ShopSCAPE Studio
// ShopSCAPE Studio
Marketing tools with AI that does the work — and still asks before it acts
ShopSCAPE Studio puts everything a marketing and customer-service team does each day in one place — ad banners, design, campaign pages, lead capture, every customer conversation across every channel, and cookie consent that stands up under Thailand’s Personal Data Protection Act.
And the AI in it doesn’t stop at suggestions. It edits real records in the system, but it shows you the change before it makes it, waits for your approval, and can undo it afterwards.
Seven toolsets, one account · AI included and ready — nothing to connect · Built for how commerce actually runs in Thailand · Running a national marketplace today
- แบนเนอร์โฆษณา
- ทดสอบ A/B
- สตูดิโอออกแบบ
- คลังไฟล์งาน
- หน้าแคมเปญ
- คิวอนุมัติงาน
- ปฏิทินแคมเปญ
- บอร์ดงานออกแบบ
- ฟอร์มเก็บข้อมูล
- บอร์ดงานขาย
- โปรไฟล์ลูกค้า 360 องศา
- กล่องข้อความรวม
- เคสและ SLA
- AI ตอบลูกค้า
- ผู้ช่วยร่างข้อความ
- คุกกี้คอนเซนต์
- สร้างภาพด้วย AI
- รายงานจากงานจริง
- นำเข้าและส่งออกข้อมูล
- Ad banners
- A/B testing
- Creative studio
- Asset library
- Landing pages
- Approvals queue
- Campaign calendar
- Design board
- Forms as a database
- Sales board
- 360° customer profile
- Unified inbox
- Cases and SLA
- AI that answers
- Drafting assistant
- Cookie consent
- Image generation
- Workload reporting
- Import and export
Your marketing team is three people. This is how it works like ten.
Everything a marketing and customer-service team touches in a day, on one set of customer records — with AI that works inside the system rather than telling you what to go and do somewhere else.
// Why small teams stall
The problem isn’t effort. It’s seven tools that don’t talk to each other.
Businesses that grow faster than their headcount all hit the same four walls.
-
Simple work always waits on someone else
Changing a price on a banner waits for the designer. Editing a campaign page waits for whoever builds the website. Adding a tracking tag waits for someone who left last year. Work that takes five minutes to decide takes five days to ship.
-
Seven subscriptions, and the customer data is still scattered
One vendor for design, another for scheduling, another for chat, another for forms. None of them knows what this customer already told you.
-
The work lives in a group chat and a spreadsheet
Who’s got it, who approved it, where the current file is, whether anyone called back the lead who filled in yesterday’s form — the answers are in a thread that has already scrolled away.
-
You want AI, but not enough to let it loose
Most AI tools offer two settings: ask it something and copy the answer somewhere else yourself, or let it run and find out afterwards what it did. Neither is something a real business can use.
// What it is
The whole marketing pipeline, on one set of customer records.
Studio didn’t start as a feature list. It started because we had to run a national marketplace with a team that never grew in proportion to the work. Every tool in it was built the week we needed it. It became a product afterwards.
The shape is a single pipeline: make the work people see → turn interest into customers → keep those customers coming back. All three stages write to the same customer records, and AI works across all three.
| Stage | Tools in this stage | The question it answers |
|---|---|---|
| 1 · Make the work people see | Media & Interactive (ad banners) · Creative Studio (design and asset library) · Landing pages · Marketing workspace (plan, approve, track) | How fast can we ship without waiting on anyone? |
| 2 · Turn interest into customers | Forms → database → lead pipeline → 360° customer profile | Yesterday’s enquiries — who owns them, and how far have they got? |
| 3 · Keep them coming back | Unified inbox across every channel · cases and SLA · AI that answers customers · cookie consent | This customer just messaged. What have we already told them? |
The rule we don’t break: anything AI changes in the system goes to a person for approval first, and has to be reversible. If the system can’t take a snapshot it could roll back to, it refuses to make the change at all. No exceptions. That is the reason a real business can let AI touch real data here.
// Stage one
Ship it yourself, in the time you used to spend queuing.
Banners you can change forever, without touching the publisher’s site
Anyone who has bought ad space on a partner site knows the problem isn’t making the banner. It’s that every promotion change means a new file and a slot in someone else’s development queue.
-
Embed the iframe once, edit the content forever
The publisher drops in one snippet and is done. After that you change prices, swap images, or pull the banner yourself from the back office, and their site updates within seconds.
