01
Individual productivity
Identify where AI can help with repetitive steps in research, writing, information management, or development. Build a workflow you can keep using.
Tools / Automation / Agent prototypes
AI Workflow Consulting
AI workflow consulting for individuals and teams. Start with a task you already do, compare tools and models, and test a knowledge base or AI agent prototype before committing to a wider rollout.
Discuss your workflow01 / Use cases
Start with a problem worth solving
02
Team workflows
Map how information moves between people and tools. Define what AI can handle, where people should review the output, and what access the tools need.
03
Product validation
If you are considering a knowledge base, an AI agent, or another AI product, test a core use case and assess its results, cost, and feasibility.
02 / Deliverables
A plan you can put to work
Document the current steps, where time goes, which problems to address first, and how to measure improvement.
Compare options by task quality, cost, data requirements, and ease of use. Explain why a tool fits and where its limits are.
Specify inputs, processing, human review, and outputs, with prompts, operating instructions, or automation configuration where appropriate.
Build or guide a test of an agreed use case such as knowledge retrieval, document processing, or agent tool calls. Record results and limitations.
Explain how to use the workflow, check its output, and troubleshoot so the individual or team can maintain it.
Separate immediate improvements from engineering work and open questions. Outline the stages and effort needed for the next steps.
03 / Process
Test on one task before rolling it out
01
Describe the workflow
Share how you do the work today, what takes longest, the tools you have tried, and the result you want.
02
Agree on scope
Choose a priority use case and agree on deliverables, collaboration, timing, and cost. Identify any engineering work separately.
03
Test the approach
Use agreed examples or de-identified materials to test the workflow and tools. Record costs, failure cases, and where human review is needed.
04
Hand over and review
Deliver the plan, agreed configuration or prototype, and instructions. Explain the results and which steps are still open.
04 / Case study
Bringing source material into an AI workflow
Personal project · Publicly available
Agent Reader · From articles to agent context
When an agent cannot read a public WeChat article directly, research can stall at copying and formatting. Agent Reader turns supported individual articles into Markdown with source information for the next research, writing or knowledge-management step.
- 01Provide the article
Submit a supported public WeChat Official Account article through the website.
- 02Prepare content and sources
Return Markdown, article information and image links, with processing notices.
- 03Continue the workflow
Copy or download Markdown, or use the same API in an existing workflow.
- 04Check the result
Review content, images and sources, and handle failures, expired caches or shared limits.
Project role
Built the article-reading service and web interface, connecting input checks, structured results, caching and error feedback into a repeatable workflow.
Design decisions
The website and API share the same reading capability. Results retain sources and notices, caching supports repeated reads, and public limits bound shared resource use.
Completed verification
Acceptance records cover real article reads, cache reuse, invalid input, and mobile-viewport checks of article display, copying and Markdown downloads.
A result you can try
Use the public website to read a supported article and get Markdown for subsequent tools. Its integration guide helps assess the fit for an existing workflow.
Supported scopeIndividual public WeChat Official Account articles are supported. Reading depends on the source site and shared public limits. Specific agent integrations and long-term success rates require further validation for each use case.
05 / Before we start
Bring one task, the steps you follow today, and the outcome you want. A short description is enough for the first email. Use de-identified examples, and we can agree on deliverables, tool costs, and maintenance before work begins.