
This case study is based on a local Indonesian skincare brand based in Bandung with steady monthly revenue of Rp 80 million at the time of onboarding. The brand name has been anonymized at the owner's request; all figures and timelines are real and drawn from the February–May 2026 period.
The real problem — not demand, but capacity
The owner was already running Instagram Ads consistently at Rp 30 million per month with a 2.4× ROAS. Demand was clearly there — Instagram DMs were coming in at 80–120 inquiries per day. But the part-time admins could only handle 60–70 chats per day each. The rest piled up, with many messages not being replied to for 6–18 hours.
Indonesian skincare buying behavior: customers dealing with active breakouts or sudden skin irritation need instant consultations. If response time exceeds 3 hours, they order a competing product on the marketplace. Every lead becomes a sunk cost — the ad spend was already paid, but the conversion was lost.
AI setup — what was deployed
- Days 1–3: Audited the top 500 admin conversations (to train brand voice and product knowledge).
- Day 4: Built the knowledge base — active ingredients per SKU, contraindications, ingredient lists, and return policy.
- Day 5: Set up AI consultation triage — helps customers identify their skin type (oily/dry/combination/sensitive) before recommending a SKU.
- Day 6: Connected AI to WhatsApp Business API + Instagram DM auto-reply + comment auto-reply.
- Day 7: Soft launch to 30% of traffic; admins continued monitoring to correct tone.
- Days 8–14: Brand voice tuning; added 40 new FAQs surfaced from real traffic.
- Day 15: Full rollout to 100% inbound.
90-day results — real numbers
| Monthly revenue | Rp 80jt → Rp 245jt (+206%) |
| Instagram DMs replied within 5 minutes | 12% → 96% |
| Average first response time | 4.2 hours → 1.1 minutes |
| Conversion from DM to order | 11% → 22% |
| Average order value | Rp 320K → Rp 385K (automated upselling) |
| Customers who reordered (90 days) | 18% → 34% |
| Spicelab AI Customer Service Pro cost | Rp 1.2jt / month |
| Net ROI at month 3 | Approx. 137× (Rp 1.2jt input vs. Rp 165jt in incremental revenue) |
The number that surprised the owner most: average order value increased by Rp 65K per order. This happened because AI consistently offered logical bundle recommendations (e.g., when a customer DM'd asking about a serum, the AI also suggested a complementary moisturizer). Human admins often skipped this step due to fatigue.
Unexpected outcomes
Things that weren't in the original proposal but emerged as significant downstream effects:
- Customers who had previously only browsed the Instagram comments section started sending DMs with technical questions — because they knew they'd get a reply. DM volume grew 4×, but conversions grew proportionally as well.
- The owner stopped being tied to their phone until 11 PM. They could now focus on product development and new brand campaigns.
- Customer data was automatically collected and organized into Google Sheets — becoming the foundation for WhatsApp broadcast campaigns that weren't previously executable due to an unstructured database.
- The part-time admin team wasn't let go — they were repositioned to handle VIP cases, complaints, and high-ticket follow-ups. Job satisfaction increased because they were no longer answering repetitive chats all day.
Challenges the typical sales pitch won't mention
Also underrated: AI performs well only if your knowledge base is solid. In the first month, the owner had to invest time rewriting FAQs and ingredient lists in a structured format. About 20 hours of owner work upfront — which ultimately became a permanent brand asset.
“I thought I was paying for software. But the biggest change turned out to be our own discipline in documenting what we know. The AI was just the catalyst.”
— Owner, case study skincare brand (April 2026)
Lessons worth extracting
- AI doesn't fix a broken product — this brand already had clear product-market fit. AI simply removed the capacity bottleneck.
- Weekly metric tracking is critical — the owner checked the dashboard every Monday morning and corrected every drift.
- Keep staff on hand for complex cases — 100% full automation is rarely optimal. An 80/20 hybrid model is the most sustainable approach.
- Brand voice training requires serious upfront effort. Skipping this results in generic AI output.
- Calculate ROI in Rupiah, not percentages. Percentage ROI can be misleading when the base is small.
For skincare brands still on the fence
Nearly every brand we audit has the same underlying issue: demand is there, ads are running, but admin capacity is the bottleneck. The differentiator isn't the tools — it's the owner's discipline during onboarding.
If your business is in a similar position (Rp 50–150 million in monthly revenue, consistent ad spend, limited admin capacity), this same pattern can typically be replicated within 60–90 days. Start by auditing your DM volume before committing to any plan.
Pertanyaan yang sering diajukan
Can Spicelab AI handle both Instagram DMs and WhatsApp at the same time, like in this case study?
Yes. Spicelab is an AI customer service platform that replies across both WhatsApp and Instagram for Indonesian businesses. In this case study, the AI was connected to the WhatsApp Business API, Instagram DM, and comment auto-reply — all in one system — so every inbound message was handled 24/7 with a consistent brand voice across both channels.
Which plan did this skincare brand use, and how much did it cost?
This brand used Spicelab's Pro plan at Rp 1,490,000 per month. Spicelab pricing is transparent across three dimensions: tier (Lite Rp 390,000, Pro Rp 1,490,000, Suite Rp 4,900,000 per month), Spark top-up at Rp 500 per reply, and the channels you activate. A free 7-day trial is available to test it first.
Will the AI stop responding if DM volume spikes to four times the normal amount?
No. Spicelab is designed so the AI never stops responding, with layered economy modes to keep the service running during surges. In this case study, DM volume increased fourfold, but customer contacts were never capped — every inquiry was handled without a backlog building up.
Can the AI sound natural and match the personality of a local skincare brand?
Yes. Spicelab uses industry-specific AI brand voice with natural Indonesian. In this case study, brand voice was trained from the top 500 admin conversations, then tuned over three weeks until the tone felt right for a young audience. The owner's discipline in reviewing daily conversations made the output increasingly accurate over time.
Can a 3× revenue result like this be replicated for my brand?
This pattern can typically be replicated within 60–90 days if your brand has clear demand and admin capacity is your bottleneck — for example, Rp 50–150 million in monthly revenue with consistent ad spend. The key is not the tools, but the owner's discipline in building the knowledge base during onboarding.
Is your brand in a similar position? Get a free diagnosis first
Spicelab's AI Business Consultant probes five dimensions of your business (finance, ops, HR, ads, and content), then delivers a PDF report identifying your biggest bottlenecks and estimated ROI benchmarked against similar case studies.


