Document extraction and structured fields
Extract document content into agreed fields with source references. Missing or uncertain information goes through human review before use.
Software development for Hong Kong SMEs
Want AI to process documents, search company information or organise enquiries? We look at your actual inputs, the results you need and what should happen when something goes wrong. Some tasks suit an AI model; others are better handled by ordinary program rules. We build a system your staff can use, with review steps for results that need checking.
Extract document content into agreed fields with source references. Missing or uncertain information goes through human review before use.
Let users search documents they are authorised to read and see summaries with sources. Define what each role can access and how the system responds when it cannot find an answer.
Organise messages, classify requests or draft replies, and connect the results to your CRM or task list. External messages and important data updates follow the review steps agreed for the project.
An assistant can search information, prepare drafts and use authorised tools in a sequence. We define the actions it can take, which need approval and where it must stop if a step fails.
These are the services we can provide. Before work starts, your proposal will set out the features, deliverables, fees and schedule for your project.
The first discussion is free. If a detailed assessment is needed, we explain what we will examine and the fee before starting. A development or repair quotation lists the features, tests, deployment and handover included. Hourly rates, fixed quotations and ongoing support follow our pricing page and your project proposal.
Ongoing costs include model/API usage, document and search storage, cloud computing and third-party tools. We estimate them for the chosen services and expected use. Responsibilities and fees for monitoring, sample evaluations, data updates and maintenance are agreed separately. A prototype test does not establish the cost of running the production system.
See pricing and engagement optionsA small set of samples or a proof of concept can show whether the outputs are useful, how errors are handled and how much review your staff need to do. If the results fit the task, we discuss production data connections, user permissions, monitoring and deployment separately. Passing the sample tests does not mean every document or question will produce the same results.
Before use, we identify what data each provider receives, where it is stored, how long it is kept and whether it can be used for model training or service improvement. Policies differ between models. We check your requirements against the chosen service terms to decide whether to remove sensitive information, restrict access or use another deployment setup.
Test fields, sources, permissions and review steps against the samples and expected answers we agree on. Include missing documents, questions without an answer and external-service failures. Before production use, agree what results are acceptable, which tasks still need human handling and the usage limits.
It depends on the task. Sample tests help establish what can run automatically and what needs a person to review it. Quotations, commitments to customers and important data changes need a confirmation process agreed in advance.
We can assess this. A common approach is RAG: search documents the user is allowed to read, then answer from the retrieved information with sources. We look at where the documents come from, how they are updated, who can access them, and what message or human follow-up is needed when no answer is found.
You may not need a separate system. We look at the AI features in your current plan and whether users can access the documents they need. Cross-platform search, specific citation formats, extra access requirements or custom workflows may call for configuration, integration or custom RAG development. Features, licences and fees depend on the selected service terms. Your proposal will state what we will deliver.
Not necessarily. Document storage and model processing can happen in different places. A system with private storage may still send questions or retrieved content to an external model. We check which service handles each part, where it runs and how long it keeps the data: documents, search indexes, model processing, chat and activity logs, and backups. The actual data flow and service terms help determine whether to remove sensitive information, restrict access or choose another deployment setup. Your proposal lists the work included.
Yes. Tell us who will use it, what problem it should solve and when you need it. We will suggest an approach and discuss the estimated cost and schedule. If a detailed assessment is needed before we can quote, we explain the assessment and its fee first.
Your proposal will cover source code, deployment, account ownership, operating instructions and acceptance tests. We also discuss who will maintain the system, third-party subscription or usage fees, and support hours for your project.
ProApp
ProApp builds apps, websites, business systems and AI systems for Hong Kong SMEs, and repairs AI-generated code. For a new project, tell us your idea. For an existing system, show us where you are stuck. The founder handles technical decisions, development and communication directly.
Tell us what you need