-
Build once, ship every size — and adjust each size for real
Presets at 970×250, 300×250 and 320×100, or anything custom from 20 to 4000 pixels. Layer positions, type sizes and background images are set per size — this is not one image scaled down.
-
Four formats in one tool
Product carousel, image slideshow, free-layer interactive, and a hybrid with a design area and a product strip side by side.
-
Several designs inside one banner
Hold up to six designs in a single unit and rotate them by weighting, while the publisher’s embed code never changes.
A/B testing that doesn’t flatter itself
Most A/B tools declare a winner too early, because they don’t guard against the three things that distort the numbers. We guard against all three.
| Count correctly from the start | One session is one impression, and one click maximum — so numbers can’t be inflated from a public endpoint. |
|---|---|
| No winner before there’s enough data | Decided at 95% confidence, and withheld until every variant clears the data threshold — 1,000 impressions per variant by default. |
| Flag a skewed split | If the audience isn’t dividing the way you set it, the system says so, instead of letting you draw conclusions from a tilted sample. |
| Measure revenue, not just clicks | The system appends a tracking token to the destination link. You place one line on your thank-you page, and the report tells you which variant actually sold. |
| Promote the winner in one click, and roll it back | No hand-editing the banner to match the result. |
Scope you should know first: eleven elements can be tested — headline copy and colour, button copy and colour, background image, background type, countdown on or off, and others. It cannot yet test a different product set or a different layout, and it can’t run at the same time as design rotation. You pick one or the other.
Design in the browser, and an asset library that refuses to delete what’s in use
-
One project, every size
Thirteen ready sizes — Pantip MALL placements, Facebook feed, story and cover — designed once, and each size can be repositioned on its own rather than auto-resized and left alone.
-
Twelve Facebook album formats, including posts where one image runs across several
Lay out one large canvas and the system slices it into ordered images — normally a Photoshop-and-crop job.
-
Share a link so a freelancer can edit, without giving them an account
Lock the layers they mustn’t touch. They add their own work and export it, but can’t save over yours, can’t see anything else, and the link can be revoked at any time.
-
The library warns you before a delete breaks something
Before removing a file, the system checks seven places it might be in use — a banner, a form, a chat page, a campaign page — and shows you the list. It prevents the classic incident where deleting an image silently breaks a live ad.
A campaign page you can finish in an afternoon
| 74 ready blocks across 11 section types | Headers, thirteen hero styles, ten content layouts, ten countdown styles, fifteen tab styles, forms, footers — pick a style and type over the copy on the live page. |
|---|---|
| What you edit is the page itself | The editor renders with the same engine as the published page. Nothing changes when you hit publish. |
| Publish and get a link immediately | Point ads at it or send it in a chat. No hosting to buy, no deployment to wait for. Set the share image and text for Facebook and LINE yourself. |
| Reporting is built in | Visits, unique visitors, click rate, conversions, average time on page, scroll depth, and a four-step funnel — collected without identifying anyone, so it sits inside PDPA by design and needs no external analytics tool to be useful. |
| Add your own GA4 or GTM ID | Enter the measurement ID; the system writes the script. You never paste code. |
| Countdowns that mean Thai time | Set 18:00 Bangkok time and it stays 18:00 Bangkok time, no matter where the person editing is sitting. |
Work as a team without chasing it through chat
-
An approval workflow the system enforces
Idea → draft → in review → approved → published. Steps can’t be skipped, authors can’t approve their own work, and sending something back requires a reason, which is kept as a comment on the item.
-
One approvals queue
A manager opens one screen and sees everything waiting across every brand, with the artwork and the per-channel copy attached. Approve or send back from there — or from the notification bell on any page in the system.
-
Write once, tune per channel
Facebook, Instagram, LINE, X and YouTube each get their own copy and images, with live character counts that warn you before you run long.
-
Plan ahead on the calendar, then publish when it’s time
Three calendar views and a cross-brand campaign timeline on a single axis — so you can see which week is overloaded and who is carrying it.
-
A drag-and-drop design board
Brief graphic and video work as cards, let the design team move them across columns, define up to twelve stages of your own, and open a job in the design tool with its name and sizes already set.
-
Workload reporting from actual events
Who moved what, how many jobs closed, which stage holds work longest, average cycle time — computed from a real event log, not from timesheets people fill in.
