AI SALES OPERATIONS
AI Sales Operations System
An end-to-end sales system that finds prospects, prepares personalized outreach, handles inbound conversations through an AI voice agent, records qualified opportunities, checks calendar availability, and books demo calls.
n8n · ElevenLabs · Pipedrive · Google Calendar · Gmail · Firecrawl · Hunter · OpenRouter
SYSTEM OVERVIEW
One system. Two sides of the sales process.
The system connects outbound prospecting with inbound sales conversations. The outbound engine identifies prospects, prepares CRM records, and creates personalized outreach drafts. The inbound engine uses an AI voice Account Executive to understand enquiries, record opportunities, check availability, and book demo calls.
Outbound prospecting and inbound voice sales share the same CRM layer.
THE PROBLEM
Sales work gets fragmented long before it gets difficult.
01
Prospecting lives in too many tools
Research, contact discovery, qualification, and CRM entry often happen separately.
02
Inbound enquiries lose context
Useful information from calls can disappear before it reaches the CRM or the person responsible for follow-up.
03
Booking creates repetitive admin
Availability checks, calendar creation, attendee details, CRM updates, and notes all have to stay synchronized.
04
AI can create new failure points
If the system guesses missing information, duplicates CRM records, or confirms a booking that failed, automation creates more problems instead of fewer.
THE SYSTEM
Outbound finds opportunities. Inbound handles the conversation.
OUTBOUND
ICP → Prospecting → RevOps → SDR → Gmail Drafts
The outbound engine starts with an Ideal Customer Profile, researches matching prospects, finds and verifies contact information, creates or reuses CRM records, and prepares personalized outreach for human review.
INBOUND
Caller → AI Voice Account Executive → Understand + qualify → Record Deal OR Book Demo
The inbound engine handles natural voice conversations, collects useful context, and chooses a controlled action depending on the caller’s intent.
OUTBOUND ENGINE
From ICP to reviewed outreach.
01
Define ICP
Niche, location and prospect criteria
02
Prospecting
Find matching businesses and contact details
03
RevOps
Check, clean and prepare CRM records
04
SDR
Generate personalized outreach from prospect context
05
Gmail Draft
Prepare communication for human review
The system does not automatically send cold outreach.
AI prepares personalized communication, while the final sending decision remains with a human.
INBOUND SALES
A voice agent that understands the enquiry before taking action.
The inbound side uses an ElevenLabs voice agent as an AI Account Executive. It speaks naturally with callers, understands what they need, captures useful project context, and decides whether the opportunity should be recorded for follow-up or moved into demo booking.
CALLER ↓ AI ACCOUNT EXECUTIVE ↓ UNDERSTAND + QUALIFY ↓ RECORD OPPORTUNITY OR BOOK DEMO
PATH 01
Interested, but not ready to book.
When a genuine opportunity does not result in a demo booking, the voice agent sends structured conversation details to the Deal Recording workflow. The workflow searches Pipedrive using the caller’s confirmed email address, reuses existing people, organizations, and open deals where appropriate, then attaches a fresh note with useful context.
01 CONFIRM CALLER EMAIL · 02 SEARCH PERSON · 03 REUSE OR CREATE ORGANIZATION · 04 REUSE OR CREATE OPEN DEAL · 05 RECORD CONVERSATION CONTEXT
Successful test execution: opportunity recorded with person, organization, deal, and note context preserved.
PATH 02
From conversation to confirmed calendar event.
When the caller wants to schedule a demo, the voice agent collects the required booking information and sends it to a deterministic booking workflow. The system checks Google Calendar before creating anything. If the requested slot is unavailable, the voice agent asks for another time. If it is available, the workflow creates the event, adds the attendee, updates CRM records, and records a booking note.
01 EMAIL · 02 DATE, TIME + TIMEZONE · 03 CHECK AVAILABILITY · 04 CREATE EVENT · 05 CRM RECORDS · 06 BOOKING NOTE · 07 RETURN STATUS
The voice agent is not allowed to say a meeting is confirmed until the booking workflow returns a successful result.
Successful test execution: demo booked for the confirmed time with CRM and calendar records created.
RELIABILITY
The happy path was not enough.
The system was tested against failure cases that appear in real voice and automation workflows.
01
Duplicate CRM records
Existing people, organizations, and open deals are reused instead of recreated.
02
Occupied calendar slot
The system returns an unavailable result and the agent asks for another time.
03
Missing booking information
The agent cannot call the booking workflow until date, time, timezone, and confirmed email are present.
04
Optional phone number
A caller can decline to provide a phone number without breaking the CRM workflow.
05
Voice email errors
The agent spells the address back character-by-character and requires explicit confirmation.
06
Failed tool actions
The agent cannot claim a booking or CRM action succeeded unless the backend confirms success.
VOICE QA
One missing letter can break the workflow.
During testing, a spoken email address was transcribed incorrectly. The agent was updated to spell the local part back character-by-character and require explicit confirmation before proceeding.
e-m-m-a dot b-o-o-k-i-n-g-t-e-s-t at example dot com
If the caller corrects any character, the agent repeats the corrected address and confirms it again before proceeding.
SYSTEM DESIGN
AI handles ambiguity. Workflows handle business logic.
AI is useful where the system has to understand language, interpret intent, extract conversational context, or choose the appropriate next action. Deterministic workflows are better for validation, CRM lookup, duplicate checks, calendar availability, record creation, and HTTP responses.
01
AI for judgment
Understanding enquiries, extracting context, qualifying intent, and writing personalized outreach.
02
Deterministic logic for actions
CRM operations, availability checks, validation, record creation, and status handling.
03
Human control where it matters
Outbound communication remains a draft until reviewed and approved.
Sometimes the better AI system is the one where AI does less.
REAL EXECUTIONS
Built, tested, and debugged beyond the demo.
DEAL RECORDING · status: recorded
DEMO BOOKING · status: booked
TECH STACK
Connected tools, controlled responsibilities.
n8n
Workflow orchestration and deterministic logic
ElevenLabs
Voice AI Account Executive
Pipedrive
People, organizations, deals and notes
Google Calendar
Availability and demo scheduling
Gmail
Reviewed outbound drafts
Firecrawl
Prospect research
Hunter
Contact discovery and verification
OpenRouter
LLM access for AI workflow tasks
OUTCOME
A sales system, not a collection of automations.
The completed system connects prospect research, personalized outreach, CRM operations, conversational AI, qualification, opportunity recording, calendar availability, and demo booking into one coordinated sales architecture. The important part was not one successful execution; it was predictable behavior when information was missing, a calendar slot was occupied, an email was misheard, or a CRM record already existed.
The system uses AI where judgment is useful and controlled workflows where reliability matters.
Let’s find out what is actually worth automating.
Tell me where your sales process is repetitive, slow, or difficult to manage.