An AI agent system that handles document work automatically from start to finish: pulling data from websites and document files, turning unstructured text into data tables, then taking the analysis further with an LLM. The output is a dashboard and an insight report, with the charts generated for you. It suits work that involves reading through the same large volume of documents week after week.
Who it is for
Organisations that regularly have to track and summarise large numbers of documents or announcements from many sources: procurement work, information management, or data analysis teams.
Collecting data from scattered sources
Most information-tracking work is spent on the same steps: visiting one site at a time, opening one file at a time, then copying the results into a spreadsheet. The AI agent takes that over, pulling data from the websites you specify (a procurement system, say, or an education information system) and reading attachments in several formats within the same process.
- Automatic collection from the websites you specify
- Supports .html, .docx and .pdf files and online spreadsheets
- You can also upload documents for the system to extract
Turning documents into data you can use
The information in a document is long-form text: a person reads it and understands it, but no system can compute with it. The agent extracts that unstructured content into data tables in a structured format, split into the fields you actually use: project type, budget, announcement date, province and target group.
Once the data is in table form, filtering, grouping and comparing across documents can be done straight away, without going back to open the original files.
- Key fields separated out of the body text automatically
- Data tables ready for further analysis
- Data from many documents brought into one common structure
Analysis and summary with AI
Once the data is structured, the system uses AI and LLMs to take the analysis further: quantitative summaries such as the number of projects, the type of work and the timescales; views by period, area or target group; and on to strategy starting points and risk analysis. The results are assembled automatically into dashboards and insight reports with charts.
- Quantitative summaries: number of projects, type of work, duration
- Analysis by period, area and target group
- Strategy starting points and risk analysis
- Dashboards and reports with charts generated automatically
Extending it into a question-and-answer assistant
The same body of data can be extended into a chat assistant your team asks questions of in natural language. Examples ThinkVerse AI has built include an assistant that explains, step by step and in Thai, how to submit a project proposal, and a security operations assistant (SOC Assistant) that answers questions from system logs such as Azure Sentinel and Cisco IPS, citing the supporting evidence alongside each answer.
- Ask directly in Thai or English
- Answers reference the data the system has collected and extracted
- Buttons for frequently asked questions, ready to tap
What it does
- Automatic collection from websites and documents
- Supports .html, .docx, .pdf and online spreadsheets
- Documents converted into tables ready for analysis
- Quantitative summaries and deeper analysis with an LLM
- Views by period, area and target group
- Dashboards and insight reports generated automatically
- Extendable into a chat-based question-and-answer assistant
- Less team time lost to repetitive work
Want something like this?
Tell us your situation and we will work out what needs adapting to fit it.