Delivered project
A managed IT provider reaches enterprises across Singapore
Public posts that show enterprises across Singapore hiring IT operations and infrastructure roles are found, classified and scored, and the newest top accounts reach the client's team in a daily digest. Nobody is contacted automatically.
Bring the job you still do by hand. We tell you what would run, what stays with a person, and where to start building it.
01Problem
The job before
A managed IT provider in Singapore wanted a steady list of enterprises across Singapore worth a conversation. Checking job boards, forums and career pages by hand took hours and missed most of them.
02Limits
Constraints
- 01
The constraint: Only public sources, read within each source's terms.
How the build meets it: The sources are public job boards, feeds, forums and career pages, each with its own schedule and health check.
- 02
The constraint: Nobody may be contacted automatically.
How the build meets it: The workflow finds, classifies and ranks. A person on the client's team decides whom to contact.
- 03
The constraint: A post seen before, or reposted, must not count twice.
How the build meets it: An evidence ledger skips posts seen before and merges reposts.
03Build
What we built
- 01A source list and search queries for enterprises across Singapore, kept in a settings table so they change without editing the workflow.
- 02Rule prefilters and an evidence ledger, so most items never reach the AI step.
- 03An AI step that classifies only new posts: signal type, relevance and organisation.
- 04A code score, a results table and a daily digest of the newest top accounts to the client's team.
The route at a glance
- Starts with
- Public posts from job boards, feeds, forums and career pages in this market, read twice a day.
- Arrives at
- Daily digest
- Rule steps
- 2
- AI steps
- 1
- Person steps
- 1
- Held back to
- No side route
The delivered build, in run order.
- Trigger: Public posts from job boards, feeds, forums and career pages in this market, read twice a day.
- Rule step: Cheap rules prefilter, and an evidence ledger skips posts seen before and merges reposts.
- AI step: AI classifies only the new posts: signal type, relevance and organisation.
- Rule step: Code scores and ranks each account and saves it to a table.
- Output: A daily digest of the newest top accounts to the client's team. Nobody is contacted automatically. It arrives at the daily digest.
- Trigger
- Rule step
- AI step
- Where the record arrives
Drawn from the delivered workflow design. No client data appears in it.
04Who does what
Where AI is used
Classifying new posts only: signal type, relevance and organisation. Rules prefilter before it, code scores after it, and the workflow contacts nobody.
Fixed rules
2 steps
- Fixed rules: Cheap rules prefilter, and an evidence ledger skips posts seen before and merges reposts.
- Fixed rules: Code scores and ranks each account and saves it to a table.
AI
1 step
- AI: AI classifies only the new posts: signal type, relevance and organisation.
A person
1 step
- A person: The client's team reads the daily digest and decides whom to contact.
05The change
Before and after the job
Before
Who did the job, and how
- Before: A person searched public job boards, forums and career pages by hand.
- Before: The search ran only when someone had time, so most signals were missed.
After
What runs now
- Runs now: The sources are read twice a day, and posts seen before are skipped.
- Runs now: New posts are classified by AI and scored in code.
- Runs now: The newest top accounts among enterprises across Singapore arrive in a daily digest, and a person decides whom to contact.
What a person still does
- A person: The client's team reads the daily digest and decides whom to contact.