AzurLab CRM: our CRM with an AI assistant and AI agents, powered by Claude

AzurLab CRM is software designed, built and operated by AD AZUR DIGITAL, live at azurcrm.azurlab.cloud. It grew out of our agency work: every day we run Facebook, Instagram and Google campaigns for our clients, and we wanted every lead they bring to be called back fast, followed through to the sale and tied to the ad that brought it.

Each client has a workspace of its own, walled off from the others, in French, Arabic (right to left) or English, with dinars, amounts excluding tax and Tunisian governorates ready from day one.

The offer. AzurLab CRM is free for every client on our social media management marketing pack: we set it up, connect their pages and forms and train their team. For outside companies, or anyone interested in the CRM on its own, it is available on quotation.

A team that follows all its leads in one place (AI-generated illustration).

In most of the businesses we work with, leads lived between an Excel file, a WhatsApp group and people’s memory. Facebook enquiries were copied by hand hours later, callbacks were forgotten, duplicates piled up and nobody knew what each dinar of advertising brought in. A lead called back the next day has often already gone to a competitor.

Before: leads scattered between files, sticky notes and phones (AI-generated illustration).

The need was a tool that is simple for salespeople, useful for managers, and an AI that saves time without ever acting out of control or taking a client’s data outside its workspace.

We built the CRM end to end around three ideas: lose no lead, tell everyone what to do, measure what pays.

What the CRM does every day

• Leads in within seconds: Facebook and Instagram forms in real time, website, WhatsApp, Excel imports; every enquiry arrives with its answers and source campaign, automatically assigned to the right salesperson;
• My day: every morning, the salesperson’s mission (new leads to call, callbacks, quotes to chase, overdue items) and a call queue that runs record after record;
• the lead record: call, WhatsApp, SMS and email from the record, message templates, clear sales stages, history of every exchange and every change;
• advertising control: cost per lead and per sale for each Meta ad, an alert when a campaign spends without results;
• quotes, payments and team: PDF quotes you know were opened, deposits, sales sent to Odoo, lead distribution, targets and coaching for each salesperson;
• everywhere: an installable app for phone and computer, voice dictation, a three-language interface.

Azur, the AI assistant

Azur, shown as the AD AZUR DIGITAL robot, is on every screen. It prepares the call (form answers, history, likely objections), summarises a record, analyses dashboards and answers follow-up questions. It works only on the client’s own workspace data, and its dashboard analyses receive figures only, never contact details.

The Azur assistant next to a record (demo workspace, sample data).

AI agents: agentic AI under control

Beyond the assistant that answers, AI agents act: they write to a new lead within minutes, even in the evening, follow up silent leads, offer a time slot, prepare a quote, send a payment reminder or wake up old contacts. Six missions are created in guided mode, in five questions: mission, opportunities, tone of voice, autonomy, summary.

Creating an AI agent, step by step (demo workspace).

Autonomy is set agent by agent: observe, assisted (each message waits for a person’s approval) or supervised. Safeguards frame everything else:

• one agent in charge of a record at a time, and control handed back as soon as a salesperson steps in;
• the agent only talks about what is on the offer sheet: an off-sheet question is noted and passed to an adviser, with nothing made up;
• never a message to an unsubscribed contact, the lead’s language respected, a global pause in one click;
• every action logged, and AI use counted in credits, with a limit per salesperson.

Claude, the AI engine

The assistant and the agents run on Claude, Anthropic’s AI model, chosen for the quality of its reasoning and of its answers in French as well as Arabic. We integrated it carefully: short thinking for conversation and deeper thinking for analyses, cached instructions to cut delay and cost, and a review of agents’ messages before they are sent. Claude also helps in our workshop: our developers rely on Claude Code to write, test and review the CRM’s code.

AzurLab CRM illustrates our AI platforms and SaaS and AI agents services.