Skip to main content
AI Concept Studio

AI AUTOMATION

Let AI automation handle repetitive handoffs.Keep important decisions with people.

From ad budget checks and quotation drafts to competitor monitoring, invoice matching and warehouse handoffs, ACS builds practical automation around the work your team already does.

Let the system handle repetition. Keep important decisions with people.

INPUT → CHECK → OUTPUT → HANDOFF

01

MARKETING / GROWTH / Ad Budget & Performance Monitoring

All six use cases ↑

Know which campaigns need attention.Without checking every account.

Collect ad data on an agreed schedule, check budget usage, selected metrics and freshness, notify the owner of issues, and prepare a weekly summary with sources.

For

Brand marketing teams, agencies and teams managing multiple ad accounts.

What you receive

Daily attention list, budget alerts with sources and periods, weekly master table and reviewable summary. Creative / hook tags can be added from an approved mapping.

Concept illustration of ad budget checks, alerts and weekly reporting.
Concept illustration · use the HTML example for data and status
Output example · synthetic data

Choose an example scenario

Autumn launch | budget alert

Period
2026-09-01 — 2026-09-10
Approved budget / duration
HK$60,000 / 30 days
Elapsed / actual spend
10 days / HK$25,000
Uniform pacing budget
HK$20,000 = 60,000 ÷ 30 × 10
Above plan
HK$5,000 · 25%
Linear full-period spend
HK$75,000

Above the illustrative 20% threshold. The owner checks budget and pacing. Linear spend extrapolation is not a performance forecast; no ads are changed.

Weekly email draft

Period: 2026-09-01 to 09-10. Synthetic Meta Ads and Google Ads sources available through 09-10 09:00 HKT. Attention: spend is 25% above uniform pacing. Master-table and creative fields appear below. Pending human review · not sent.

Master-table fields
Date, source, campaign, spend, approved budget
Creative / hook mapping
Approved mapping only; no creative classification or conversion results supplied

Google Ads data is stale

Meta Ads last available
2026-09-10 09:00 HKT
Google Ads last successful
2026-09-09 09:00 HKT
Fixed example check
2026-09-10 09:00 HKT
Combined spend / conclusion
Withheld pending source refresh

The owner checks the connection and reporting period. Missing data is neither zero nor evidence that nothing needs attention.

Workflow, starting scope & human checks

From input to handoff

Authorised ad data + approved budget → period and freshness checks → pacing calculation → agreed rules → internal notification → weekly draft.

Suggested starting scope

One Meta Ads account, one Google Ads account, one budget table, three alert rules, one internal channel and one weekly format. GA4, CRM outcomes and creative analysis are additional scope.

Key rules

Client-agreed rules cover pacing, CPA over a specified window, campaigns with no spend and source-update failures. They are not universal platform standards. Account for attribution delays; do not judge performance from an unfinished day or sum platform-attributed conversions as deduplicated orders.

AI, rules & people

Rules calculate and check data; AI drafts from checked records; people confirm causes, budget changes and external conclusions.

Integration & exception boundaries

Confirm account authorisation, available data and periods. Withhold combined conclusions for stale sources. Alerts do not change ads.

02

MARKETING / GROWTH / Enquiries, Quotes & Follow-up

All six use cases ↑

After an enquiry, keep the requirements,quote and next step together.

Collect enquiries from designated email, forms or authorised messaging channels. Extract quantity, budget, dates and missing details. Once complete, draft from approved prices and remind the owner to follow up.

For

B2B services, corporate gifts, wholesale and teams with human quote approval.

What you receive

Enquiry and opportunity table, owners and next steps, approved-price quote drafts, reply drafts for review, and unassigned or follow-up reminders.

Concept illustration of enquiry intake, quote drafts and follow-up.
Concept illustration · use the HTML example for data and status
Output example · synthetic data

Choose an example scenario

Corporate gift enquiry | details missing

Customer request
300 corporate gifts with a logo, about HK$150 each, delivered to Wan Chai before 25 September
Date check
Year and delivery requirements unconfirmed
Missing details
Company, email, exact address, product, logo file and split delivery
Owner
Sales owner

Awaiting details · no formal quotation prepared.