Two limits worth stating plainly: there is no post scheduler in the system — you plan ahead on the calendar and publish when it’s time. And publishing from the system reaches Facebook and Instagram once an administrator has connected the page and Meta has approved the app. LINE, X and YouTube are used for planning and copywriting, which you then post at the destination.
// Stage two
From the form to the closed deal, on one pipeline.
Most businesses collect leads with a free form, chase them in a spreadsheet, and keep the customer history in the head of whoever spoke to them. Three separate things, and none of them can tell you where yesterday’s enquiry stands today.
Forms that are already a database
| Type the questions; the table builds itself | Nothing to set up in advance. And if you delete a question later, the answers people already submitted stay intact — they don’t disappear with the field. |
|---|---|
| Eleven question types, thirteen file formats | Enough for application forms, surveys, and issue reports, with supporting documents collected in the same form. |
| Thirteen-digit national ID numbers keep their leading zero | The classic failure of spreadsheets and free form tools. Here the value is stored exactly as entered, and checked for length. |
| Respondents can fix their own answer for 30 minutes | Someone who mistyped a phone number corrects it from the thank-you page — and the corrected contact details flow through to the deal already sitting with the sales team. That is work your team no longer does by hand. |
| Styled to the standard of a campaign page | Eight Thai typefaces, theme colours, and a header banner pulled from your own design work. A form that looks like the brand gets filled in more often than a plain white one. |
| Caps, closing times, and duplicate handling | Nobody has to sit up to close a form at midnight. |
| PDPA consent captured in the form | A consent checkbox with wording you control, recorded with a timestamp and policy version, and attached to the customer profile automatically. |
A sales board a team of two can actually run
-
Drag cards between columns
Every deal is a card with its value, confidence, and owner. The board header totals open deals and pipeline value.
-
Start from one of four templates
Merchant onboarding, campaign recruitment, upsell, new business — then rename the stages to whatever your team actually says.
-
Choose whether leads arrive automatically or after screening
Turn on the switch and every form submission becomes a deal, or tick and send up to 500 at a time. This handles the real case: a thousand applied, you’re only chasing two hundred.
-
Revenue reporting that back-edits can’t distort
Deal value is captured at the moment of closing, so correcting a number later doesn’t change last month’s revenue. Pipeline value, win rate, average cycle time, and per-person performance are all available.
-
Deal-level permissions
Everyone can advance stages and log contact. Closing a deal, changing its owner, or editing its value is limited to the deal owner or a manager — which is what makes the reporting trustworthy.
-
A 360° customer profile
One screen: every channel they’ve reached you on, every case, open deals, a timeline, and their PDPA consent status. If one person exists twice because they messaged on two channels, merge the profiles.
Bring old data in, and take your data out. Import from CSV or a pasted table, with Buddhist Era dates converted automatically. Export to a file that opens correctly in Excel with Thai text intact.
Selling on Pantip MALL already? Two more things.
-
A product row priced live, on every page load
Paste the product link and the system pulls the image, name, price, original price, and discount into a card. The price a customer sees is the price at that second — no incorrect prices on a live campaign, and no going back to edit the page when a promotion changes.
-
Turn a review video you already have into a shop
Drop in a YouTube or Facebook video and attach products to moments in the clip. When the video reaches a product, that product’s card rises over it, ready to buy. The customer never leaves the video to go looking.
These two features connect to products listed on Pantip MALL. If you sell elsewhere, everything else on this page works as normal.
// Stage three
Every conversation in one inbox. Every case with an owner.
The unified inbox
-
Five channels, one screen
LINE Official Account (several accounts), Facebook Messenger (several pages), email, the web chat widget, and a standalone chat page with its own link.
-
You have to claim a case before you can reply
The compose box stays locked until you do — which ends two agents answering the same customer at once, and means there is always a name against every case. A supervisor can take a case over, but the system records it every time. Nobody replies without a trace.
-
Recall a message you just sent
Wrong person, wrong wording — recall it inside the window you set, five seconds by default. For LINE and Facebook the message isn’t actually sent during that countdown, so a recall in time means the customer never saw it.
-
An SLA clock that knows Thai public holidays
Set business hours per day, and the clock pauses outside them and on the national holidays already built in. Your SLA figures stop being punished for hours the team wasn’t working.
-
Closing a case produces the data
Pick a category, a resolution, and a cause at close, and it flows straight into reporting. Nobody fills in a separate summary form afterwards.
