Banking
In the CRM, the assistant summarizes each customer complaint and suggests a category. The adviser confirms or corrects it before saving.
AI integration · Companies and public bodies
AI integration inside the tools you already use
You keep your software. We connect artificial intelligence models to your ERP, CRM, document management system or email, and your teams use them without leaving the screens they know. Someone on your team approves each sensitive action.
Your current software, still without AI
Your employees paste extracts from client files into a public chatbot, with no rules and no record.
Your case officers search several shared folders for a circular before they can answer a citizen.
Your advisers log complaints as free text in the CRM and have no time to sort them by type.
A software vendor tells you to replace your ERP to “get AI”. Your project would start with a migration that takes months.
You don’t know what an AI model will cost once the whole team uses it.
What you get
Your teams keep their tools. We add AI to them, with access controls and cost tracking.
An assessment of your tools
We review your software and its APIs, then rank the possible integrations by expected benefit and risk.
Model selection
We test commercial and open models on your real cases. You choose based on quality, cost and how sensitive your data is.
Connectors to your tools
We link the model to your ERP, CRM, document management system, email or intranet. Your teams see its suggestions in the screens they use.
Safeguards in place
We carry over the access rights from your tools, log every exchange and test the answers before rollout.
Usage costs estimated in advance
We estimate this monthly cost before rollout. You then track actual model usage on a dashboard.
Use cases by sector
Public administration
From the intranet, an officer asks a question. The assistant answers and cites the circular or internal memo it relies on.
Retail and distribution
In the customer service tool, the model drafts a reply from the order history. The agent reviews it before sending.
How we work
Tool assessment
We map your tools and the data they exchange, then choose the first integration point with you.
You receive
an assessment report, a ranked list of integrations and a first estimate of usage costs.
Prototype on one tool
We connect a model to a single tool, for one use case, with your real data and a small group of users.
You receive
a working integration and its first indicators, so you can make a go or no-go decision.
Rollout
We extend the integration to other tools and other teams, with tests, access rights and logging.
You receive
the integration in production, documented, with operating procedures for your teams.
Monitoring and maintenance
We monitor answer quality and monthly cost. We update the connectors when your software moves to a new version.
You receive
a monitoring dashboard and a regular review with your team.
Data and safeguards
We apply Moroccan law 09-08 and choose with you the hosting that suits your data. The assistant follows your access rights: each user sees the data they are allowed to see, nothing more.
Before the tool sends a message, changes a file or records a decision about a customer, it waits for approval from someone on your team.
We log requests, answers and approvals. If something goes wrong, you can trace who asked what and who approved it.
We set monthly usage caps with you, so the model bill stays within the planned budget.
Technologies we use
- Commercial modelsGPT, Claude, Gemini
- Open modelsLlama, Qwen, hosted on your servers if needed
- n8n to orchestrate exchanges between tools
- Your software’s REST APIs and webhooks
- Python and TypeScript for the connectors
- PostgreSQL and vector search for your documents
- Hosting in Morocco, on your servers or in the cloud
Frequently asked questions
No. We connect the model to your software through its APIs or webhooks. If a tool offers neither, we flag it during the assessment and work from scheduled exports or an intermediate database.
We test OpenAI’s GPT models, the ones behind ChatGPT, alongside Claude, Gemini and open models, on your own cases. You choose based on the test results. We build the integration so you can switch models later without rebuilding the connectors.
It leaves Morocco if you choose a model or hosting outside the country. Transferring personal data abroad may require authorization from the CNDP, Morocco’s data protection authority, under law 09-08. For sensitive data, we offer an open model hosted in Morocco or on your own servers.
We charge a fixed fee for the assessment. The integration price depends on the number of tools to connect and the data volume, and we quote it at the end of the assessment.
For a prototype limited to one tool and one task, allow from a few days for a simple case to a few weeks. The exact schedule also depends on access to your systems, and we set it with you at the end of the assessment.
We can, under a monthly support plan: connector updates, quality and cost monitoring, and a model change if needed. If you prefer, your teams take it over with the documentation we hand them.
AI solutions delivered fast, built to last.
Let’s talk about your project
Describe your need in a few lines: the process involved and the tools you use today. For a public tender, include its reference. We reply by email or by phone.
Based in Rabat, we work across Morocco, Africa, the Middle East and Europe, on site or remotely.