AI Workflows

AI Workflow for Sales Prospecting and CRM Automation

Automate your entire sales prospecting pipeline — lead research, outreach drafting, CRM updates, and follow-up — with a chain of AI prompts that work together.

FreeLast tested: 2026-07-03Audience: Sales teams, solopreneurs

Why sales prospecting needs a workflow, not a prompt

Most salespeople use ChatGPT the same way: one-off prompts for cold emails, sporadic research, and a CRM they update only when forced. The problem is that each step lives in isolation. A workflow chains them together — the output of one prompt becomes the input of the next, and the CRM stays current without manual data entry.

This article documents a repeatable AI workflow for sales prospecting. It assumes you have access to a general-purpose LLM (ChatGPT, Claude, Gemini) and a CRM with an API or CSV import. Total setup time: under 30 minutes.

The four-stage sales prospecting pipeline

Each stage produces structured output that feeds the next stage. The whole chain runs in under 15 minutes once the prompts are tuned.

Stage	AI task	Output
1. Lead research	Scrape company website + recent news, extract relevant signals	Structured lead profile (JSON)
2. Outreach draft	Generate personalized email sequence from lead profile	3-email sequence (cold → follow-up → break-up)
3. CRM entry	Format lead data into CRM-ready fields	CSV row or API payload
4. Follow-up scheduler	Generate timing + trigger conditions for each touchpoint	Calendar reminders + trigger rules

Stage 1: Lead research prompt

Start with a prospect's company URL. This prompt extracts the signals that matter for a sales conversation.

You are a sales researcher. Given a company URL, extract: 1. What they do (one sentence) 2. Recent signals: funding, product launches, leadership changes, hiring spikes (last 90 days) 3. Likely pain points their current tech stack creates 4. Decision-maker role to target 5. A relevant trigger event for outreach Format as JSON with keys: summary, signals, pain_points, decision_maker, trigger_event Company URL: {paste URL here}

Tip: Feed the LLM the company's "About" page and a recent blog post or press release. Most LLMs can handle 2-3 pages of context per run. For a deeper dive, combine with Brave search or a web research tool to pull recent news automatically.

Stage 2: Outreach sequence from research data

Pass the JSON output from Stage 1 directly into this prompt. It generates a short sequence without generic filler.

You are a senior SDR writing personalized outreach. Using this lead profile: {lead_profile_json} Write a 3-email sequence: - Email 1 (cold): reference the trigger event, state a specific hypothesis about their pain point, offer one concrete insight - Email 2 (follow-up, day 4): add new angle based on industry trend, include social proof - Email 3 (break-up, day 10): leave the door open, one final relevant resource Rules: no flattery, no "{firstName}" tokens, max 120 words per email, end each with a single low-friction CTA.

The output is copy-paste ready. If you use an AI email tool like AI email marketing automation, you can pipe these drafts directly into your campaign builder.

Stage 3: Automated CRM entries

The same research data can generate CRM-ready records. This prompt works with any CRM that accepts CSV imports (HubSpot, Pipedrive, Salesforce) or REST API calls.

Convert this lead profile into CRM fields: {lead_profile_json} Output as a CSV row with headers: company_name, industry, website, decision_maker, title, trigger_event, pain_points, lead_source, notes Lead source: "AI workflow outreach" Notes: include the key signal and suggested next step

For teams using API-connected CRMs, automate this step entirely: feed the CSV output into your CRM's import endpoint via a scheduled script. Manual data entry drops to zero.

Stage 4: Follow-up timing and triggers

The final stage ensures no lead falls through the cracks. This prompt generates a lightweight schedule based on the lead's signals.

Given this lead profile, recommend a follow-up schedule: {lead_profile_json} Output: - Optimal send times (timezone-aware) - Trigger conditions for skipping/accelerating the sequence (e.g., "if they reply to email 1, skip email 2 and move to demo booking") - Calendar block template for the close attempt (30 min) Consider: industry vertical, role seniority, trigger event urgency.

Export the schedule to your calendar or task manager. For high-volume prospecting, pair this with AI customer support tools that can handle inbound qualification while your sequenced outreach runs.

Putting it all together: a single-run workflow

Here's how the four stages connect in practice. Total time for one prospect: under 10 minutes of AI interaction, plus 2 minutes of human review before hitting send.

Research: Paste company URL → get JSON lead profile (3 min)
Outreach: Feed JSON → get 3 personalized emails (2 min)
CRM: Feed JSON → get CSV row, import to CRM (3 min)
Schedule: Feed JSON → get timing + triggers (1 min)
Review: Human reads all output, adjusts tone, queues emails (2 min)

For bulk prospecting (50+ leads), script the JSON-to-CSV and JSON-to-outreach steps using an LLM API. The landing page copy approach we tested uses the same chaining technique but for marketing content — the pattern transfers directly.

Limits and notes

Accuracy: Lead research depends on the quality of the source material. If the company's website is thin or the LLM has no recent training data on them, the pain points will be generic. Always verify trigger events before email 1 lands in someone's inbox.

CRM integration: Stage 3 assumes your CRM has a CSV import or API. For closed CRMs (some enterprise Salesforce instances), you'll need a manual copy-paste step. Consider a middleware tool like Zapier or Make as the bridge.

Compliance: Each region has its own cold email rules. GDPR in Europe, CAN-SPAM in the US. The workflow does not check compliance — review your outreach against local regulations before scaling.

Scaling ceiling: This workflow works well for 10-50 prospects per week. Above that, invest in API-based automation and a proper sales engagement platform.

Browse all articles →