-
The conversation history stays with the customer
Next time they message, the team can see everything. The customer doesn’t repeat themselves.
AI that answers customers from your own knowledge base
The number one risk in switching on a customer-facing bot is that it makes something up. So ours answers only from the knowledge base your team writes. If the answer isn’t there, it doesn’t guess — it says it will pass the question to a colleague.
| It can’t go live until it has been tested | The “let AI reply to customers” switch stays locked until you have run at least one test in a simulator that reaches no customer. Turning it off, by contrast, works instantly and always. |
|---|---|
| The bot and your people can’t collide | The moment an agent claims a case, the AI goes quiet, and any action that puts a human back in control clears the AI state automatically. Every handover is logged as a note, so you can see which part of a thread was a person and which was the bot. |
| Retrieval built for Thai | Written Thai has no spaces between words, which is why ordinary search misses things. Ours segments Thai properly and weights human-curated tags at twice the value of body text. |
| A per-case “let AI take it from here” button | Finished the hard part but don’t want to release the case? One button and the AI handles the routine follow-ups. When the case closes, it picks up automatically if the customer writes back in the same thread. |
| A drafting assistant beside the compose box, invisible to the customer | Draft the next reply, summarise the case in a few lines, or rework what you’ve typed — shorter, more formal, grammar fixed, or translated. The output is always a draft. A person sends it. |
| If the model call fails, nothing is sent | No fallback answer is invented for the customer. The system stays quiet, records why, and leaves it to a person. |
Scope you should know: the system can recommend who should take a case, based on who is online, on shift, and how much they’re carrying — but a person assigns it. Work is never routed automatically. Cases are not escalated to a manager automatically when an SLA is breached. Satisfaction scores are collected from the web chat widget only.
Cookie consent — the compliance job most teams leave half-done
If your site runs Google Analytics, the Meta pixel, or the TikTok pixel, and there’s no consent banner holding them until the visitor agrees, that’s an exposure sitting open every day.
-
One line of code to install
Paste it into the page or deploy it through Google Tag Manager. A button then checks whether the code is actually live — which solves the classic case of a beautifully configured banner nobody ever installed.
-
Trackers are held until consent is real
Google Analytics, the Meta pixel, the TikTok pixel and Google Tag Manager are emitted in a form the browser cannot execute, and only swapped for the real script once the visitor agrees.
-
All seven Google Consent Mode v2 signals
Google can still model your measurement when a visitor declines — so you aren’t choosing between compliance and marketing data.
-
Accept and reject look the same
We wrote a test into the build that fails if anyone makes the reject button quieter. Thailand’s PDPA requires consent to be freely given; consent that isn’t freely given doesn’t bind. It is also the point auditors catch most often.
-
Consent evidence that isn’t itself personal data
What was accepted, under which policy version, and when — with no raw IP address and no email. Exportable to a spreadsheet with Thai text intact.
-
Change the policy and consent is re-requested automatically
You never have to wonder whether old consent still covers you.
How the cookie scan works: we open your site in a real browser twice — once without accepting, to prove the site is clean before consent, and once after, to build the full cookie declaration. Known cookies are classified from 24 built-in rules. The scan is provided as part of an installation cycle, not as a self-service button on day one. The banner displays in one language at a time; visitors cannot switch it themselves.
// Agentic AI
AI that works inside the system — not AI that tells you what to go and do.
The AI tools businesses use today come in two shapes. In the first, you ask a question, get an answer, and open another screen to act on it yourself. In the second, you let it run and find out afterwards what it changed.
We built a third. The AI reaches real data and real tools, but every time it wants to change something it has to show you the change first — this value, becoming that value. You approve it, then it acts. And there is always a way back.
The AI is included and ready — you connect nothing
-
No AI account to open, no card to add, no key to paste
Most AI-branded tools ask you to open an account with a model provider and bring the key back. That is where most people give up. Here you sign in and the AI works. That part is ours to run.
-
We match the model to the job
Answering customers, drafting copy, and editing records are different problems that suit different models. We pair them up, and move to a better model when one arrives, without you doing anything.
-
The system isn’t tied to one provider
It supports Anthropic (Claude), OpenAI and Google Gemini — 23 models in total — switchable from a settings screen with immediate effect and no redeployment. That architecture is what keeps us from being locked in, and keeps that risk off you.
