AI-Powered Ticket Routing for Support Automation

AI Solution for Automatic Support Ticket Routing

Every day, support teams drown in a flood of tickets: billing, technical support, complaints, feature requests. Operators manually read each message and decide where to send it. The result: first response time in hours or even days, misrouting up to 20%, and frustrated customers. We automate this process using LLMs and reduce response time by 3–5 times.

How AI Solves the Routing Problem

The system, based on a Large Language Model (GPT-4o, LLaMA 3, Mistral), analyzes the ticket text and instantly determines: topic, priority, complexity, language, customer segment. It then assigns the ticket to the right team or specific agent, considering their current workload. Classification accuracy exceeds 90% from the first week, and fine-tuning on your historical data pushes it to 97%. The model uses few-shot prompting and dynamic context (client history, SLA, time of day).

Routing System Architecture

class TicketRoutingDecision(BaseModel):
    team: str                    # billing, tech_support, sales, escalation
    priority: Literal["P1", "P2", "P3", "P4"]
    assignee_id: str | None      # specific agent or None for auto-assignment
    reasoning: str               # explanation of decision
    suggested_response_template: str | None

def route_ticket(ticket: Ticket) -> TicketRoutingDecision:
    context = build_context(ticket)  # client history, current team load
    return llm_classify(ticket.text, context)

The pipeline includes preprocessing (lemmatization, stop-word removal), embedding (text-embedding-ada-002, 1536-dimensional vector), and similar ticket search in a vector database (Pinecone, Qdrant) for few-shot examples. This boosts stability on rare categories. For embeddings we use text-embedding-ada-002 (OpenAI).

Model Comparison for Routing

Load Balancing

Routing must consider current agent load. Algorithm: primary classification by competence → select least loaded agent with required competence. Load data: open tickets per agent, average handling time, status (online/offline). For urgent cases (P1), escalation to on-duty engineer with Slack/Telegram notification is implemented.

Integration with Helpdesk Systems

• Zendesk: Triggers API for automatic tagging and assignment
• Freshdesk: Webhooks + API for update ticket
• Jira Service Management: REST API, automatic rules
• ITSM systems: ServiceNow, OTRS — via REST API

All integrations use Zendesk API and similar APIs for other systems. We build custom connectors for systems without public API (via webhooks or email parsing).

Why AI Is Better than Manual Routing?

What Metrics Do We Track?

We monitor not only speed but quality: Routing Accuracy (correctly routed tickets), First Response Time (median and p95), Misrouting Rate (manually reassigned tickets), Agent Utilization. Data visualized in Grafana + helpdesk system dashboards. Additionally, we calculate economic impact: support cost reduction by 30–50% and ROI within 3–6 months.

What's Included in the Work?

We deliver a turnkey solution:

• Audit of current processes and historical tickets (sample of at least 1000 tickets).
• Selection and fine-tuning of LLM (GPT-4o or open-source model for confidentiality).
• Development of classification pipeline with few-shot examples.
• Integration with your helpdesk system via API.
• Configuration of load balancing and escalation rules.
• A/B testing for 1–2 weeks.
• Documentation, team training (2 sessions of 1 hour), 1 month of support.

Implementation Timeline?

Timeline: from 2 weeks (basic version with GPT-4o-mini) to 6 weeks (custom model, integration with ServiceNow). Cost is calculated individually — depends on ticket volume, number of integrated systems, and need for fine-tuning.

How Do We Guarantee Quality?

We set target metrics in SLA: Routing Accuracy >90%, Misrouting Rate <5%, First Response Time reduction of 60%. We conduct A/B testing before full switchover. We provide a routing accuracy guarantee in the contract.

Order a pilot project: we conduct a free audit of your support system, assess current metrics, and prepare a proposal. Contact us to discuss details. Our experience: 5+ years in AI and ML, over 30 projects for retail, fintech, and telecom.

NLP Development: Text Classification, NER, Embeddings, and Information Extraction

We often receive a task: process 50,000 support tickets — currently all manual. Dataset — 3,000 labeled examples, 12 categories, imbalance: one category occupies 40% of the sample, three at 1-2% each. Baseline accuracy — 78%. Sounds decent until you look at recall for rare classes: 0.31, 0.44, 0.28. These classes — complaints and churn threats — are most important to the business.

This is a typical NLP development project. The problem is not the algorithm but that accuracy is the wrong metric. Our experience across 30+ projects shows: we start by analyzing business metrics and only then choose the model.

Why accuracy is not the right metric for rare classes?

Accuracy ignores imbalance. If the "churn" class appears in 2% of cases, the model can predict "all good" and get 98% accuracy — but the business loses clients. Solution: F1 macro (averaged over all classes) or weighted F1. For NER — strict entity F1 (exact matches only). We guarantee: after choosing the correct metric, model quality becomes measurable and predictable.

Text Classification: From BERT to Distillation

BERT-like models are the standard for classification. ruBERT-base or ruBERT-large from DeepPavlov for Russian. multilingual-e5-large — for multiple languages in one pipeline. XLM-RoBERTa-large — a strong multilingual backbone.

Fine-tuning for classification: add a classification head on top of the [CLS] token, train for 3-5 epochs with lr=2e-5, weight decay=0.01. For imbalance — weighted CrossEntropyLoss or focal loss with gamma=2.0. Contact us — we will show a code snippet.

