The Prospect-to-Meeting Pipeline
In Lessons 1 through 7, you built every piece of the sales workflow: research, scoring, outreach, sequences, briefs, and follow-up. Each lesson focused on one capability in isolation. But NexaFlow's reps do not execute isolated commands. When a new prospect appears, they run a pipeline: research the company, score the lead, draft outreach if the score is hot, build a multi-touch sequence, prepare a pre-call brief when the prospect responds, and write a follow-up after the call. Six stages, one session, one prospect.
This lesson runs that complete pipeline for a fresh prospect — Crescent Freight, the fifth demo company from your Lesson 1 dataset that you have not yet worked with. You will execute all six stages, trace the data flowing between them, and then run the same pipeline with a deliberately weak ICP to see what happens when the foundation is broken.
The Pipeline
Each stage uses a specific skill from the Revenue Engine extension. Each stage produces output that the next stage consumes.
| Stage | Action | Skill | Input From | Output To |
|---|---|---|---|---|
| 1 | Research the prospect | prospect-research | Company name + context | Stage 2 |
| 2 | Score the lead | lead-scoring | Research brief | Stage 3 |
| 3 | Draft outreach | outreach | Research + score + ICP | Stage 4 |
| 4 | Build sequence | sequence | Research + outreach + ICP | Stage 5 |
| 5 | Prepare pre-call brief | pre-call-brief | Research + score + outreach history | Stage 6 |
| 6 | Write follow-up | follow-up | Meeting notes + prior context | Pipeline complete |
The agent researches, drafts, and recommends. The sales professional decides and sends.
Step 1: Research
Crescent Freight is a mid-size freight forwarding company based in Karachi. You generated their profile in Lesson 1 but never ran intelligence on them. Start the pipeline:
Use the prospect-research skill to research Crescent Freight
Karachi — full intelligence brief
The prospect-research skill auto-activates and produces a structured brief: company overview, key personnel, financial signals, technology stack, pain points, and engagement signals. Review the output against the hallucination detection rules from Lesson 1.
What to check before moving to Step 2:
- Identify the primary buyer (likely VP Operations or Director of Logistics based on NexaFlow's ICP)
- Flag any financial claims that cannot be verified from public sources
- Note the technology signals — are they running legacy systems or modern infrastructure?
- Check for timing signals — new leadership, contract wins, system failures, expansion
The research brief is the foundation. Every downstream stage will reference it. If the research is thin, everything downstream will be thin. If the research is specific, everything downstream will be specific.
Step 2: Score
Feed the research brief into the scoring model:
Use the lead-scoring skill to score this lead: Crescent Freight,
Karachi. Use the research brief from Step 1.
The lead-scoring skill evaluates three dimensions — Fit, Timing, and Engagement — using the ICP you built in Lesson 2.
What to expect: The agent produces a three-dimension score with classification and routing. Your output will vary, but look for these sections:
| Section | Intent | What to Verify |
|---|---|---|
| Fit score | ICP match based on industry, size, geography | References Crescent's actual demo-data.md profile |
| Timing score | Active buying signals | Hiring, expansion, or contract signals from research |
| Engagement score | Prior interaction with NexaFlow | Likely lower — Crescent has not reached out to you |
| Total + Classification | Composite score with HOT/WARM/CULTIVATE/NOT YET | Matches the threshold table from Lesson 3 |
| Routing recommendation | Next pipeline action | Should match the classification tier |
The scores depend on your demo-data.md content and ICP configuration. The teaching point is the dimension analysis — which dimension is weakest and what that means for your outreach strategy. If Engagement is lowest, your outreach needs to work harder to establish relevance. Compare this score to your Meridian score from Lesson 7 to see how engagement level changes the outreach challenge.
Routing decision: HOT — proceed to outreach.
Step 3: Outreach
The score is hot. Now draft a first-touch message:
Use the outreach skill to draft a LinkedIn DM for the VP
Operations at Crescent Freight. Reference the new warehouse
expansion at Port Qasim and their hiring for an operations
manager.
The outreach skill produces a Five Laws-compliant message. Review it against the five constraints from Lesson 5:
| Law | Check |
|---|---|
| Law 1: Specific verifiable reference | Does it cite the Port Qasim expansion or the ops manager hiring? Both are verifiable signals. |
| Law 2: Lead with prospect, not product | Does the first sentence describe Crescent's situation, not NexaFlow's product? |
| Law 3: One ask, one clear next step | Is there exactly one call-to-action, not two or three? |
| Law 4: Hard word limits (under 150 words) | Count the words. |
| Law 5: Zero jargon | Could a non-technical VP read this without stumbling on industry buzzwords? |
If any law fails, iterate. Prompt the agent to fix the specific violation. The outreach skill enforces the Five Laws, but enforcement is not perfect — you are the final auditor.
