This is what an AI system
built to work looks like

Invoices, purchase orders, delivery notes, receipts and other operational documents arrive as PDFs, images or email bodies. The system reads the document, identifies the customer or supplier by tax ID or name, maps each line to a catalogue item — and creates a draft with the original attached.

• Customer recognition by tax ID, email or name
• Every line mapped to the right catalogue item
• Accuracy that improves with every coordinator fix

Every project — customer onboarding, product development, an ops project — with a type, status, owner, milestones and tasks. The same data in four views: list, kanban, Gantt and dashboard.

• Every task has an owner, due date and priority
• Tasks also come from the assistant and from visits
• Recurring tasks that schedule themselves

A dedicated app for reps in the field: a weekly visit diary, visit recommendations from purchase history, and every customer detail in your pocket. And between customers there is driving mode: talk to Nova by voice, hands-free, to ask what the next customer bought or update a task.

• A dedicated rep app with a smart visit diary by area and city
• A visit recommendation with a reason per customer
• Voice conversation with Nova, even while driving

A private area where the customer sees where they stand, places a new order, and tracks it — without phoning anyone. Staff can also browse on the customer's behalf.

• Overview, new order, order tracking and finance
• Customer sign-in and self-service onboarding
• A direct link to the customer's own account data

Eight dashboards: target, pace, forecast and revenue mix. A smart projection by purchase cycles — not averages. And instead of just showing the gap, the system proposes how to close it: customers to win back, declines to stop, quotes to close — each with a value and an owner.

• Annual target, what's achieved, and pace vs. today's requirement
• "How do we reach the target?" — a plan that becomes tasks
• Slice by customer, agent, domain and region

Each workflow is a trigger (a request opened, a status changed, a schedule), conditions and stages of actions: Data, AI and human approval. 'Build with the assistant' writes a whole workflow from a plain-language description.

• Data, AI and Flow stages with human approval
• A knowledge base that builds itself from every approved reply
• No reply goes out on its own — it stops for your approval

Agents that check the data on their own and propose actions: a customer drifting away, a stalled quote, a weekly summary for the manager. Pick a ready-made agent from the gallery, or describe what you need in plain language and the system builds it.

• 7 ready-made agents: customer keeper, quote follow-up, manager summary and more
• Each agent sees only what its user may see, and every step is logged
• Every change waits for your approval before it happens

Agents connect to other systems over the MCP protocol, starting with Priority ERP. The admin decides which tools the agents may use, and which count as writes that need approval.

• Priority ERP connected, including an order-sync check
• One-click connection to monday.com, Notion, Atlassian, Stripe and more
• Keys are stored on the server and never shown on screen

Build the year's budget by customer, product and month, compare versions and scenarios, and make one version the official target every dashboard measures against. Reps add expected deals, and the system marks them as won when the sale arrives.

• Volume and price changes per customer tier, spread across the months automatically
• Versions, comparison and a next-year plan for management
• The year as a race: cumulative actuals vs. budget, and the projected landing