-
Usage ceilings are built in
Commands per minute per user, steps per command, and concurrent jobs are all capped. Nothing runs away with itself.
Three rules we don’t break
| Propose, approve, then act — and always reversible | Every data-changing command becomes a proposal showing the old value and the new one. If the system can’t take a snapshot it could restore from, it refuses to make the change at all — so “already changed, can’t undo” is structurally impossible. |
|---|---|
| The AI can never exceed the permissions of the person directing it | Permissions are re-checked every time it calls a tool, and it sees only the workspaces that person actually belongs to. Hand it to a new joiner and it works within a new joiner’s boundaries. |
| Everything it does is auditable | Every tool call, every approval, and every undo is recorded. Who asked for what, when, and what happened. |
Four things the AI does
| An assistant inside the screen you’re working in | A chat panel on the page you’re already on. Ask in plain language: check which products are missing images, update this batch of prices from the file, rename the Songkran images and move them into a folder. Work that used to be clicked one row at a time finishes in a sentence. |
|---|---|
| The customer-facing bot | Answers from your team’s knowledge base, on LINE, Facebook, email and the web. |
| The drafting assistant | Drafts replies, summarises cases, adjusts tone — invisible to the customer. |
| Image generation and a content production canvas | Generate images from a description with up to six reference images, landing straight in the shared library — plus a drag-and-drop canvas for chaining several AI steps together, such as source material, then copywriter, then image generator. |
Where we think differently
-
The AI calls the system’s own tools, it doesn’t click the screen for you
25 tools across four working screens, wired directly to real data — more accurate than screen automation, and it doesn’t break when the interface changes.
-
Images are generated deliberately without text, so a person sets the Thai type
Every image model still mangles Thai script. So ours generates a clean image and hands the words over as an editable text layer in the design tool. You get the speed of AI without losing correct Thai typography.
-
Prompt injection is handled
Customer messages and tool output are wrapped and labelled as data, not instructions. A customer typing “ignore all previous instructions” into a chat gets nowhere.
-
Bounded by the operator, every call
There is no separate service account the AI runs as. It borrows the permissions of whoever is directing it, and those are re-checked on every tool call rather than once at sign-in.
⟦PLACEHOLDER — live web search. The brief claims the assistant can search the web and show which pages it opened. That capability is enabled at the model-provider account level, which is now our account and therefore our decision, and nobody has confirmed it is switched on. The claim is held rather than hedged.⟧
⟦PLACEHOLDER — the “nothing to connect” promise above rests on four things being operationally true at delivery, not merely intended: that onboarding configures a model key for every customer, that at least one of them is an Anthropic key (the in-page assistant that edits records runs on nothing else), that a fair-use position exists now that we carry the cost, and the web-search decision above. Until those are confirmed, this page promises something the system may not do for a given customer — which is worse than not advertising it.⟧
// Proof
We run on it ourselves, every day.
Studio isn’t something we built and then waited for customers to try. Our own team runs the marketing and customer service of a national marketplace on it every day. We find the bugs before customers do.
Still being built
The product ships continuously rather than in annual releases, and the team that answers your questions is the team writing the code.
⟦PLACEHOLDER — the release number and update count. The brief offers both as a “this is real and still moving” signal. They are a strong signal precisely because they are specific, which is also why nobody should publish them until someone owns keeping them current.⟧
⟦PLACEHOLDER — customer card. Awaiting written consent (G4): a two-line description and one metric the data owner has reviewed. How far we may name the marketplace we run is part of the same question.⟧
We only show work we’re allowed to talk about. If a number hasn’t been confirmed by whoever owns the data, it doesn’t appear here.
// Standards
Assessed by organisations that don’t work for us.
Every vendor says its system is secure. The question that’s actually useful during evaluation is who assessed it, and against whose criteria.
The national digital service catalogue
| What it is | Thailand’s register of digital products and services, maintained by the Ministry of Digital Economy and Society together with the Digital Economy Promotion Agency. |
|---|---|
| Why it matters | Public bodies can procure from this register through a transparent, auditable process — and for any buyer, being listed means having passed a screening someone else administered. |
Business Development Service — Office of SMEs Promotion
| What it is | The Business Development Service measure run by Thailand’s Office of Small and Medium Enterprises Promotion, under which the state co-funds business development services for smaller businesses. |
|---|---|
| How we’re involved | We’ve joined the programme. Being an approved provider within it means passing review by the agency responsible for the country’s SME sector. |
| What it means for your business | Businesses that meet the measure’s criteria can apply for support toward services covering sales-channel development and digital adoption. Our team can help work out which category your case falls under. |
⟦PLACEHOLDER — two things held here. The digital product quality mark is omitted entirely rather than hedged, because nobody has confirmed whether it was issued for the ShopSCAPE platform alone or covers Studio as well. And the English names of the three public bodies above need checking against their own published usage before an English page carries them.⟧
A certificate doesn’t make a system better on its own. What it does is let the security and PDPA conversation start from existing documentation instead of from zero.
