AI Automation Without Losing Judgment: A Playbook

Most teams don’t have a data problem. They have a judgment problem.

That’s the big idea behind the recent e27 story about Kevin Lee (ex-PayPal risk leader) and TrustPlus AI: credit underwriters weren’t failing because they lacked information, but because they were drowning in preparation work—spending 16+ hours per deal gathering data, spreading financials, checking adverse media, and writing memos. By the time they reached the decision, they were depleted.

If you run a Singapore SME, this should feel familiar. Not because you’re underwriting credit, but because digital marketing has the same hidden flaw: teams spend most of their energy preparing dashboards, exporting reports, resizing creatives, compiling weekly updates, and responding to “Can you pull the numbers?” Instead of doing the work that actually changes outcomes—positioning, offers, targeting, experiments, and customer understanding.

This post is part of the “AI dalam Insurans dan Pengurusan Risiko” series, and the underwriting case is a useful lens: it shows how to apply AI in risk workflows without accelerating bad decisions. Then we’ll translate that into a practical, SME-friendly marketing playbook.

The hidden flaw AI exposes: the “judgment deficit”

The flaw is simple: when experts spend most of their time on prep, the quality of their judgment drops—no matter how good their tools are.

In the e27 article, Lee describes a workflow where highly trained underwriters spend only ~20% of their time on true judgment. The rest is clerical work: copying numbers from PDFs to spreadsheets, validating inconsistencies, scanning news, formatting credit memos, and building internal documentation for approvals.

Here’s the sentence that matters for any industry:

Speed doesn’t create risk. Poorly supported judgment creates risk.

In insurance and risk management, this shows up everywhere:

• Underwriting: too much time formatting submissions, not enough time understanding the real risk drivers
• Claims management: too much time triaging documents, not enough time detecting fraud patterns and policy intent
• Fraud detection: too many alerts, not enough prioritisation and investigative capacity
• Risk assessment: reports everywhere, decisions nowhere

AI, used badly, can make this worse—by producing faster reports that no one trusts, or by pushing teams to approve/decline faster without accountability.

Used well, AI does something more valuable: it protects human judgment by removing the prep bottleneck.

What TrustPlus AI gets right (and why finance teams trust it)

TrustPlus’s approach isn’t “AI replaces underwriters.” It’s “AI restores underwriters.”

Based on the article, the platform automates the preparation phase and compresses the workflow from 16+ hours to under 2, covering:

• Financial spreading across languages and country formats (reported 95%+ accuracy)
• Adverse media review across 100+ sources, summarised with citations
• Industry and business model analysis
• Credit memo drafting tailored to each institution’s process
• Audit trails for every source, output, and human override

The most important design choice is structural:

AI prepares. Humans approve.

That sounds obvious, but it’s where most “AI adoption” breaks down—especially in regulated environments like banking, insurance, and risk.

TrustPlus also leans into process trust before outcome trust, which is a good rule for any AI workflow that affects revenue, compliance, or reputation:

• Complete audit trails: you can trace what the AI used
• Explainability: citations, page references, and sources
• Human authority: humans make the final call
• Security posture: bank-grade expectations (SOC 2, GDPR/CCPA mentioned)

This is relevant to AI dalam insurans dan pengurusan risiko because it shows what “responsible automation” looks like in practice: not flashy model talk, but workflows, controls, and accountability.

The SME marketing parallel: your team is doing “credit memo work”

Marketing has its own version of manual spreading and memo formatting. It just looks different.

Common “prep traps” I see in SMEs:

• Reporting that takes half a day (exports, pivot tables, screenshots, slide decks)
• Campaign builds that are mostly repetitive setup (UTMs, naming conventions, resizing)
• Content production bottlenecks (turning one idea into 8 formats manually)
• Lead follow-up delays (slow routing, inconsistent qualification)
• Data scattered across tools (Meta, Google, HubSpot, Shopify, WhatsApp, spreadsheets)

The result is the same as underwriting:

• The team spends more time explaining performance than improving it
• Decisions are made when people are tired (or rushed)
• High-impact work—positioning, messaging, offers, segmentation—gets postponed

If your marketing feels “busy but not effective,” you’re likely facing a judgment deficit.

A practical playbook: automate prep, protect judgment

The goal isn’t to automate marketing decisions. It’s to automate the steps that steal time from good decisions.

1) Automate collection and normalization first (your “spreading”)

Answer first: If your data isn’t consistent, AI will produce confident nonsense.

Underwriters had to normalize financials across formats. SMEs have to normalize marketing and sales data across platforms.