Reply draft · review required · not sent

Thank you for your enquiry. Please confirm the gift selection, delivery year and date, exact address, and any split delivery, and provide your company, contact email and logo file so we can check availability and pricing.

GIFT-A | quotation draft

Example assumptions
Product, quantity of 300 and required details confirmed; plate and shipping fees approved
Goods
300 × HK$128 = HK$38,400
Logo plate fee
HK$600
Shipping
HK$280
Total
HK$39,280

Pending sales approval · not sent. A person confirms stock, print requirements, delivery, terms and customer commitments.

Workflow, starting scope & human checks

From input to handoff

Enquiry → requirements and missing fields → update the same case → approved-price calculation → sales review → internal follow-up.

Suggested starting scope

One entry point, opportunity table, price list, quote template, two reminder rules and one approval step.

Key rules

Remind a supervisor if unassigned within agreed working hours. After a formal quote is sent, check for a reply or next step after client-defined working days. Stop or adjust for won, declined, replied or contact-later cases. Updates to one conversation stay in the same case.

AI, rules & people

AI structures requirements and reply drafts; rules calculate approved prices and track status; sales confirms the product, stock, print, delivery, terms and commitments.

Integration & exception boundaries

WhatsApp and similar channels require suitable business connections, authorisation and messaging rules; private WhatsApp is not read arbitrarily. This page shows internal reminders and drafts only; it sends no quotes or follow-up messages.

03

MARKETING / GROWTH / Competitor Price & Promotion Monitoring

All six use cases ↑

Track price, offer and availability changes.Without opening every page daily.

Monitor specified product pages for price, pack size, offer conditions and page availability. Notify the owner of relevant changes with previous and current records, source and time.

For

Retail, ecommerce, brands and teams tracking market prices.

What you receive

Price history, price and offer-condition alerts, page availability alerts, read-failure records and a concise weekly change summary.

Concept illustration of competitor pack, price and offer monitoring.
Concept illustration · use the HTML example for data and status
Output example · synthetic data

Choose an example scenario

Price change | retailer A

Product
Yuzu sparkling tea 330ml × 6
Previous → current
HK$96 → HK$78
Reduction
HK$18 · 18.75%
Calculation
(96 − 78) ÷ 96 × 100 = 18.75%
Comparison basis
Same volume, pack and currency; public price; page shows purchasable
Fixed example time · not live
2026-09-10 09:00 HKT

Notify the owner to review; no changes to your own prices.

View source evidence

Evidence presentation example. Previous example: 09-09 09:00, HK$96; current example: 09-10 09:00, HK$78. Source location: retailer A’s confirmed product page (no real URL or captured evidence provided).

Retailer B | conditional offer

Pack
330ml × 6
Members only
HK$72

Requires membership; not a public price available to everyone.

Retailer C | not directly comparable

Pack
300ml × 6
Price
HK$68

Volume differs from the monitored product; this is not a same-product price drop.

Retailer A | read failed

Last successful record
HK$78 · 2026-09-10 09:00 HKT
Fixed attempted check
2026-09-10 15:00 HKT
Current price / availability
Unknown; recheck required

Retained data is stale. Read failure does not mean out of stock and is not filled with HK$0.

Workflow, starting scope & human checks

From input to handoff

Confirm product pages → record specifications, price, conditions and time → like-for-like comparison → source-linked alert → weekly summary.

Suggested starting scope

20 confirmed URLs on up to three sites, two checks a day, one history table and a set of price or status rules. A discussion starting point, not a fixed package or availability guarantee.

Key rules

Record product identity, volume, pack count, retailer, currency, public price, membership or offer conditions, minimum quantity, page availability, time and source evidence. Separate differing packs and member offers; retain and mark stale data on failure.

