Admissions operations, from first enquiry to enrolled student

Role

Operator-builder. Senior Course Advisor and interim Regional Admissions Manager.

Scope

Enquiry through enrolment handover for 10 to 12 Course Advisers across four countries and 7+ university partners.

Stack

Cloud-hosted n8n, HubSpot, Gmail, Dialpad, Google Workspace, DocuSign, Xero, OpenAI and Gemini.

Status

Developed against live operations. Components reached different levels of production maturity; these diagrams are sanitised reconstructions.

The problem was fragmented context. A prospect's email sat in Gmail, their history and ownership in HubSpot, call evidence in Dialpad, and the current course offer in spreadsheets. Agreements, payment details and the Student Services handover lived elsewhere again. Delays accumulated in the joins between those tools.

I built the joins. Using cloud-hosted n8n, I turned the existing stack into an event-driven admissions process. Rules handled routing and administration. OpenAI and Gemini helped where the work involved language or interpretation. People retained the decisions that committed the provider or the applicant.

• 24h+ → ~10haverage enquiry reply turnaround
• 2 days → same dayenrolment handover turnaround
• 1 + 1one Student Services role removed, one redeployed
• 400+new leads in a typical month

1. The operation behind the workflow

One field carried too much. Prospects entered HubSpot as contacts, then moved through open, second call, third call, contacted, warm, hot, cold and qualified values in one general lead-status property. Those values mixed call attempts, engagement, sales temperature and qualification. Ownership could follow an automatic assignment or the adviser who booked and held the consultation.

The offer kept moving. Cyber Security, Data Science & AI and Software Engineering courses ran full time or part time, remotely or in person, across timezones and university partners. Cohorts rotated each quarter, with current dates, prices, cutoffs and webinars maintained in shared sheets, and ran with 4 to 18 students. A valid answer depended on that live operating context.

Volume exposed every delay. About 10 to 12 advisers worked across New Zealand, Australia, Singapore and the United States. Each handled roughly 20 to 40 email enquiries a week, while the team made about 40 to 80 calls and held 1 to 6 booked hour-long consultations a day. The Admissions Manager took about 20 more shared-inbox enquiries directly; the team closed 3 to 12 enrolments a month. Manual copying created small delays at every stage.

2. One connected customer journey

Every event gained context first. An inbox watcher covered the shared enquiry address and individual adviser accounts. n8n matched the sender to HubSpot, resolved the owner, read the relevant engagement history and custom properties, then combined that context with approved course information and the current timetable.

flowchart TD
    A["New email, call or webinar event"] --> B["Identify HubSpot contact and owner"]
    B --> C["Assemble current context"]
    C --> C1["CRM history and properties"]
    C --> C2["Course knowledge and sales playbook"]
    C --> C3["Dates, cutoffs and webinar schedule"]
    C1 --> D["Prepare draft or next action"]
    C2 --> D
    C3 --> D
    D --> E{{"Adviser reviews communication"}}
    E --> F{"Business outcome"}
    F -->|Missed first call| F1["Recovery sequence"]
    F -->|Webinar attended or missed| F2["Matched follow-up sequence"]
    F -->|Consultation complete| F3["Notes, email and next tasks"]
    F -->|Prerequisite route| F4["Preparatory study and completion evidence"]
    F1 --> G["Managed follow-up queue"]
    F2 --> G
    F3 --> G
    G --> H{{"Adviser confirms enrolment"}}
    F4 --> P{{"Admissions reviews completion evidence"}}
    P --> H
    H --> I["Prepare agreements, payment records and handover"]
    I --> Q{{"Human confirms documents and payment terms"}}
    Q --> J["Student Services receives the enrolment pack"]

    classDef human fill:#f6ad55,stroke:#c05621,color:#1a1a1a;
    class E,H,P,Q human;
  

People kept the commitments. The system could prepare a reply, queue a task or assemble a document pack. An adviser still sent customer communication, Admissions still confirmed the prerequisite route, and a person checked enrolment documents and payment terms before handover.

3. The system reacted to business events

The trigger was operational. Each workflow began with a change the team already understood, then gathered the information needed for the next action.

• New enquiryMatch the sender to a contact and owner, gather live context and prepare a reply. The adviser checks and sends the draft.
• Missed first callPlace the contact into a recovery sequence and schedule the next attempt. The sequence sets timing; the adviser makes the call.
• Webinar outcomeSend attendees and no-shows into different follow-up journeys. The team approves the sequence; contact ownership stays in HubSpot.
• Consultation completeTurn the transcript into draft notes, tailored follow-up and next tasks. The adviser corrects transcript errors before anything becomes a record.
• Second or third call dueCreate a prioritised HubSpot task and Calendar block. A manager activates the structured mode and sets the operating priority.
• Enrolment confirmedPrepare agreements, student records, payment details and the handover. People confirm admissions and financial terms before documents move.

