Intelligence that works with you.

When a support agent opens a customer ticket, Aelita assembles the full context automatically: active contracts, recent billing events, device signal quality, and past interaction history. She drafts a response tailored to the specific issue and suggests the most likely next action, whether that is a plan change, a technician dispatch, or a billing adjustment.

• Context pull from contract, device, and billing history
• Draft response with customer-specific language
• Recommended next action based on issue pattern
• Ticket classification and priority suggestion

Customers ask "why is my bill higher this month?" and agents spend 10 minutes cross-referencing ledger entries. Aelita answers in seconds. She reads the invoice, identifies the delta, traces it to the specific event (plan change, addon activation, proration), and produces a plain-language explanation the agent can send directly or adapt.

• Invoice explanation in plain language
• Anomaly detection across billing cycles
• Smart Price Matrix conflict detection
• Pro-rata calculation verification

Aelita processes incoming support tickets automatically: reading the customer's message, pulling their contract and billing context from the ERP, running network diagnostics where relevant, and drafting a categorized response. Tickets are routed through session-based workflows (drafting, HQ routing, verification) with learning hooks that improve accuracy over time based on agent corrections.

• Automatic ticket classification and priority assignment
• Customer context assembly from ERP data
• Draft response generation with routing suggestions
• Self-learning from agent corrections and overrides

When a customer calls outside business hours, the IVR system captures their details and hands them to Aelita. She performs a full diagnostic sequence (billing status, ONU signal quality, active outage zones), then creates an osTicket entry with the diagnosis attached and sends the customer an email summary. No human involvement needed until morning, and the technician starts their day with full context.

• IVR call capture with customer identification
• Automated billing, ONU, and outage diagnostics
• osTicket creation with diagnosis report
• Customer email notification with status summary

When a customer complains their connection is down and their address falls inside a maintenance window your engineers already announced, Aelita doesn't treat it as a new fault. She checks the complaint against the scheduled work, recognises the customer is in scope while the window is open, and leads the reply with a calm acknowledgement: planned work is under way, here is when it is expected to finish. No needless diagnostics, no escalation to the NOC, no truck roll for an interruption you planned on purpose. The moment the window has not yet started or has already closed, she treats the report as a genuine fault again, so a real problem is never waved away as maintenance.

• Matches the complaint against the announced maintenance schedule before reacting
• Inside the window: reply leads with the planned work and its expected end time
• Outside the window: handled as a real fault, full diagnostics run as normal
• Distinct from outage clustering — this is planned work, not an unexpected drop

When two or more customers in the same neighbourhood drop offline within a short window, Aelita recognises the pattern as a single infrastructure event, not a handful of unrelated tickets. She groups the affected customers, flags the OLT port and location, and raises one coordinated incident so the NOC can dispatch a cable team once instead of chasing each report separately.

• Automatic clustering of related wire-down events on the same port
• Two-or-more threshold; single-customer drops stay out of the noise
• One incident note for the whole cluster, with every affected contract attached
• Saves agents the manual work of cross-referencing reports

Before a technician is dispatched, Aelita reviews the customer's network stack: ONU signal levels, OLT port status, WiFi configuration, DHCP profile, and recent reboot history. She flags issues that can be resolved remotely (a misconfigured WiFi channel, a DHCP conflict, an ONU that just needs a reboot), saving a truck roll.

• ONU/OLT telemetry summary and interpretation
• WiFi configuration verification
• DHCP profile review and conflict detection
• Remote reboot guidance before truck roll

Network engineers draw infrastructure diagrams: fiber routes, OLT placements, splice closures, patch panels. Aelita converts these drawings into structured topology data that feeds into the coverage map and network inventory. No manual data entry from a Visio export or a whiteboard photo.

• Infrastructure drawings to topology data
• Node and link extraction from diagrams
• Coverage map integration
• Whiteboard and Visio format support

A warehouse clerk photographs a supplier invoice. Aelita uses Claude Vision to extract every line item (product names, quantities, unit prices, totals, tax breakdowns) and maps them to your inventory catalog. She pre-fills the Goods Received Note and highlights discrepancies against the original purchase order: wrong quantities, price mismatches, missing items.

• Supplier invoice photo or PDF to structured data (Claude Vision OCR)
• Line item extraction: product, quantity, price, tax
• Auto-reconciliation against purchase orders
• Discrepancy highlighting and GRN pre-fill

Aelita continuously monitors billing cycles for exceptions that need human attention: failed payment retries, contracts with expired payment methods, suspended accounts approaching grace period limits, and credit notes that exceed threshold amounts. Each exception comes with a recommended action and priority level.

• Real-time billing exception monitoring
• Escalation queues with priority ranking
• Recommended resolution actions per exception
• Pattern detection across exception types

Aelita cross-references coverage maps with contract data to identify addresses within your network footprint that aren't yet customers (coverage-qualified leads). She generates segmented promotional lists based on plan type, location, or contract age, and normalizes messy address data for campaign targeting.

