In this article
Why Start in 2026
2026 is the inflection point for AI sales in Thailand. Three forces have aligned:
1. Pricing has fallen into SME range. Three years ago AI voice agents started at over ten thousand baht per month. Today they begin at ฿2,990/month — comparable to a part-time employee for a couple of hours a day.
2. Thai language is production-ready. LLMs that understand Thai and speak naturally are now in production, not the awkward word-by-word transliteration tools of years ago.
3. Customers expect faster. HubSpot research shows leads contacted within 5 minutes convert 21× higher than those contacted after 1 hour. Humans cannot. AI can.
5 Steps Without IT
Step 1: Identify your biggest bottleneck
Before deploying AI, answer: "Where is our sales bottleneck?" For Thai SMEs, the answer usually falls into one of three categories:
- Too many leads, can't follow up — Facebook/Google Ads generate leads faster than the team can call
- Inbox replies are slow, customers churn — LINE/Messenger volume overwhelms admins
- Appointments no-show too often — no reminder system in place
Pick the biggest, most measurable one. Do not try to solve everything at once.
Step 2: Choose a "no-code" platform
AI sales platforms split into two camps:
- Developer Platforms (Vapi, Retell AI) — per-minute pricing, code required, USD pricing — built for dev teams
- SME Platforms (such as caiku.ai) — flat monthly pricing, no code, live in 24 hours — built for business owners
SMEs without IT teams should choose SME platforms — fewer features, but actually usable from day one.
Step 3: Connect channels (LINE / Facebook / Phone)
Each channel takes around 30-60 minutes to connect — done through Settings, not code:
- LINE OA — connect via Channel Access Token (copied from LINE Developers Console)
- Facebook Page — connect via Facebook OAuth (click "Connect" and approve)
- Phone — use the platform's provided number or port your own
Step 4: Upload your knowledge base
The AI must "know your business" before it can answer customers. Upload:
- Product/service catalogue with pricing
- FAQ — at least 20 common customer questions
- Hours, address, contact channels
- Active promotions and campaigns
- Edge cases that should escalate to a human (complaints, disputes)
Step 5: Test → Go live → Iterate
Before going live, test 20-30 calls/conversations yourself plus your team:
- Call yourself: ask real-customer-style questions. Does AI answer well?
- Test edge cases: angry customer, off-topic questions, things not in FAQ
- Check sales handoff flow: do leads land correctly in your CRM?
After launch, review the dashboard every 3 days in the first month. Adjust the knowledge base as you spot weak responses.
Realistic Starting Budget
Most SME owners worry about cost. Here are real numbers.
Starter — small shops
- Price: ฿2,990/month (around ฿3,500 with VAT)
- Fits: small shop, 1-3 staff, fewer than 200 leads/month
- Comparable to: half a day per month of part-time wages
Growth — mid-sized SMEs
- Price: ฿5,990/month
- Fits: 10-30 staff, 200-1,000 leads/month, multi-channel
- Comparable to: one full-time admin per month
Quick ROI math
Assume a real estate broker with 300 Facebook leads/month converting at 5% = 15 deals/month. With AI replying within 5 minutes, conversion rises to 10% = 30 deals — an extra 15 deals × ฿20,000 commission = +฿300,000/month.
Investment: ฿2,990/month → ROI ~100×.
Calculate ROI for your business →
What to Prepare First
Before subscribing to any platform, prepare these four things.
1. Your current process as a flowchart
Sketch: "Lead arrives → admin does A → if reply B, route to C → close at D." If you cannot draw it, your process is not standardized yet — fix that before adding AI.
2. An updated knowledge base
FAQs, pricing, policies must all be current. AI will repeat whatever you upload, including errors.
3. A CRM or spreadsheet to receive qualified leads
Where will AI send qualified leads? If you do not have a CRM yet, at minimum a Google Sheet your team can access.
4. A team to receive handoffs
AI qualifies — but closing is human work. Who handles the qualified leads? Set responsibility before going live.
5 Mistakes to Avoid
1. Deploying AI before fixing a broken process
AI scales process — including bad ones. If your flow is broken, fix it manually first, then add AI.
2. Expecting AI to be "smart from day one"
The first month, AI will fumble plenty of cases. That is normal. Expect 4-6 weeks of knowledge base tuning before AI hits its stride. Plan for iteration.
3. Using voice AI where customers prefer text
Some businesses (Gen Z e-commerce) have customers who hate phone calls. Use AI for DM/LINE instead. Pick the channel that matches customer behavior.
4. Not testing edge cases
If you only test the happy path, AI will collapse on the first angry customer or curveball question. Test angry, off-topic, complaint, and competitor-comparison scenarios.
5. Not monitoring after launch
AI is not "set and forget." Review conversation logs weekly. Spot weak responses. Update the knowledge base continuously.
Pre-Launch Checklist
Use this before pressing "Go Live."
- ☐ Identified the most urgent use case (e.g., "Follow up Facebook Lead Ads in 5 min")
- ☐ Current process drawn as a flowchart
- ☐ Knowledge base updated (FAQ ≥ 20 items, pricing, policy)
- ☐ All needed channels connected (LINE/Facebook/Phone as applicable)
- ☐ Tested with self + team, ≥ 20 conversations
- ☐ Edge cases tested (angry customer, off-topic, complaint)
- ☐ Responsibility assigned — who handles AI-qualified leads
- ☐ CRM/Sheet ready to receive qualified leads
- ☐ KPIs defined for month one (response time, qualified rate, conversion)
- ☐ Calendar blocked to review the dashboard every 3 days in month one
Ready to begin? Try caiku.ai's 14-day free trial — live in 24 hours, no credit card required. Or book a demo for a guided walk-through of use cases in your industry.