What to implement:

• Standardize tracking: UTMs, lead source fields, campaign naming conventions
• Single source of truth: a CRM (even a lightweight one) that records lead source, pipeline stage, and outcome
• Basic event tracking: form submits, WhatsApp clicks, bookings, purchases

What AI can do safely here:

• Auto-tag and categorize incoming leads by intent and source
• Summarize daily/weekly performance from connected platforms
• Flag tracking anomalies (sudden drop in conversions, broken landing page, spend spike)

Judgment stays human: “Should we change the offer?” “Is this audience wrong?” “Are we attracting bad-fit leads?”

2) Build “process trust” into your AI marketing workflows

Answer first: If you can’t explain why something happened, you can’t scale it.

TrustPlus wins because it makes outputs auditable. SMEs need the same idea, minus the heavy compliance.

Borrow these controls:

• Keep a record of prompts, creative versions, and hypotheses
• Require citations for AI-written market/competitor claims (no citations, no publishing)
• Establish approval rules: AI drafts, humans approve
• Maintain a change log for major campaign shifts (targeting, budget, landing page)

A simple policy that works:

• AI can write drafts and summaries. Humans own claims and decisions.

3) Use AI to compress “time-to-insight,” not to increase posting volume

Answer first: Posting more isn’t a strategy; learning faster is.

A lot of SMEs adopt AI and end up producing more content that converts the same (or worse). The better use is to shorten the loop between:

Signal → interpretation → action

Examples:

• AI summarizes call transcripts and WhatsApp chats into objections, motivations, and common questions
• AI clusters leads by reasons they didn’t buy (price, trust, timing, features)
• AI drafts 3 new offer angles based on real objections—then you test them

This is the marketing equivalent of giving an underwriter time to focus on the “top 20% cases that drive 80% risk.”

In SME marketing terms: focus on the few experiments that drive most revenue.

4) Redesign roles: stop paying “chef wages” for chopping vegetables

Answer first: If a skilled person spends their week on admin, you’re buying the wrong output.

The e27 article compares AI to a sous-chef. That’s spot on.

For SMEs, this often means:

• Let AI create first drafts of ad copy, landing page sections, and email sequences
• Use templates and automation for weekly reporting, not bespoke decks
• Auto-generate meeting notes and action items, then assign owners

Then redeploy human time into:

• Offer design (bundles, guarantees, pricing tests)
• Creative direction (real customer stories, founder POV, proof)
• Conversion rate optimization (landing pages, forms, WhatsApp scripts)
• Sales enablement (objection handling, case studies)

Where AI goes wrong: “accelerating bad judgment”

Answer first: The fastest way to waste ad budget is to scale the wrong message efficiently.

Kevin Lee’s warning is the right one:

The real danger isn’t AI replacing judgment. It’s AI accelerating bad judgment without accountability.

In SME marketing, this happens when:

• AI writes claims you can’t substantiate (trust damage)
• You optimize for cheap leads instead of qualified leads (pipeline pollution)
• You automate follow-ups that feel robotic (brand erosion)
• You let tools decide winners based on too little data (false confidence)

A simple safeguard:

• Don’t scale until you have proof: conversion rate, close rate, refund rate, or repeat purchase—pick the metric that actually matters to your business model.

“People also ask” (quick answers for SME owners)

Should SMEs use AI for marketing if they have limited data?

Yes—if you use AI to reduce prep time, not to “predict customers.” Start with summarization, categorization, and workflow automation.

What’s the first AI workflow to implement?

Automate reporting and insight capture: weekly performance summaries, lead-source tracking, and call/chat summarization into objections and FAQs.

How do you keep AI outputs trustworthy?

Use “process trust”: require sources for external claims, keep version history, and make humans approve final messaging and budget changes.

What this means for AI dalam insurans dan pengurusan risiko

AI in underwriting, claims, and risk management is heading toward a clear direction: workflow-first automation that compresses preparation time while keeping human accountability intact. That’s the real lesson from TrustPlus.

And it’s directly applicable to Singapore SMEs trying to grow through digital marketing. When you treat AI as the engine for admin work—data prep, summarization, documentation—you recover the scarce resource that actually improves performance: clear thinking.

If you’re planning your 2026 growth targets, here’s a strong stance: don’t chase more dashboards or more content. Chase more judgment per hour. Your marketing will get sharper, and your team will stop operating in permanent catch-up mode.

What would your results look like if your team reclaimed even 10 hours a week from reporting and rework—and reinvested it into testing offers and improving conversion?