5 AI Surfaces That Transformed a Singapore Platform's Commercial Team

This is an anonymised account of a 12-week AI workforce redesign at a Singapore digital media platform with a community of roughly 170,000 members. Four commercial roles were redesigned and five AI surfaces were built and shipped to production. For each surface: the problem it solved, what it does operationally, and the outcome it produced.

1. AEO Intelligence

The problem. The platform's SEO team was optimising for a search landscape that no longer exists. AI search answers cite sources differently than ranked links do, and the team had no visibility into whether — or where — the platform was being cited by external LLMs. They were flying blind on the channel that was quietly becoming their acquisition front door.

What it does. AEO Intelligence pairs Search Console data with external-LLM citation tracking: it monitors where and how the platform appears in AI-generated answers, flags citation drops, and surfaces the content gaps that cause them. It runs unattended and reports on a schedule.

The outcome. The SEO/AEO analyst role was reconfigured into an AEO Strategist: the data-pull work is automated, and the human time goes to interpretation and content strategy — Anchor work that the old role never had hours for.

2. Prospecting Assistant

The problem. Cold-call prospectors spent about 70% of their week on list research, manual enrichment, and sequence maintenance. Fifteen percent of the role was actual conversation with advertisers.

What it does. The Prospecting Assistant runs a CRM-and-enrichment pipeline: it compiles and enriches prospect lists, drafts outreach, and scores interest signals as replies arrive. Humans work the scored queue, not the spreadsheet.

The outcome. The role became the Advertiser Solutions Consultant — 10% pipeline oversight, 25% AI-assisted work, 65% consultative selling. More qualified conversations per week than the old role produced in a month of list-building.

3. Account Health Monitor

The problem. Account managers were reactive to churn, not predictive of it. Warning signs lived in reporting feeds nobody had time to read until the advertiser had already left.

What it does. The Account Health Monitor ingests reporting feeds, flags churn-risk accounts against learned patterns, and assembles a pre-meeting brief for every account conversation — context, history, risk flags, recommended agenda.

The outcome. The advertiser-retention manager role became the Account Growth Partner. Monitoring is automated; the human hours go to intervention and expansion conversations — the work that actually retains revenue.

4. Client Research Briefing

The problem. Every client meeting began with manual research: an hour or more trawling an advertiser's public footprint before a single strategic sentence was written.

What it does. A daily public-footprint crawl per advertiser, distilled into a weekly digest and a per-meeting brief. Research that took an hour of human time now arrives before the calendar invite does.

The outcome. Meeting preparation time fell to near zero. The reclaimed hours moved into the meeting itself — preparation became strategy time.

5. Forum Moderator Copilot

The problem. Roughly 75% of moderator time went to queue triage: sorting, tagging, and routing posts before any judgement was applied. The community-strategy work — the reason the role exists — was starved.

What it does. A layered moderation pipeline: automated first-pass classification, crisis routing that escalates sensitive threads immediately, and a copilot that drafts moderation actions for human confirmation. Nothing is actioned without a human; the queue simply stops consuming the day.

The outcome. The forum moderator role became the Community Health Lead — focused on community strategy, member health, and escalation judgement rather than queue throughput.

Engagement scope

Twelve weeks, S$78K, WDG(JR+) co-funding eligible — with the grant evidence trail (task-audit baseline, Job Redesign Workbook, training cohort, analytics dashboard) produced as a by-product of the engagement from week one. The workforce analytics dashboard has been live since week eight, reporting hours reallocated per role, skills-assessment closure rate, and revenue-per-FTE delta.

Five surfaces, four redesigned roles, one dashboard — and a task mix that actually changed. That is what an AI workforce programme looks like when the redesign comes first.

If your commercial team is spending more than half its time on tasks that should be automated, email hello@qlabshq.com.

Build order and dependencies

The five surfaces were not built in the order they are listed here. Build order followed data readiness: the Prospecting Assistant went first because CRM access was provisioned in week one and its output — a scored queue — unblocked the redesigned sales role earliest. AEO Intelligence followed, needing only Search Console. The Account Health Monitor waited on reporting-feed integration; the Client Research Briefing on the crawl infrastructure the Monitor introduced; the Moderator Copilot last, because layered moderation pipelines need the classification learnings from the earlier surfaces.

The dependency lesson generalises: sequence surfaces by data readiness and by which redesigned role they unblock first, not by perceived importance. A surface that unblocks a redesigned role in week four is worth more than a more impressive surface that lands in week ten.

What the dashboard showed at month six

The mid-term re-audit at month six compared the live task mix against the redesigned targets in the Job Redesign Workbook. Hours reallocated per role had moved materially for all four redesigned roles; skills-assessment closure had reached full-cohort completion; revenue-per-FTE showed its first positive quarterly delta against baseline. Two surfaces required amendment: the Account Health Monitor's churn-risk thresholds produced false positives on seasonal advertisers and were retrained, and the Research Briefing digest was shortened after moderators reported digest fatigue — an Amplify surface tuned down is still an Amplify surface.

That is what sustainment looks like in practice: not a launch announcement but a re-audit with amendments in it. The workbook was recertified against the re-audit, and the amendments carry dates — which is exactly what a grant assessor or an internal audit wants to see.

What we would do differently

Two things. First, time-logging should have started in the first week alongside the interviews rather than after them; the two-week validation delayed the baseline sign-off by a gate. Second, the Moderator Copilot's human-confirmation workflow should have been designed with the moderation team in the room from the first session — confirmation points designed without the confirmers produce workflows they route around, and we rebuilt two of them in week nine.

Neither lesson changes the method. Both change the calendar — which is where engagements are actually won and lost.

Data governance and the PDPA-first posture

Five surfaces touching CRM, Search Console, reporting feeds, public web sources, and forum data raise the question every Singapore platform asks first: what is the data posture? The engagement recorded a lawful basis for every source before build began, minimised each surface to the fields its function requires, and kept personal data processing on local hardware where the source allowed — so member data never transited a third-party processor for the moderation and community surfaces.

Two design consequences followed. The Prospecting Assistant enriches from public sources and CRM records under legitimate interest, and stores scored signals rather than raw correspondence. The Forum Moderator Copilot classifies and routes on local models; only escalated threads reach human moderators with full context, which is where judgement belongs anyway. PDPA compliance was not a review stage at the end of the build — it was a constraint inside each surface's design, which is the only place it can be cheap.

If your commercial team is spending more than half its time on tasks that should be automated, email hello@qlabshq.com.