Per-file chat
Per-file chat lets your team:
- Ask about one specific point without reopening every document
- Instantly find the document or clause a conclusion is based on
- Keep a conversation history attached to that file
Enterprise AI agents that analyze your files
A MarketerZ enterprise AI agent reads your files, applies your business rules, and produces a traceable analysis with citations. Your team keeps the final call, and saves time on every file.
The problem with reviewing files by hand
Invoices, estimates, photos, contracts: every file arrives with documents to cross-check and rules to apply. Reviewing that volume by hand takes time and leaves room for mistakes, especially when the information is scattered across several tools.
Teams end up spending more time hunting for information than making decisions. Turnaround slows down, and the record of what was actually checked often gets lost in email threads.
- Documents to review one by one: photos, invoices, estimates, contracts
- Business rules and contract clauses to cross-check by hand
- Decisions that are hard to justify later, for lack of a clear record
If you are not sure where to start, our AI consulting work helps you identify which files are worth automating first.
What a MarketerZ enterprise AI agent does
A MarketerZ enterprise AI agent reads the documents in a file - photos, invoices, estimates, contracts - and checks them against the business rules you give it. It produces a structured analysis: what qualifies, what is missing, and the associated estimate.
This AI document analysis agent does not replace your team. It prepares the groundwork and backs up every conclusion. The final decision stays with a person, file by file.
- Reads attached documents: photos, invoices, estimates, contracts
- Applies business rules you configure together with us
- Produces a structured, well-supported analysis
Traceable analysis, with cited sources
Every conclusion the agent reaches is backed by an identified source: a contract clause, a line on an invoice, a term in the general conditions. Your staff can check exactly where each part of the analysis came from with one click.
That traceability changes the nature of the work. A traceable AI agent documents its reasoning instead of simply handing down a verdict. That is what makes decisions consistent across a team, while leaving a useful record if a decision is ever challenged.
This approach follows our security and compliance practices, covered in full on a dedicated page.
Per-file chat and an admin dashboard
Every file has its own conversation thread. Your team can ask a question directly on the file in question and get an answer grounded in the documents already analyzed.
Admin dashboard
The admin dashboard gives managers a full overview:
- Status and progress of files being processed
- Document management and knowledge base updates
- A search test tool for documents already ingested
- Conversation history and the pipeline's active configuration
Automated evaluations before every release
Before every update, the agent is tested against reference cases (golden cases): known files with an already-established correct answer. This confirms that a change to the pipeline has not made the analysis worse.
These automated evaluations rely on sets of business questions built for your activity. They run before every release, so you can trust each new version before it touches a real file.
Cloud or on-premise, based on your data constraints
The agent can run in the cloud or directly on your own servers, depending on your data constraints. An on-premise AI deployment through Docker Compose keeps your documents and their processing on your infrastructure, start to finish.
A data loss prevention (DLP) policy applies either way, along with the option of a fully local mode: models hosted locally through Ollama, with no calls to an external API. That choice is made together with you, based on how sensitive your documents are.
Under the hood
A MarketerZ enterprise AI agent runs on an open stack, with no dependency on a proprietary black box:
- Orchestration: LangGraph drives the analysis pipeline step by step
- Document search: PostgreSQL with the pgvector extension powers RAG (retrieval-augmented generation) over your documents
- Language models: the Mistral API or local models through Ollama, depending on your preferences and data constraints
- API: FastAPI exposes the chat, document ingestion, and the admin dashboard
- Deployment: Docker Compose, for a cloud or on-premise install on your own infrastructure
This stack is entirely open source. You are never locked into a single vendor.
Case study: the HomeAssur claims analysis agent
HomeAssur, a French insurer specialized in short-term rental coverage, has used a MarketerZ enterprise AI agent to analyze its insurance claims since August 2026. The agent reviews attached documents - photos, invoices, estimates - checks eligibility against the contracts and general conditions, then proposes a compensation estimate.
Every analysis cites the clauses and documents it relies on. Claims staff save time on the initial review and keep the final decision on every file. The agent's AI Act status has been assessed: it is not classified as high-risk, and its transparency obligations are met.
See the full story in our HomeAssur case study.
Frequently asked questions
It is a program that reads your documents, applies your business rules, and produces a well-supported analysis with citations. It differs from a virtual assistant: it works through complex files rather than answering simple questions.
Photos, invoices, estimates, contracts, and any other text or image document relevant to your business. The list of accepted document types is configured together with you, based on your files.
Every analysis cites its sources, so you can trace a conclusion back to the clause or document it came from. Before every release, the agent is also tested against reference cases (golden cases) to check the quality of its answers.
Yes. An on-premise deployment through Docker Compose is one option, alongside cloud hosting. The choice depends on your data constraints and is made together with you.
Assess an AI agent for your files
Tell us about your files and your business rules. We will assess together whether an enterprise AI agent can save your team time, during a free 30-minute consultation.
Assess an AI agent for my files