AI, rules & people

Rules compare specifications and prices; AI summarises checked changes; people confirm sources, product mapping and commercial responses.

Integration & exception boundaries

Assess access and site terms first. No guarantee every site can be read reliably and no bypassing login, payment or other restrictions. Page availability is not warehouse inventory; your own prices are not changed automatically.

04

MARKETING / GROWTH / Campaign Assets & Launch Readiness

All six use cases ↑

Images approved, copy still pending.Keep the launch handoff on hold.

Keep images, copy, landing pages, versions and launch information in one campaign record. Identify missing items and notify owners; prepare a clear launch handoff once agreed conditions are met.

For

Brand marketing teams, agencies and teams with multiple reviewers.

What you receive

Deliverable and version register, asset-specific revisions, formal approvals, deadline reminders and a handoff pack of approved assets, copy, links, schedule and ad details.

Concept illustration of campaign versions, approvals and launch readiness.
Concept illustration · use the HTML example for data and status
Output example · synthetic data

Choose an example scenario

Not ready for launch

Scheduled launch
2026-09-25 12:00
Total / complete / pending
7 / 2 / 5
Main KV
V03 · Approved
4:5 image
V03 · Approved
9:16 image
V02 · Latest direction missing
Chinese copy
V02 · Revision needed · offer date conflict
English copy
V01 · Pending review
Landing page
— · Entered, pending checks
Ad budget & dates
— · Entered, pending approval
Offer end in copy
2026-09-20

Chinese copy needs review: launch is after the offer end date. Confirm the offer or scheduling date.

Not ready for launch

Scheduled launch
2026-09-25 12:00
Total / complete / pending
7 / 1 / 6
Main KV
V04 · New version needs approval
4:5 image
V03 · Approved
9:16 image
V02 · Latest direction missing
Chinese copy
V02 · Revision needed · offer date conflict
English copy
V01 · Pending review
Landing page
— · Entered, pending checks
Ad budget & dates
— · Entered, pending approval
Offer end in copy
2026-09-20

Chinese copy needs review: launch is after the offer end date. Confirm the offer or scheduling date.

Ready for scheduling handoff

Scheduled launch
2026-09-25 12:00
Total / complete / pending
7 / 7 / 0
Main KV
V03 · Confirmed
4:5 image
V03 · Confirmed
9:16 image
V02 · Confirmed
Chinese copy
V02 · Confirmed
English copy
V01 · Confirmed
Landing page
— · Confirmed
Ad budget & dates
— · Confirmed
Offer end in copy
2026-09-30

This simulated scenario resolves the date conflict and all pending items. Ready for handoff only; nothing is published or advertised.

Workflow, starting scope & human checks

From input to handoff

Deliverable list → assets and versions → version-linked feedback → formal approval → date and missing-item checks → launch handoff.

Suggested starting scope

One campaign type, asset location, deliverable checklist, approval path, deadline reminder set and standard handoff pack.

Key rules

“Change the headline” is not approval. Record the authorised approver, time and version. New versions do not inherit approval; a final filename is insufficient. Preserve retrievable versions or snapshots. Date conflicts belong to the copy item; ambiguous dates need human review.

AI, rules & people

Rules track versions, counts and dates; AI organises feedback and tasks; authorised people approve specific versions and scheduling.

Integration & exception boundaries

Handoff is ready only after the date conflict and every pending item are resolved. This page does not upload, schedule or run ads.

05

OPERATIONS / Supplier Invoice Matching

All six use cases ↑

An invoice arrives.Match the order and receipt before accounting review.

Collect supplier documents from a designated inbox or folder, structure invoice numbers, dates and lines, and match purchase and receipt records. Flag duplicates, price differences or incomplete receipts for review, not payment.

For

SMEs and operations teams checking purchasing, receiving and supplier payments.

What you receive

Invoice register, original-file and source mapping, duplicate and difference lists, review and approval records, and accounting-system bill drafts if included.