• Coverage-qualified lead identification
• Segmented promotional list generation
• Address normalization and deduplication
• Churn risk scoring for retention campaigns

A daily automated review of team activity across all notes and tasks. Aelita identifies overdue work, unassigned tasks, missing schedules, and quality issues, delivering a prioritized summary to team leads so nothing falls through the cracks.

• Daily quality audit across all teams and departments
• Tiered severity: neglected, overdue, needs attention
• Actionable recommendations per flagged item
• Intelligent caching: only re-analyses changed notes

Aelita searches your internal documentation library (SOPs, guides, vendor manuals, policy documents) using v9 semantic search powered by a local voyage-4-nano embedding model. Multilingual: an English question finds the relevant Bulgarian SOP. ACL-filtered: agents only see documents they have access to.

• Semantic vector search across vault and SOP library
• Multilingual: English ↔ Bulgarian retrieval
• Per-tenant scoping; staff only see their own SOPs
• ACL hard filter at retrieval; never leaks restricted content

When a staff member hits the "Report a Bug" button, Aelita walks them through it. She checks whether the issue is truly a system defect or actually a customer-specific case that belongs on a walk-in note, asks clarifying questions to fill in the missing context, assembles a clean ticket, and routes it to the right owner. Fewer vague bug reports, faster fixes.

• Decision help: defect vs. customer-specific case vs. training question
• Clarifying questions to capture steps, URL, and affected account
• Clean, reproducible tickets instead of one-liners
• Automatic routing to the owning team

Atlas is Aelita's welcome mat for new hires. On day one, a new agent, billing clerk, or technician can open a single illustrated guide that walks them through the ERP by concept (money, network, people, places) so they learn which module opens for which task. Discoverable from inside the Knowledge Base, designed to be read end-to-end in one sitting.

• One guided tour of the whole ERP, organised by concept
• Role-agnostic: works for agents, finance, NOC, HR, warehouse
• Discoverable from inside the Knowledge Base on day one
• Printable and shareable as an onboarding packet

When staff ask "what promos are running right now?", Aelita doesn't guess. She pulls the current active promotions directly from the pricing engine in real time. Every answer reflects today's live rules: promo names, validity dates, applicable plan groups, discount details. No stale documentation, no outdated spreadsheets.

• Real-time promotion data from the Smart Price Matrix
• Always current; reflects today's active rules, not last week's
• Promo details: validity periods, plan groups, discount terms
• Staff can confidently quote offers to customers on the spot

When a staff member writes a new note, Aelita reads the text and suggests the right category — so notes are filed correctly the first time. Correct filing keeps everything downstream accurate: reporting, escalation, and team routing all depend on the category being right. The suggestion is advisory; the agent stays in control of the final pick and can accept it or choose another.

• Reads the note text and suggests an appropriate category
• Notes filed correctly the first time, not re-classified later
• Keeps downstream reporting, escalation, and routing accurate
• Advisory only — the agent always makes the final choice

Management-facing reports track every AI interaction: session counts, token breakdowns, month-to-date spend against a monthly credit, and per-model and per-feature cost. Alongside the spend view sits an answer-quality queue where supervisors rate Aelita's help answers and flag any that need the knowledge base improved. A credit-usage banner gives a green / amber / red read on the budget at a glance. Governance and budget oversight are built in, not bolted on.

• Spend report: month-to-date against a monthly credit, by model and by feature
• Activity dashboard: session counts, token breakdowns, and ticket-response stats
• Answer-quality queue: Good / Needs-fix / Fixed verdicts on help answers
• Needs-fix flags point back to the knowledge that should be improved

Aelita now takes a deeper thinking pass on every multi-step job (verifying invoices, drafting ticket replies, running daily operations reviews) so her answers hold up across complex cases, not just simple ones. Document scanning also gained full-colour vision, letting her read faded handwriting, recognise ID cards by their colour cues, and handle higher-resolution supplier invoices without losing small-print details.

• Deeper reasoning on multi-step reviews and verifications
• Full-colour document vision; better on faded text and coloured forms
• Higher-resolution scanning for detailed supplier invoices
• More consistent output quality across long, context-heavy tasks

Aelita is not a one-shot Q&A box. Each interaction opens a persistent conversation session. Ask a question, get an answer, then follow up with "tell me more about that" or "what about the billing part?" and Aelita remembers the full context of your prior exchange. No need to repeat yourself or re-explain the situation.

• Persistent context across follow-up questions
• Per-user sessions; your conversation stays yours
• Automatic session recovery if connection drops
• Clear session anytime to start fresh

Every response Aelita generates can be copied to your clipboard with a single click. Paste it directly into a ticket reply, an email to a customer, an internal note, or a Slack message. No manual selecting and formatting. The response is ready to use the moment Aelita produces it.

• One-click copy on every AI response
• Formatted text ready for tickets, emails, and notes
• Eliminates manual copy-select-paste workflow
• Works across all Aelita capabilities