, data, or {{SOLUTION_CATEGORY}}
• Job postings mentioning specific tools → reveals architecture decisions

STEP 3 — FITS Scoring

Score the account before writing the research file:

Total FITS Score: /100

• 80+: Tier 1 — prioritize immediately
• 60–79: Tier 2 — strong, worth outreach
• 40–59: Tier 3 — possible, lower priority
• Below 40: Tier 4 — stretch, deprioritize

STEP 4 — Synthesize and Save

Output path: accounts/{account-name}/research/{YYYY-MM-DD}-initial-research.md (Refresh: {YYYY-MM-DD}-research-update.md)

Required sections (chain interface — write-outreach reads these):

# {Company} — Research
**Date:** {YYYY-MM-DD} | **Mode:** Full / Lite

---

## 2026 Key Signals
[Fresh, time-sensitive findings. Lead with the most recent.]

## Company Overview
[2-3 sentences: what they do, size, stage, HQ]

## Tech Stack (Confirmed)
| System | Technology | Source |
|--------|-----------|--------|

## Leadership Contacts
| Name | Title | Relevance | Source |
|------|-------|-----------|--------|

## People Publishing About AI / {{SOLUTION_CATEGORY}}
| Name | Title | Topic | Link | Why It Matters |
|------|-------|-------|------|----------------|

## AI / {{SOLUTION_CATEGORY}} Job Postings
| Role | Tools Mentioned | Posted | Implication |
|------|----------------|--------|-------------|

## Pain Signals (Research-Confirmed)
1. [Pain] → Source: [URL]

## Sales Opportunity Analysis
**FITS Score:** /100 | **ICP Tier:** 1/2/3/4
**Sales Motion Route:** Classic / Sprint / Fast / Unknown
**Value Driver:** Make Money / Save Money / Go Fast / Be Safe
**Why Anything:** [What breaks if they do nothing?]
**Why {{PRODUCT_NAME}}:** [Which capability matches their pain?]
**Why Now:** [What external event creates urgency?]

## Top 3 Outreach Hooks
1. [Specific person + specific signal + specific angle. Ready to use.]
2.
3.

## Recommended First Contact
**Who:** [Name + title]
**Channel:** LI / Email / Cold Call
**Hook:** [One-liner]

## Sources
- [URL or document with description]

STEP 5 — Update account-brief.md

After saving the research file:

• Add/update FITS score and ICP tier
• Add any new confirmed contacts
• Update "Last Updated" date
• Update "Top Pain Signal" and "Top Hook" fields

Quality Gates

• Internal knowledge checked before web research
• At least 5 web queries run (including deep signal layer in full mode)
• Tech stack confirmed from source (not assumed)
• At least 1 pain signal is research-confirmed with a cited source
• FITS score calculated
• Sales motion route hypothesis stated with reasoning
• Output file saved with today's date in name
• All required chain interface sections present (Company Overview, Tech Stack, Pain Signals, Leadership Contacts, Top 3 Outreach Hooks)
• account-brief.md updated
• No invented metrics — every number has a source