Concept illustration of supplier invoice, purchase and receipt matching.
Concept illustration · use the HTML example for data and status
Output example · synthetic data

Choose an example scenario

INV-7821 | procurement / accounting review

Purchase order PO-1048
40 × HK$96 = HK$3,840
Receipt GRN-0881
38 boxes
Invoice INV-7821
40 × HK$102 = HK$4,080
Quantity difference
2 boxes
Unit price difference
HK$6
Received × PO unit price
38 × HK$96 = HK$3,648
Comparison difference
HK$4,080 − HK$3,648 = HK$432

Payment not approved. The difference is not confirmed overbilling; check partial delivery, price changes and receipt records. Same box unit throughout; tax and freight excluded.

INV-7821 | duplicate check

Matching identity
Example supplier A + INV-7821
Original record
INV-7821 / PO-1048 / GRN-0881
Handling
Keep original record; human review required

Same supplier and invoice identity reappeared; no second payable record is created.

Source field needs review

Source reference
INV-7821
Invoice unit price
Unverified; not filled automatically
Difference
Withheld

Check the source document before comparing amounts; payment is not approved.

Workflow, starting scope & human checks

From input to handoff

Receive → retain original and source → extract and check fields → duplicate checks → match orders and receipts → list differences → procurement / accounting review → bill draft if scoped.

Suggested starting scope

One company, one invoice inbox, three to five main supplier formats, one matching table and one human review path. Accounting connections depend on tools and permissions.

Key rules

Test languages, layouts, handwriting and unclear documents. Leave uncertain fields unverified. Do not create duplicate payable records for the same supplier and invoice identity; a difference is not confirmed overbilling.

AI, rules & people

AI extracts fields; rules match identity, quantity and amounts; procurement / accounting checks exceptions, tax, account coding and payment approval.

Integration & exception boundaries

Check what existing tools already handle before rebuilding. No guarantee of perfect extraction, automatic payments or automatic acceptance of supplier bank-detail changes.

06

OPERATIONS / Branch Orders & Warehouse Handoff

All six use cases ↑

Branch orders come in.The warehouse receives an organised picking list.

Bring branch orders into one entry point and organise by product code, box / unit conversion and cutoff. Check availability, prepare allocation suggestions and picking lists, and track shortages separately without retyping.

For

Wholesale, retail, multi-branch and central-warehouse ordering teams.

What you receive

Branch order and status records, cutoff demand totals, pending allocation and shortage lists, location-ordered picking lists, and dispatch / handoff records within scope.

Concept illustration of branch order consolidation and warehouse handoff.
Concept illustration · use the HTML example for data and status
Output example · synthetic data

Choose an example scenario

Example batch | 14:00 cutoff

Branch A
T-001 3 / T-002 1 boxes
Branch B
T-001 2 / T-002 1 boxes
Branch C
T-001 2 / T-002 1 boxes
Branch D
T-001 1 / T-002 0 boxes
Branch E
T-001 2 / T-002 1 boxes
T-001 total
10 boxes
T-002 total
4 boxes

Five branches, two products. T-001 is short by 2 boxes / 48 units. Allocation awaits supervisor confirmation; not dispatched.

View suggested picking list

Available inputs already account for existing reservations; do not deduct twice. A suggestion is not a reservation, dispatch or delivery.

A-01 · T-001
8 boxes / 192 units
A-03 · T-002
4 boxes / 48 units

T-001 | allocation pending

T-001
Requested 10 boxes · available 8 · suggested 8 / 192 units · short 2 / 48 units
T-002
Requested 4 boxes · available 6 · suggested 4 / 48 units · short 0 / 0 units

T-001 is short by 2 boxes / 48 units. Agreed rules or a supervisor determine allocation. No branch delivery notes created; not dispatched.

Separate exception | product needs review

Input
Unrecognised product code (excluded from the batch)
Product master
T-001 / T-002
Handling
Owner checks the master; no automatic similar-name matching

No guessed unit conversion, batch-total changes or picking records.

