AI and automation
We treat AI as a tool that speeds up your work, not a showpiece. First we find the work that wastes your time, then we automate it safely.
Scope
What we cover
We start AI projects with the work your team loses time on every day. We build assistants that answer from your company documents and cite their sources, flows that run on their own between forms and emails, and tools that speed up content production. We also built Hub, the AI tools center of our own product Parisedu, and its video studio VideoTube, from the interface to the infrastructure.
Assistants that know your company
We turn your procedures, contracts and support records into a searchable knowledge base (RAG). The assistant shows the document behind every answer, and when the documents don’t cover a question, it says so instead of guessing.
- Document collection and update flow
- Assistant interface that cites sources
- Permissions that set who can see which documents
- Answer quality test set built from real questions
Workflow automation
We turn information that gets moved by hand between forms, emails, spreadsheets and the CRM into flows that run on their own. AI handles the classifying, summarizing and drafting; critical decisions stay with a person who approves them.
- Current process map and automation plan
- Flows built with n8n, Make or custom code
- Error alerts and activity log
- Usage guide for your team
Chatbots and support assistants
We build bots that give your customers a first answer on your site, WhatsApp and Telegram. When the bot can’t resolve something, it hands the conversation, with its history, to the right person on your team.
- Bot persona and response rules
- Site and messaging channel integration
- Handoff flow to a human agent
- Conversation and topic analytics panel
Content production tools
We speed up text, image and video production with tools tuned to your brand’s tone and templates. In Parisedu VideoTube, users describe what they want, the film is built scene by scene, and it can then be edited on the same timeline.
- Brand tone and template library
- Production interface built for your team
- Approval and publishing flow
- Usage and cost tracking
Data, reporting and summaries
We bring your sales, support and marketing data into one place and turn it into regular summaries. The report doesn’t just list numbers; it also says what changed and why it matters.
- Connecting your data sources
- Automatic weekly and monthly summaries
- Alerts for unusual changes
- Dashboard and email report
AI agents
We build agents that use tools to carry out multi-step work on their own, like research, data collection or opening tickets. What each agent can access, when it asks for approval and what it logs are defined up front.
- Task definition and permission limits
- Tool and API connections
- Approval points and activity log
- Evaluation and monitoring panel
Document reading and data extraction
We read the fields you need from documents like invoices, contracts and application forms and pass them into your systems. Fields the system can’t read with confidence are flagged for review.
- Field schema for each document type
- Reading and validation flow
- Transfer to accounting or ERP
- Accuracy report
Examples
The kind of work we ship
Concepts we designed in the studio to show what work in this area looks like. Client names and data are examples.
Try it
Let us show you instead
We built this small piece as an example of the service. Play with it for a bit, then let’s talk about what we could build for you.
See the other eight demos
Our work
VideoTube and Nova, its AI director
In VideoTube, you tell Nova what you want and the film comes together scene by scene. Then everything can be edited on one timeline.
Parisedu case studyProcess
How it runs
Finding the work
We sit down with your team and list the repetitive tasks that cost time. We start with the one that saves the most and carries the least risk.
Data and access
We decide which documents and systems will be used, who can access what, and which data must not leave your company.
Prototype
We build a first version that works with your own examples, and your team tries it in their daily work.
Testing and improvement
We measure the answers and flows against the test set and close the errors and edge cases.
Launch and monitoring
We connect it to your systems and put it live, then keep monitoring usage, cost and answer quality.
Why Gosedu
Working with us
We use it in our own product
We built Parisedu Hub and VideoTube. We learned how to choose models and balance cost against speed while working with real users.
Real work, not a demo
We start by picking work we can measure: hours spent, response time, error rate. We talk about success in those numbers, not in a demo.
You stay in control
The assistant shows its sources, automations ask for approval at critical steps, and every action is logged. You know where your data is processed and who has access to it.
Tools we use
- OpenAI APILanguage, image and speech models
- Anthropic APILong documents and reasoning
- LangChainAssistant and agent framework
- pgvectorVector search on Supabase
- PineconeScalable vector database
- n8nWorkflow automation
- MakeFlows with ready-made integrations
- PythonData processing and services
- WhisperSpeech to text
- Next.jsAssistant and panel interfaces
FAQ
Common questions
Is our data used to train models?
We work with model providers’ enterprise APIs; on these services, the data you send isn’t used for model training by default. For sensitive data we can mask personal information, and if needed we can work with open-source models running on your own servers.
What if the assistant gets something wrong?
We set the assistant up to answer only from your documents and to show its source in every answer; when the documents don’t cover a question, it says it doesn’t know. Before launch we build a test set from real questions and measure answer quality.
Which AI model do you use?
We aren’t tied to a single model. We choose between OpenAI, Anthropic and open-source models based on the type of work, the language, speed and cost, and build a setup that lets you switch models when needed.
Does it work with our current systems?
Most CRM, email, spreadsheet, e-commerce and accounting systems can be connected through an API. During discovery we review the tools you use and tell you up front how the connection will work and where its limits are.
What drives the cost? Is usage billed separately?
Setup cost depends on the scope of the work, the number of systems to connect and whether you need a custom interface. Model usage fees come on top and depend on volume; before launch we estimate the monthly cost and make it trackable from a dashboard.
Let’s start
Let’s talk about your project.
Tell us in a few sentences what you want to build, and we’ll come back with a roadmap and a quote.