4. AI only where language required it

Rules carried the routine load. Contact assignment, status changes, follow-up enrolment, calendar scheduling, course completion triggers, document movement and handover all had explicit conditions. They did not need a model to improvise.

Models handled language. OpenAI and Gemini supported enquiry drafting, transcript extraction, CRM note preparation and the identification of coaching moments. Retrieval-augmented generation kept drafts grounded in approved course information and sales language rather than relying on a model's memory.

The playbook came from practice. I used a structured AI pass over thousands of earlier sales-email conversations to identify how the team answered recurring questions, when a consultation added value, and which approved value and payment-plan language belonged in a reply. That became reusable guidance for the advisers reviewing each draft.

Fresh facts stayed separate. HubSpot supplied the person's context. The knowledge layer supplied course and sales guidance. Sheets supplied live dates, cutoffs and webinars. The adviser saw the assembled result on the contact record, with ownership tagged, and made the final call on what to send.

5. One call became several usable records

One transcript fed several jobs. Dialpad calls fed one extraction flow that prepared the HubSpot note and a tailored follow-up email. It also pulled agreed course dates and payment-plan details so the right document set could be prepared for review.

Coaching used the same evidence. The flow marked discovery, programme detail, tailored pitch points, objections and the close, then compared adviser and lead talk time. Managers could configure phrases or topics for a current coaching focus rather than commission a new analysis each time.

Structured fields supported coaching. They could be compared across calls to show objection mix, conversation balance and movement over time. The design turned one call into sales follow-up, CRM hygiene and coaching evidence without asking the adviser to recreate it three times.

6. When autonomy needed tightening

Managers needed a control mode. During a performance review, open pipelines could not depend on an adviser deciding when to revisit each contact. I built a manager-triggered mode that translated open, second-call and third-call contacts into a prescribed daily queue.

flowchart LR
    M{{"Manager activates structured mode"}} --> A["Open, second-call and third-call leads"]
    A --> B["Apply timing rules and manager priorities"]
    B --> C["Create HubSpot tasks and fixed Calendar blocks"]
    C --> D["Adviser works the prescribed queue"]
    D --> E{"Completed?"}
    E -->|Yes| F["Update lead status and next due action"]
    E -->|No| G["Return incomplete work to the queue"]
    F --> A
    G --> A

    classDef human fill:#f6ad55,stroke:#c05621,color:#1a1a1a;
    class M human;
  

The calendar became the queue. n8n applied due dates and manager priorities, then created the HubSpot tasks and fixed Google Calendar blocks. Completion updated the lead state; incomplete work returned to the queue. While this mode was active, it deliberately replaced adviser self-scheduling.

7. Closing the gap after the sale

Enrolment crossed a boundary. Once a sale closed, the active HubSpot lead became a separate student or customer process. The handover had to carry the contact history, chosen course and date, student details, documents and payment arrangement without losing the decisions already made during the consultation.

The pack assembled from source records. The workflow prepared contracts or student agreements, DocuSign documents, Google records, payment-plan fields, Xero-linked finance steps and Calendar scheduling. It also drafted the handover from adviser to Student Services from the full HubSpot profile for a person to check.

Alternative entry stayed procedural. Applicants without the standard degree prerequisite were directed to designated free, self-paced courses. Completion triggered evidence collection and an adviser case for the organisation's approval process. The automation handled evidence and administration; academic approval remained with the organisation.

8. What changed for the operation

Customers waited less. Average enquiry reply turnaround moved from just over 24 hours to about 10 hours. Enrolment handover moved from roughly two working days to the same day, which meant Student Services received the agreed context while it was still current.

Administrative capacity shifted. The automation enabled one Student Services role to be removed and another to be redeployed. HubSpot remained the source of truth while the automation moved work between the tools the team already used.

9. Evidence and transfer

The source systems stayed with the employer. This programme was developed against the provider's live operation, with individual workflows reaching different levels of production maturity. I do not retain or publish employer-owned workflows, customer records, logs or credentials. These diagrams are sanitised reconstructions; the scope and approximate outcomes are available for verification by my former direct manager.

The same handoffs recur elsewhere. Enquiry, qualification, appointment, follow-up, paperwork, payment and delivery handover appear in education, professional services and other relationship-led businesses. I find where context is lost between the tools a team already uses, decide what belongs in rules and what benefits from a language model, and leave consequential decisions with the right person.