Agents that work inside your systems
We build AI agents connected to the ERP, prepare teams to work with them, run them and give the company control over what they do.
- Customer emailQuote request: 2 × 40' HC, Leixões → Luanda, general cargocompras@industria-exemplo.pt · 09:31
- Reads the request and checks Sage X3Existing customer, 30-day payment termsCurrent carrier rate and next sailing on 14/11
- Prepares the quoteOcean freight 2 × 40' HC€3,700Bunker adjustment factor (BAF)€620Terminal handling (THC)€380Bill of lading issue€75Total€4,775
- A person approves11% margin, below the 12% limit: goes to the sales repApproved by Ana R. · 09:38
- Sends and recordsEmail to the customer, quote in Sage X3 and opportunity in the CRMNobody types in any data
the cycle
Build, enable, run and govern
An AI pilot that nobody runs or controls stalls halfway. That is why the same team handles all four stages.
Build the agents
Agents connected to the ERP and to each sector’s process, with a person approving at the critical points.
Enable the teams
Training, workshops and follow-up, until teams work with AI every day.
Run and observe
Every agent decision keeps its trace, cost and outcome, and the Sevways team looks after the agents in production.
Govern
Inventory, rules, permissions, internal communication and the evidence compliance requires, in one place.
agents
The agents, one by one
Quotes and Proposals
Reads the quote requests customers send by email, prepares the proposal and the reply, answers the customer and records it in the CRM. A person approves before it is sent.
Container Tracking
Reads the customer's request, looks up the container status in the ERP and on the shipping lines' websites, compares it with the usual times for the route and prepares the reply. A person approves before it is sent.
Transport Deviation Alerts
Compares each shipment with transport patterns and, on its own initiative, warns the team of delays and interruptions. The agent raises the alert and people decide what to do.
Customer Accounts
Answers customer emails about their account and raises internal alerts for collections and management, based on payment habits and news about each customer.
Customs Compliance
Checks the cargo and port-call documents against customs rules and flags what is missing or does not match before it goes to customs. A person approves before submission.
Vessel Data Entry
Automatically fills in each vessel's data in the system from the documents and available sources, so the team only has to confirm.
Natural Stone Quoting
Reads requests for natural stone works and projects and prepares the quote piece by piece, with costs calculated from the factory's industrial costing in the ERP.
Production Manager's Assistant
Schedules cutting and manufacturing orders across the factory's machines in the most efficient way and warns when a disruption outside the usual pattern appears.
MCP Connectors
Connect Claude, ChatGPT or Gemini to the data in Ports & Shipping and Stonify OnManage.
See AI agents
How each agent runs, the path to production and the test with real documents.
human in the loop
How an agent runs in production
Trigger
An event in the ERP, an email or a user request starts the run.
Gathering
The agent reads data from authorised systems, with the permissions of the user who invokes it, and gathers the context it needs.
Proposal
It produces the result (an entry, a reply or a document) with a justification.
Validation
Above the defined value limit, or when in doubt, a person approves, corrects or rejects.
Logging
Actions and decisions remain auditable and feed the process metrics.
AI enablement
Teams that work with AI every day
The AI Act asks companies to make sure that people who use AI understand what they are using. We start with management and reach every team.
Management workshop
Where AI pays off in the company, what the risks are and where to start.
Applied training
Hands-on sessions on each department's own documents and processes, not on generic examples.
Assistants on internal data
Search and answers across contracts, manuals and support history, with the source cited.
Follow-up
Internal champions in each area and regular reviews, until the company maintains on its own what is up and running.
agent operations
We run the agents for you
An agent in production needs someone to watch it, tune it and respond when it fails. The same team that built it does this.
Production monitoring
Daily monitoring, correction of deviations and tuning of rules and value limits.
Monthly review
What the agents did, what they cost, what went through human validation and what should change.
Safe new versions
Model or rule changes tested on real cases before they go into production.
product · MAJISTI
MAJISTI, the company's AI control tower
One product with two modules. Governance sets what each agent may do; observability shows what it did. Under development.
Governance
Inventory of agents and models, rules and permissions, internal communication and evidence for the AI Act and for audits.
Observability
Trace of every decision, cost per task, measured quality, AI infrastructure health and alerts.
See MAJISTI
The two modules, who it is for and how it works.
in production
Who already works this way
“Today AI warns us before an error reaches the customer. A few years ago, this seemed impossible.” Paulo Henrique Tomás, Managing Director, Grupo Ibero Linhas.
Got a project? Let us talk.
Answer within 24 working hours.