Step 4: Sequence
One touch is not enough. Build a multi-touch sequence for Crescent:
Use the sequence skill to build a 6-touch, 21-day outreach
sequence for Crescent Freight. Include exit conditions for reply,
bounce, unsubscribe, and silence.
The sequence skill generates a full cadence. Review the output against what you learned in Lesson 6:
Exit conditions to verify:
- Reply: Stop the sequence. Any reply — positive, negative, or "not now" — triggers a human handoff.
- Bounce: Remove from sequence. Invalid contact.
- Unsubscribe: Remove immediately. Legal requirement.
- Silence after 6 touches: Exit to nurture. Do not keep pushing.
Check for personalisation decay across the sequence. Touch 1 should reference the Port Qasim expansion. Touch 6 should still reference something specific to Crescent — not a generic "following up on my previous message." If the later touches decay into templates, that is Over-Automation from Lesson 6. Flag it and ask the agent to maintain specificity.
Step 5: Brief
Assume Crescent's VP Operations responds to Touch 2. She agrees to a discovery call next Thursday. Prepare:
Use the pre-call-brief skill to prepare for my discovery call
with Crescent Freight. The VP Operations responded to our LinkedIn
outreach about the Port Qasim warehouse expansion. Call is
scheduled for Thursday.
The pre-call-brief skill assembles meeting preparation from every upstream stage — research context, scoring rationale, outreach history, and engagement signals.
What to verify in the brief:
- Does it reference the specific hook that got the response (Port Qasim expansion)?
- Does it include discovery questions drawn from the research brief?
- Does it carry forward any hallucination warnings from Stage 1?
- Does it suggest talking points based on Crescent's ICP fit dimensions?
The brief should feel like a document prepared by someone who has been following this prospect for weeks. That is the pipeline working — six stages of accumulated intelligence converging into one preparation document.
Step 6: Follow-Up
Assume the call went well. Crescent's VP confirmed the expansion timeline, mentioned they are evaluating two vendors including NexaFlow, and wants a technical demo next week. Write the follow-up:
Use the follow-up skill to write a follow-up email for Crescent
Freight. Call notes: VP confirmed Port Qasim warehouse goes live
Q3. Evaluating NexaFlow and one competitor. Wants a technical demo
next Tuesday. Key concern: integration with their existing SAP
dispatch system. Action: send SAP integration case study before
demo.
The follow-up skill generates the email. Apply the Context Loss test from Lesson 7: does the follow-up reference specific conversation points, or is it a generic "great speaking with you" template?
Specific references to check:
- Port Qasim Q3 timeline
- SAP dispatch integration concern
- Technical demo on Tuesday
- SAP integration case study as the deliverable
If the follow-up misses any of these, you have Context Loss. The fix is operational: include the call notes in your prompt so the skill has the intelligence to reference. The agent uses what you give it. It cannot reference a conversation it was not told about.
Garbage In, Garbage Out
You just ran a complete pipeline with NexaFlow's calibrated ICP — the one you validated against 20 closed-won deals in Lesson 2. Now run the same pipeline with a deliberately weak configuration.
The Weak ICP
Open sales-marketing.local.md and replace the ICP with a minimal version:
icp:
firmographic:
industries: ["Any"]
company_size:
employees_min: 1
employees_max: 100000
geography:
primary: ["Global"]
No technographic signals. No timing signals. No persona profiles. No negative signals. No data sources. This is the ICP equivalent of "sell to everyone."
Running the Same Pipeline
Run the same six stages for Crescent Freight with the weak ICP. At each stage, compare the output to what you got with the strong ICP:
| Stage | Strong ICP Output | Weak ICP Output | What Changed |
|---|---|---|---|
| Research | Specific intelligence brief | Same — research does not use ICP | No change yet |
| Score | 76/100 — meaningful dimensions | 95/100 — everything scores high because nothing is filtered | Score inflated, signal destroyed |
| Outreach | References Port Qasim, ops manager hiring, specific pain | Generic "we help freight companies" message | Personalisation lost |
| Sequence | 6 touches with Crescent-specific hooks | 6 touches with generic templates | Personalisation decays from Touch 1 |
| Brief | Actionable preparation with discovery questions | Vague preparation with no targeted questions | Meeting advantage lost |
| Follow-up | References specific call points | Generic template regardless of call quality | Context reduced to minimum |
The research brief is identical — prospect-research does not reference the ICP. But from Stage 2 onward, every output degrades. The score is meaningless because everything is a fit. The outreach is generic because the ICP provides no targeting guidance. The sequence decays because there are no prospect-specific hooks to sustain personalisation. The brief is vague because the scoring rationale is empty. The follow-up defaults to a template because there is no upstream intelligence to carry forward.
