AI solutions
Build Zalo, Facebook Messenger and Telegram bots or automate a workflow around the channels and languages your business uses in Vietnam.
AT A GLANCE
Test one workflow with sample data, human review and visible operating costs.
Separate costs / conditions: Model/API, hosting and operating fees are separate. Zalo OA, multiple channels, CRM and other workflows need a separate quote.
Compare service prices →Cost depends on the amount and complexity of the work. A smaller scope may reduce the total; we confirm this after reviewing your requirements. Work fees exclude any applicable taxes; domains, hosting, paid tools/APIs and advertising are separate. The final quote confirms scope, total, taxes and timing before work begins.
Sample workflows
These illustrative scenarios use sample data to define inputs, checks and outputs. They are pilot ideas, not reported client results.
Input: sample form submissions. AI suggests a service category and owner; staff review unclear requests. Output: a categorised request list with the original text retained. Track AI usage, any connection fees and review time.
Input: a question and an approved sample document set. Check answers against their sources; missing information goes to the responsible person. Output: an answer with references or a request for help. Track AI usage, storage, document updates and cases needing review.
Input: a sample enquiry and confirmed service information. AI creates a draft; staff check the facts, any prices and the next step before sending it. Output: an approved reply. Track generation costs and editing time to assess whether the workflow is useful.
BOT DEVELOPMENT
Collect enquiries, answer questions and connect customers with your team. Choose the platform your customers use and start with one specific workflow.
For Zalo Official Accounts
Build a chatbot for your business account to collect and handle customer enquiries in Zalo.
Requires a business Zalo Official Account (OA). We check verification status, integration permissions, the service plan and messaging rules before agreeing the scope.
For business Facebook Pages
We build Messenger chatbots to handle incoming messages on your Facebook Page and support initial enquiries.
Requires a business Facebook Page and appropriate integration permissions. We check Meta’s review requirements and messaging rules before launch.
For customers and internal teams
Develop a bot for customer requests, team notifications and task updates in Telegram.
Customers must start a conversation with the bot before it can message them privately. Integrations and notifications follow each team’s access permissions.
Menus, forms and fixed answers can run on scripted flows without AI. We can add AI to handle natural-language questions or search approved documents. Unresolved requests go to your team.
We separate development costs from platform, hosting and optional AI fees, based on the agreed scope and expected usage.
Data
Define permitted sources and what each connection may read or change.
An example AI workflow with defined data sources, a focused task and a staff review step.
List permitted sources, the person who confirms the content and how updates are made. Plan for missing, conflicting or outdated documents before the pilot.
Identify connected systems and permissions for each operation. Consequential actions, such as editing records or sending content externally, follow the agreed approval steps.
Agree what information may leave your systems, where it is stored, retention periods and access permissions before using real business information in a pilot.
Pilot
Assess errors, review effort and running costs before expanding.
Agree normal examples, out-of-scope questions, missing information and connection failures. Record expected results, observed errors and cases requiring staff involvement.
Decide which outputs need approval, how users report errors and how work returns to a manual process when an output is unsuitable or unavailable.
Separate the build cost from AI usage, platform fees, hosting, integrations, data updates and staff review time. Estimate from pilot volume; agree usage limits, who monitors them and what happens if consumption exceeds expectations.
HOW WE WORK
Describe the current task, its users and its difficulties. Compare an AI approach with simpler rule-based automation before selecting a pilot.
Agree sample data, access limits, evaluation criteria and actions requiring approval. Identify the person who can judge whether the outputs are useful.
Test representative situations and review errors, costs and human checking effort. Use the findings to decide whether to improve, stop or expand.
When the results meet the agreed criteria, we deploy the workflow, hand over user guidance and agree monitoring responsibilities. Ongoing support is scoped separately.
ILLUSTRATIVE SCENARIO
An example of defining requirements and checking the output. This is an illustrative scenario, not a deployed project or a measured client result.
FAQ
A few useful answers to help define the right scope.
No. A task with clear rules may be better handled by conventional automation. AI is worth testing when inputs involve variable language or unstructured information.
No. It can misunderstand input or produce unsupported information. Approved sources, representative testing and human review help manage these limitations but do not remove them.
No. Use only necessary, permitted information. An initial pilot can use sample data with sensitive details removed while access and handling arrangements are defined.
Possible costs include AI usage, storage, integrations and monitoring. They depend on processing volume and selected services, and are estimated separately from implementation work.
LET’S DEFINE THE REQUIREMENTS
Describe the task, the channels you use and the systems it needs to connect to.