Common problems, and how we solve them
Six blueprints for problems we see again and again in growing businesses — what we build, how it works and what you get.
Solution blueprints at a glance
Blueprints show how we typically solve these problems; every engagement is scoped to your business.
AI document processing
Invoices and receipts read, checked and entered for you — with a person approving each batch.
How it works: AI document processingCustomer-support agent on Zalo OA and web chat
Routine customer questions answered around the clock; anything tricky goes to your team.
How it works: Customer-support agent on Zalo OA and web chatSales CRM and follow-up agent
Every lead in one place, and follow-ups drafted on time for your sales team to send.
How it works: Sales CRM and follow-up agentInternal knowledge assistant
Staff ask a question and get an answer from your own documents, with the source linked.
How it works: Internal knowledge assistantOperations dashboard
Today's sales, orders, stock and cash on one screen, updated as the work happens.
How it works: Operations dashboardLegacy system modernization
An ageing system rebuilt step by step, without stopping the business that depends on it.
How it works: Legacy system modernization
AI document processing
The problem
Accounting staff type supplier invoices and receipts into spreadsheets or accounting software by hand. It is slow, error-prone, and piles up at month end.
What we build
A document pipeline that reads invoices and receipts from email, upload or scans, extracts the fields you need, checks them against your rules, and prepares entries for approval.
How it works
- Documents arrive by a dedicated email address, upload page or shared folder.
- The AI extracts supplier, dates, line items, totals and tax details.
- Rules flag duplicates, mismatched totals and unknown suppliers.
- Your accountant reviews and approves the batch in one screen.
- Approved entries are pushed to your accounting system or exported.
What you get
- Extraction pipeline tuned on your real documents
- Review and approval screen with a full audit trail
- Integration or export to your accounting tool
Typical stack
- OCR
- LLM APIs
- Python
- PostgreSQL
- Cloud storage
Customer-support agent on Zalo OA and web chat
The problem
The same questions about prices, order status, opening hours and policies arrive all day across Zalo, web chat and email, and replies depend on who happens to be free.
What we build
A support agent that integrates with Zalo Official Account and your website chat, answers from your approved knowledge and order data, and hands conversations to a person when needed.
How it works
- We collect your FAQs, policies and product information into a reviewed knowledge base.
- The agent answers customers on Zalo OA and web chat, citing only approved content.
- Order and booking questions are answered from your live system.
- Complaints, refunds and unclear cases are handed to staff with the full conversation.
What you get
- Agent connected to Zalo Official Account and web chat
- Handover inbox for your team
- Conversation logs and an editable knowledge base
Typical stack
- Zalo OA API
- LLM APIs
- Retrieval (RAG)
- Node.js
- PostgreSQL
Sales CRM and follow-up agent
The problem
Leads arrive from the website, calls, Zalo and events, and are tracked in notebooks or spreadsheets. Follow-ups slip, and no one can see the real pipeline.
What we build
A lightweight CRM shaped around your sales process, plus an agent that reminds reps, drafts follow-up messages and summarises each deal's history.
How it works
- Leads from your website, forms and chat land in one pipeline automatically.
- Each lead moves through stages that match how you actually sell.
- The agent drafts follow-ups based on the last conversation and the next step.
- Reps edit and approve each message before it is sent.
What you get
- CRM with your pipeline stages, roles and reports
- Follow-up agent with approval before sending
- Lead capture from your website and chat channels
Typical stack
- Next.js
- TypeScript
- PostgreSQL
- LLM APIs
- Email & chat integrations
Internal knowledge assistant
The problem
Procedures, price lists and policies are spread across drives, chat history and a few senior people's heads. New staff keep asking the same questions.
What we build
A private assistant that searches your company documents and answers staff questions with links to the source, respecting who is allowed to see what.
How it works
- We connect the document sources you choose and set access rules by team.
- Documents are indexed and kept in sync as they change.
- Staff ask questions in plain language and get answers with source links.
- Unanswered questions are logged so owners can fill the gaps.
What you get
- Assistant with source citations and access control
- Connectors for your document storage
- Report of unanswered questions to improve your docs
Typical stack
- Retrieval (RAG)
- Vector search
- LLM APIs
- Python
- Single sign-on
Operations dashboard
The problem
Owners wait for someone to compile a spreadsheet to know how the business is doing, and the numbers are already out of date when they arrive.
What we build
A dashboard that pulls data from your sales, accounting and operations tools into one place and shows the numbers you decide matter, with alerts when something drifts.
How it works
- We agree on the handful of numbers you actually run the business by.
- Data is pulled automatically from each system on a schedule or in real time.
- The dashboard shows trends by branch, team or product.
- Alerts notify the right person when a number crosses a threshold you set.
What you get
- Dashboard for owners and managers, on desktop and mobile
- Automated data connections to your existing tools
- Alerts and a scheduled summary report
Typical stack
- Next.js
- PostgreSQL
- Data pipelines
- Charting
- Cloudflare
Legacy system modernization
The problem
A system built years ago still runs core work, but it is slow, hard to change, and only one person or an outside vendor knows how it works.
What we build
A modern replacement delivered in stages: we document the old system, move one module at a time, and keep both running side by side until the switch is safe.
How it works
- We map what the current system does, including the undocumented parts.
- We agree on the order of modules to replace, starting with the highest risk or value.
- Each module is rebuilt, tested against real data and run in parallel.
- Data is migrated and the old module is retired once you sign off.
What you get
- Documentation of the current system
- A modern, maintainable codebase you own
- Migrated data and a staged cut-over plan
Typical stack
- TypeScript
- Node.js
- PostgreSQL
- Docker
- CI/CD
Ready to see what one system and a few good agents could do for you?
Tell us where the work gets stuck. We will tell you honestly what we would build first — or whether you need to build anything at all.