// Getting started
We set it up. You start working.
-
Our team configures it during onboarding
LINE Official Account, Facebook pages, email, business hours and holidays, and the AI — ready before you log in. There is nobody you need to open an account with first.
-
A checklist that ticks itself as you actually use it
On first sign-in the system asks what you came here to do, then builds a starting checklist to match. Each item ticks itself when you genuinely do it, rather than waiting for you to tick it — so you always know what’s left.
-
Turn on only what your team needs
Most teams start with two areas and open the rest when they’re ready.
-
Your data stays yours
Export customers, leads and reports to CSV that opens correctly in Excel with Thai text intact. Nothing is held hostage.
⟦PLACEHOLDER — the handover. Product detail, packages and trials belong on the official ShopSCAPE Studio site, and this block is where the reader is handed to it. That URL is unconfirmed, so the only route out of this page is our team.⟧
// Why ShopSCAPE Studio
Built where the work actually happens.
We build on Google Cloud as a certified implementation partner. Multi-cloud through one team is the point: BytePlus and Alibaba Cloud where the workload calls for it.
-
We built it because we needed it
Every tool here exists because we had to run a national marketplace with a team that never grew in proportion to the work. None of it came from a competitor’s feature list.
-
Built for how Thailand actually works
Retrieval that segments Thai text correctly, so a search finds the answer sitting right there. Thai public holidays inside the SLA clock, so your service figures aren’t punished for days nobody worked. Buddhist Era dates converted on import, so a decade of spreadsheets loads cleanly. Sixteen Thai typefaces embedded automatically. Exports that open in Excel without turning Thai into symbols. None of this is what foreign tools optimise for, and all of it is what costs a Thai team time every day.
-
AI that’s ready, and safe enough to point at real data
Because we supply and run it, and because every change goes through a person and can be undone.
-
The person who fixes it is the person who wrote it
Not a ticket in a foreign support queue in another timezone.
-
Assessed by Thai public bodies
Both for the digital service register and as an approved provider in a programme the state administers.
-
Growing out of it doesn’t mean leaving
When the business needs a full commerce platform, a multi-vendor marketplace, or order management, ShopSCAPE sits in the same architecture. You extend into it. You don’t start again.
// How this compares
Against the three things you are probably doing now.
| Five to seven foreign tools | A retained agency | Spreadsheets and group chat | ShopSCAPE Studio | |
|---|---|---|---|---|
| Customer data | Split across vendors | Held by the agency | In a file anyone can edit | In one place, and the service and sales teams see the same records |
| Speed of a change | Fast, if you’re fluent in all of them | Queued | Fast, but leaves no trace | Fast, with a record of who changed what |
| Thai | Usually a translated interface and little else | Not applicable | — | Down to word segmentation and the public holiday calendar |
| PDPA | Another subscription | Usually not included | Absent | In the system — consent banner and consent evidence both |
| AI | Each vendor has its own, each needs configuring | Rarely transparent | None | Included, ready, and able to work inside the system |
| Who fixes it | Offshore support | The agency | You | The team that wrote the code |
We’ll tell you plainly if your problem is still small enough for the free tools to handle. Selling someone more system than they need costs us as much time as it costs them money.
// Questions
What teams ask before they start.
How big does a business need to be for this to fit?
A better test than size is how many tools you’re already running. If design is one vendor, chat is another, forms are a third, and cookie consent is a fourth — and you’re still copying data between them by hand — that’s the point worth a conversation.
Do we have to use ShopSCAPE as our commerce platform too?
No. Studio works whatever you sell on. Two features are the exception, both tied to products listed on Pantip MALL: the live-priced product row on campaign pages, and product cards on video.
Do we need our own AI account?
No. We supply and manage that. You sign in and the assistants and the customer-facing bot work — no model provider to sign up with, no card to add, and no quota to watch.