Workflow, starting scope & human checks

From input to handoff

Authorised branch submission → deduplication and master checks → 14:00 cutoff totals → availability checks → supervisor allocation review → location-ordered picking suggestions → warehouse confirms actual dispatch.

Suggested starting scope

One central warehouse, up to five branch roles, one product master, one cutoff process and one picking output. Full purchasing, accounting, returns, batch / expiry tracking and route planning are separate.

Key rules

Branches access authorised records only; repeated submissions do not create new orders. SKUs must match the master: T-001 is 24 units per box at A-01; T-002 is 12 at A-03. Late orders enter the next batch or an approved exception. Agreed rules or a supervisor allocate shortages.

AI, rules & people

Rules handle master data, conversions, deduplication and batches. AI may summarise the handoff, not guess SKUs or priority branches. Supervisors approve allocation; warehouse staff confirm actual dispatched quantities.

Integration & exception boundaries

Available quantities already account for reservations. A picking suggestion is not a reservation, dispatch or delivery; no individual delivery note before branch allocation is confirmed. This website example creates no warehouse or permissions backend.

START SMALL

Connect one piece of work.Then decide how far to grow.

You do not need to rebuild the whole company at once. Choose repeatable work with clear inputs and outputs, build a usable first version, then adjust based on real use.

  1. 01

    Show us how it works today.

    Use sanitised examples to show where information comes from, who handles it and what needs to be delivered.

  2. 02

    Define the first useful workflow.

    Agree the entry point, tools, rules, outputs and human handoffs without putting every feature into phase one.

  3. 03

    Build, then test the exceptions.

    Test missing data, duplicate inputs, connection failures, version changes and decisions that need a person, as well as the normal path.

  4. 04

    Hand it over, then decide what comes next.

    Document operation and handover, observe actual use, then decide whether to add sources, rules or departments.

Make one workflow work well before expanding.

DELIVERY & SCOPE

Connected and understandable.With someone to handle exceptions.

01

Working workflows and useful outputs.

Connect agreed sources and tools to produce the notifications, records, drafts, reports or action lists your team needs.

02

Exception handling and handover.

Know the current step, when work should stop and who takes over, with the necessary rules, tests and operating guidance.

03

Clear ongoing arrangements.

List setup, tool usage, ongoing maintenance and new requirements separately so future costs are clear.

Build scope
Depends on sources, formats, volume, rules and output requirements.
Running costs
Based on the tools, hosting, third-party services and AI usage needed.
Ongoing support
Monitoring, maintenance, rule changes and new connections are scoped separately.

Acceptance checks cover correct outputs, missing-data and exception flags, duplicate handling, effective human checks, and whether the team can inspect and take over. Time or cost savings require measurement.

BEFORE WE BEGIN

Before we begin.

Can we start without a complete SOP or AI knowledge?

Yes. Start with a sanitised input and the output you want. We map the steps together; you do not need a flowchart in advance.

Do we need to replace our existing tools?

Not necessarily. We check your tools, account plans, formats and permissions before deciding what to retain, connect or add. Integration methods are confirmed before building.

Are these six ready-made packages?

No. They are relatable use cases. Scope depends on sources, volume, rules, outputs and exceptions, and can begin with a small version.

How are data and account permissions handled?

We agree required data, tools and authorisation, then set appropriate access and retention arrangements. Test data can be sanitised first; live connections follow project requirements.

What if AI is wrong or a connection fails?

Results do not all pass through automatically. Workflows include appropriate checks, failure records, alerts, retries or human handoffs. Important commitments and actions require confirmation. Ongoing monitoring and support are scoped separately.

Does the build fee include tools and maintenance?

These are listed separately: one-off setup, third-party tools or usage, ongoing maintenance and new features. Undefined scope is not described as all-inclusive.

Can our team operate and manage it afterwards?

Delivery includes operating and handover material within the agreed scope. Tool accounts, administration rights, third-party services and future change arrangements are confirmed at the start.

WhatsApp, tell us what you want to solve