The pipeline amplifies whatever you put into it. A strong ICP at the top produces specific, actionable output at every stage. A weak ICP at the top produces generic output that gets worse at every stage. This is why Lesson 2 — ICP calibration — is the most important lesson in this chapter. Everything downstream depends on it.
After running the weak-ICP comparison, restore sales-marketing.local.md to the validated ICP from Lesson 2. Do not leave the weak config in place — every command you run from this point forward will use whatever configuration is active.
The Time Comparison
Track how long each pipeline stage takes with the Revenue Engine versus doing it manually.
| Stage | AI-Assisted Time | Farah (Manual) | Ahmed (Skips It) |
|---|---|---|---|
| Research | 2 minutes (generate + review) | 45 minutes (Companies House, LinkedIn, news search) | 4 minutes (skim LinkedIn headline) |
| Score | 30 seconds (generate + evaluate) | 10 minutes (mental scoring against criteria) | 0 minutes (gut feeling) |
| Outreach | 2 minutes (generate + Five Laws audit) | 20 minutes (draft, revise, check specificity) | 3 minutes (send template) |
| Sequence | 2 minutes (generate + review exit conditions) | 45 minutes (plan 6 touches, write each one) | 0 minutes (no sequence, single touch only) |
| Brief | 2 minutes (generate + verify intelligence) | 30 minutes (assemble from notes, CRM, research) | 0 minutes (walks in cold) |
| Follow-up | 1 minute (generate + context check) | 15 minutes (write specific follow-up from notes) | 2 minutes (generic "thanks for the call") |
| Total | ~10 minutes | ~2 hours 45 minutes | ~9 minutes |
Farah spends nearly three hours on a single prospect. She produces excellent output because she does the work. Ahmed spends nine minutes and produces minimal output because he skips most stages.
The pipeline gives every rep Farah's output quality in Ahmed's timeframe. That is the value proposition of the Revenue Engine: research depth in minutes, not hours. But only when the ICP is calibrated. With a weak ICP, the pipeline produces Ahmed's output quality — generic, unfocused, no competitive advantage — regardless of how fast it runs.
What You Built
- Complete prospect-to-meeting pipeline for Crescent Freight (all 6 stages, one session)
- Understanding: the pipeline amplifies config quality — strong ICP produces strong output at every stage, weak ICP produces garbage at every stage
- Time comparison: AI-assisted pipeline (~10 minutes) versus manual preparation (~2 hours 45 minutes)
Flashcards Study Aid
Try With AI
Use these prompts in Claude or your preferred AI assistant to practise the pipeline skills from this lesson.
Prompt 1 (Reproduce)
Run the full 6-step pipeline for Crescent Freight (Karachi):
1. Use the prospect-research skill — full intelligence brief
2. Use the lead-scoring skill — three-dimension lead score
3. Use the outreach skill — LinkedIn DM referencing a specific trigger
4. Use the sequence skill — 6-touch, 21-day cadence with exit conditions
5. Use the pre-call-brief skill — preparation (assume prospect responded)
6. Use the follow-up skill — post-call email (assume call confirmed timeline)
At each step, document: what data arrived from the previous step,
what the agent added fresh, and your total elapsed time.
What you are learning: How to execute the complete pipeline as a connected workflow rather than isolated commands. By tracking data flow between stages, you see where the pipeline preserves intelligence and where it drops context. The elapsed time gives you a baseline to compare against manual preparation.
Prompt 2 (Adapt)
Run the pipeline for a prospect in a DIFFERENT market — pick a
company in the UAE or UK instead of Pakistan.
At each stage, note where the output changes (compliance language,
cultural tone, channel selection, data availability) and where it
stays the same (pipeline structure, scoring dimensions, exit
conditions).
What you are learning: How market context affects pipeline output while pipeline structure stays constant. The scoring dimensions (Fit, Timing, Engagement) work the same way regardless of geography, but the outreach tone, compliance requirements, and data availability change significantly between Pakistan, UAE, and UK markets.
Prompt 3 (Apply)
Run the pipeline for a real prospect from your own network.
At each stage, rate the output:
- READY: could use as-is, no edits needed
- NEEDS EDITING: close but needs tweaks before sending
- REDO: not usable, would need to regenerate with better context
What is your average across all 6 stages? Where did the pipeline
produce the strongest output? Where did it need the most correction?
What you are learning: How to evaluate pipeline output quality against your professional standards. A pipeline that produces all-READY output is configured well. A pipeline that produces mostly-REDO output has a config problem upstream — usually the ICP, the brand voice, or the competitive intelligence. Your ratings tell you where to invest configuration time.