Will the bot make things up to our customers?
It answers only from the knowledge base your team writes. If the answer isn’t there, it says it will pass the question on rather than guessing. And if the model call fails, it sends nothing at all instead of inventing a fallback. On top of that, the go-live switch stays locked until you’ve tested it at least once.
Can the AI break something?
Every change is a proposal you approve first, showing the old value and the new one, with an undo. If the system can’t take a snapshot it could restore from, it won’t make the change at all. And it can never exceed the permissions of the person who asked.
Can we publish to social media from the system?
You can plan ahead on the calendar, write per-channel copy, and route it for approval. Publishing from the system reaches Facebook and Instagram once an administrator has connected the page. LINE, X and YouTube are used for planning and copy, which you then post at the destination.
Can freelancers or an agency work in it without seeing our customer data?
Yes. There’s a permission level that shows external users only the campaigns assigned to them, and a design share link that lets someone edit without an account at all, with the layers you don’t want touched locked first.
How secure is our data?
Provider keys and passwords are encrypted before storage and are never sent back to the browser. Each team sees only its own data, enforced on the server rather than by hiding menus. Back-office pages are set to be excluded from search indexing by default. Every significant change is recorded and auditable.
⟦PLACEHOLDER — data residency, log retention, single sign-on, restore capability and penetration testing. These are the questions an enterprise security questionnaire opens with, and none of them is answered in the source material. They matter more on the English page than the Thai one.⟧
Can we bring data over from spreadsheets?
Yes. Drop in a CSV; the system reads the headers, infers each column’s type, converts Buddhist Era dates, and shows you a preview before anything is created.
If we leave, can we take our data?
Yes. Customers, leads and reports export to CSV that opens correctly in Excel with Thai text intact.
Does this work for a larger organisation?
Yes. There are three layers of permissions, brand-separated workspaces enforced server-side, an audit log that can’t be edited after the fact, and two-factor authentication. Organisations running several brands, or working alongside an agency, are the case this was designed around.
What does it cost?
Package detail lives on the product’s own site. Until that address is published here, talk to our team about which fits your case.
Tell us about the job your team repeats every week.
Standard demos aren’t much use. Tell us which piece of work your team does every week that still waits on somebody else, and we’ll show you how many minutes it takes here.
| Where most teams go next | Why |
|---|---|
| ShopSCAPE | The business has reached the point of needing a full commerce platform, a marketplace, or order management |
| Agentic AI & Generative AI | You want AI working beyond marketing and customer service |
| Commerce Platform | You want to know how a real build project runs if we do it for you |
Innovate for the better tomorrow.
// Corporate update
Our
Move
-
Buying the platform and building on it used to be two conversations with two suppliers. It is one conversation now: we resell Google Cloud, Alibaba Cloud and BytePlus, and the same engineers who size the environment stay with it through production support.
For teams already running with us, nothing changes technically — the difference is commercial. Licensing, quota and billing sit with the people who know what the workload actually does.
-
A certified implementation partner works to the cloud provider's published reference architectures. In practice that means your landing zone, IAM model and network layout look like something any Google Cloud engineer can pick up — including the next team you hire.
It also means the review gates are not ours to waive. Where the reference architecture asks for separation of duties or a break-glass path, it gets built.
-
Hospital data arrives in fragments — HIS exports, lab feeds, scanned forms, free-text notes in Thai and English. Before a model sees any of it, someone has to answer where each field came from, who consented to what, and which records must never leave the country.
We build that layer first. It is slower to demo and it is the reason the pilots survive contact with a real ward.
-
Live commerce moves fast enough that the recommendation loop has to close in the same session. That puts the weight on the event pipeline, not the model: what counts as a view, when a cart event lands, how quickly the feature store sees it.
We treat the BytePlus components as a stack to be wired properly rather than a switch to be flipped. The lift comes from the wiring.
-
The site you are reading ships as static HTML, one stylesheet and one script, served by a Node process with a strict content security policy. There is no analytics tag, no font CDN and no tracker.
It is partly a statement of taste and partly a working sample: the same restraint we bring to a client's platform, applied to our own front door.
-
Warehouses fill up faster than they get governed. By the time a model needs a feature, nobody can say which of the four revenue columns is authoritative, and the project stalls in a meeting about definitions.
The fix is unglamorous: contracts on the ingest side, lineage through the transformations, and one owner per domain. Do that and the AI work stops being archaeology.