Imbalance case study. Dataset — 3,000 examples, imbalance 1:20. Solution: class_weight via sklearn + CrossEntropyLoss. Additionally — augmentation of rare classes via backtranslation (ru→en→ru through MarianMT). Recall for rare classes rose from 0.31 to 0.67 with a slight drop in accuracy (76%→74%). Full NLP development end-to-end took 3 weeks.

Distillation for production. BERT-large gives F1 0.89, but inference on CPU — 180ms. Distillation into DistilBERT or ruBERT-tiny2 reduces latency to 25ms with F1 0.84. Export to ONNX Runtime provides an additional 1.5-2x speedup. DistilBERT achieves 7x lower latency than BERT-large with only a 5% drop in macro F1 – a typical production trade-off.

How to choose between BERT and LLM for your task?

For most classification and extraction tasks, BERT-sized models offer the best trade-off between cost and performance. Shift to LLMs only when the task demands generation, complex reasoning, or zero-shot generalization.

NER: Named Entity Recognition

NER — extracting persons, organizations, locations, dates, amounts, document numbers. For general categories (PER, ORG, LOC), pre-trained models work well. For specialized ones (medical terms, legal concepts) — fine-tuning is needed.

Data annotation. The main cost of an NER project. For a quality model — 500-2,000 labeled sentences per entity type. Tools: Label Studio (open source) or Prodigy (by spaCy creators). IOB2 format — standard.

Architecture. Token classification on top of BERT: each token gets a label (B-PER, I-PER, O). spaCy 3.x with transformer pipeline — a convenient production choice.

Nested entities. Standard IOB models cannot handle nested entities (organization inside an address). For such tasks — span-based NER: SpanBERT or SpERT. More complex but correct.

Post-processing is mandatory. The model predicts tokens — normalized entities are needed. Date — dateparser. Amounts — regex + validation. Names — deduplication via rapidfuzz. Included in our standard delivery.

Sentiment Analysis and Opinion Mining

Binary classification positive/negative works out of the box with BERT. Complexity — aspect-based sentiment analysis (ABSA): "the restaurant has good food but terrible service." For ABSA: aspect extraction (NER) + sentiment per aspect. Joint models BERT-for-ABSA — quality on Russian data is lower due to dataset scarcity. RuSentiment, SentiRuEval — main resources.

For production with simple positive/negative/neutral: distil models are enough. Three classes, balanced dataset, 2,000+ examples — F1 macro 0.82-0.87 in 1-2 days.

Text Summarization

Extractive summarization (select sentences) — TextRank or BM25 without training. Fast, no hallucinations. Good for long documents.

Abstractive (generates new text) — seq2seq: mT5, mBART, FRED-T5, ruT5-large. For production via LLM API (GPT-4, Claude) — often the best cost/quality/speed trade-off.

Embeddings: Vector Representations of Text

Embeddings are the foundation of semantic search, deduplication, clustering, RAG. Quality critically affects downstream tasks.

Models. E5-large-v2, BGE-M3, multilingual-e5-large — strong multilingual embedders. sentence-transformers/paraphrase-multilingual-mpnet-base-v2 — fast option. For Russian: ru-en-RoSBERTa (Skoltech) performs well on semantic textual similarity.

Embedding quality evaluation uses the MTEB benchmark as standard. But top results on MTEB don't guarantee success on a domain dataset — we build domain-specific eval.

Fine-tuning embeddings. If standard models don't give the required Recall@k — contrastive learning on domain pairs with MultipleNegativesRankingLoss. How to perform this for domain data:

1. Collect 500–2,000 semantically similar pairs from your domain.
2. Apply MultipleNegativesRankingLoss with a batch size of 32–64.
3. Train for 1–3 epochs using AdamW (lr=2e-5).
4. Evaluate Recall@k on a held-out domain test set.

This approach yields a 5–15% improvement in Recall@k in practice.

Dimensionality and storage. E5-large: 1024 dim, float32 — 4KB per vector. For 10M documents — 40GB. Quantization int8 reduces to 10GB. FAISS IVF_PQ — more compact but with losses. Included in our deployment recommendations.

Information Extraction

Structured extraction is a frequent task. Examples: key contract terms, technical characteristics, dates and amounts from invoices.

1. Regex + rule-based. For INN, OGRN, amounts, dates — more reliable than neural networks. No data required.
2. NER + post-processing. For variable formats.
3. LLM with structured output. GPT‑4 / Claude with JSON schema — for complex documents. Cost: minimal per document. For 10k+ documents/day — we calculate the economics.

We guarantee a hybrid: regex/NER for typical fields + LLM for edge cases. Our guarantee is backed by years of production experience and more than 30 projects.

Work Stages

Prototype on existing data — 1-3 weeks. Production system with CI/CD — 1.5–2.5 months. Cost is calculated individually — get a consultation for a project estimate.

What's Included

• Model and pipeline architecture documentation
• Access to the model via REST API (FastAPI + ONNX)
• Client team training (2-hour webinar + Q&A)
• Accuracy guarantee on the agreed test set
• Months of post-delivery support (bug fixes, adaptation to new data)

Our Experience

Years of NLP projects from classification to RAG systems. The team includes ML engineers experienced with Hugging Face, spaCy, LangChain, MLOps. We use vLLM, Kubeflow, Weights & Biases — a production stack, not toys. Contact us to evaluate your NLP project within two days — request a free consultation on your text processing pipeline.