Casebook · Finance & Business Services · New Zealand

A NZ Mortgage Adviser Team: Client Touchpoints That Start with a Birthday Email

Partly liveRecords as of 9 Oct 2026BEE Sigma Delivery Team

The client is a small New Zealand mortgage adviser team serving the Chinese-speaking community. We connected an AI agent to two systems they already used: their business Outlook mailbox and Lark Base. Status: partly live. The client birthday email, connected between 25 and 28 August 2026, runs daily and is the only workflow our records confirm as live. Xiaohongshu content support was deployed with training on 3 September 2026; a daily mortgage brief has been sent as a one-month trial since 5 August 2026.

IndustryFinance & business services (mortgage advice)
MarketNew Zealand, Chinese-speaking clients
Delivery statusPartly live: birthday emails running; Xiaohongshu support deployed with training; daily brief on trial
Period15 Jun 2026 – 14 Sep 2026 (one quarterly engagement)
Systems connectedBusiness Outlook mailbox (Microsoft Graph authorisation), Lark Base; Xiaohongshu posted manually, no API connection
Decisions kept by peopleMailbox permission scope approved by the client's administrator; Xiaohongshu posts reviewed and published by the team
01

How it worked before

Client birthdays depended on an adviser remembering and writing a note by hand. Xiaohongshu is the team's main channel for new clients, but the account had too little past content to settle on a voice, so every topic and image started from scratch.

Before and after
Before and after
02

What AI does now, and what people do

The birthday email is a scheduled job: the agent checks daily and sends greetings through the authorised mailbox. For Xiaohongshu we set up a topic pool in Lark Base and fixed rules for cover and content images, so advisers can draft posts by chatting with the agent. A daily 08:00 NZT brief collects mortgage, OCR and refix news as topic input.

  • AI does: Sends client birthday emails daily; drafts Xiaohongshu posts and images from the topic pool and image rules; compiles the 08:00 market brief.
  • People do: Approve mailbox permissions; review and publish Xiaohongshu posts by hand; decide which brief items are worth turning into content or passing to clients.
Process: what the AI does and what people do (gold = people)
Process: what the AI does and what people do (gold = people)
03

Timeline

All dates come from project chat and meeting records.

  • 2026-06-15Kick-off and first deployment.
  • 2026-07-20Collaboration channel working; image style tuning began and the topic pool table was set up.
  • 2026-08-0345-day review meeting.
  • 2026-08-05Daily 08:00 mortgage market brief started as a one-month trial.
  • 2026-08-25Business mailbox authorisation completed, approved by the client's administrator.
  • 2026-08-28Automated client birthday email confirmed and connected.
  • 2026-09-03Xiaohongshu content skill deployed, with a training session.
  • 2026-09-08The model in use was retired by its provider and the agent stopped responding; we upgraded the runtime and switched models the same day.
  • 2026-09-14First quarterly engagement ended.
Timeline: every point comes from project records
Timeline: every point comes from project records
04

Current status and limits

Only the birthday email is confirmed as running in our records. Xiaohongshu support is deployed and taught, but we have no data on post volume, reach or leads, so we make no claim about results. The daily brief is a trial, and our records show no decision beyond it. A loan report layout request was not delivered within our records and is not counted here. The brief is a news digest, not financial advice.

05

What we learned

When an account has little past content, a fixed skill may not fit yet; drafting by conversation with the agent is easier to adopt, and rules can be captured once content builds up. The outage on 8 September 2026 also taught us that models have a lifecycle: a retired model shows up as silence, not an error, so troubleshooting should start by checking the model's status.

Scope note: an anonymised case is a summary of project records with client names, people and amounts removed. This page draws on those records; client names, people and amounts are withheld. Dates and figures come from project records and some have not been confirmed in writing by the client. The delivery status is as labelled at the top; results vary with each business’s systems, data and involvement and are not a promise to other businesses.

FAQ · Quick answers

Where should a mortgage adviser team start with AI?

In this project, the first workflow confirmed as live was the client birthday email: small, rule-based, and needing only mailbox authorisation and a scheduled job. It was connected on 28 August 2026 and runs daily. Content and brief features came later and are still at deployment or trial stage.

Is it safe to connect an AI agent to a business Outlook mailbox?

Authorisation was approved by the client's own administrator and completed on 25 August 2026. We explained read-only and read-write permission levels separately so the client could choose the scope. Related Lark Base tables are not shared through public links.

Does the AI post to Xiaohongshu by itself?

No. There is no API connection to Xiaohongshu. The agent drafts posts and images from the topic pool and image rules; the team reviews and publishes them by hand. We have no performance data on posts, so we promise no follower or lead growth.
More anonymised cases
← All anonymised cases

Hand your first workflow to the agents

One free scan shows how visible you are in the AI era; the AIM assessment finds your best angle of adoption.