Generative AI Customer Support and Alltegrio at Call & Contact Centre Expo 2024

67 views • • 7 minutes • 4.6

Marharyta K.
Marharyta K.

AI Expert

Generative AI customer support is one of Alltegrio’s central areas of work as an AI product development company. Built well, it gives customers support that answers from a company’s own knowledge at any hour and hands the conversation to a person when it should. To stay close to how contact centres actually operate, the Alltegrio team attended the Call & Contact Centre Expo at ExCeL London on November 27-28, 2024, and spent both days listening to speakers and talking with contact centre and CX teams.

Before the event, we invited people who would be in London on those dates to meet us. If you missed us there, you can still reach us via contact@alltegrio.com or the Contact Us form.

The Alltegrio team at the Call & Contact Centre Expo 2024 in London.

At the expo we discussed how generative AI is changing customer experience and how customer expectations are shifting with it. The conversations were useful because they came from people running support operations every day: what their agents spend time on, which questions customers repeat, and where earlier chatbot projects disappointed them.

We exchanged stories with contact centre teams, shared some of the AI customer experience projects we had worked on, and discussed how those approaches could be adapted to other businesses. Some situations called for an entirely new approach, which is often where the most useful projects start.

Generative AI Customer Support at Call & Contact Centre Expo 2024

In 2024, the Call & Contact Centre Expo ran alongside the Customer Experience (CX) Expo, bringing contact centre leaders, CX teams and technology vendors into one venue. The program covered engagement, task automation, faster resolution, customer satisfaction, and sales and loyalty, with two days of seminars and practitioner talks.

The Alltegrio team attended as guests rather than exhibitors, which gave us time to walk the exhibition floor and sit in on sessions. We focused on the move from scripted chatbots to assistants built on large language models, and the practical questions that come with it:

  • how to keep answers grounded in a company’s own knowledge rather than the model’s general training;
  • when and how to hand a conversation to a person, with the context carried over so the customer does not repeat themselves;
  • how to measure quality beyond containment rate, the share of conversations that never reach an agent.

We came back with ideas for the customer service and support solutions we were working on at the time.

You can also take a look at some of our generative AI for customer support case studies: a generative AI bot for customer service that answers from tickets, historical records and product manuals, and an interactive chatbot with an animated avatar built for a telecommunication company’s mobile app.

Talk to Us About Your Support Team

The expo has ended, and we are happy to discuss AI for your customer support online at any convenient time.

Contact Us

How to Use AI to Improve Customer Experience

Chatbots are older than most people think. ELIZA, written by Joseph Weizenbaum at MIT in the mid-1960s, imitated a psychotherapist by matching keywords and reflecting the user’s words back as questions. It understood nothing, yet some users treated it as if it did. Today’s assistants built on language models can actually work with a company’s knowledge, which makes them useful for customer experience, provided they are kept to that knowledge and know when to hand over.

We help our clients apply generative AI to customer support in the following ways:

  • Intelligent chatbots built on language models
    Current support chatbots use large language models that answer from the company’s own help content, product data and past tickets, instead of following fixed scripts. They offer 24/7 support in multiple languages and take the repetitive questions off the support team, while anything involving refunds, complaints or account changes goes to a person.
  • Ticket triage and support analytics
    If your business uses a customer experience platform, AI can be added to it. It reads incoming tickets, sorts them by topic and urgency, drafts a first reply for the agent and routes the ticket to the right team, so requests are handled faster. The same data shows which issues come up most often and where customers get stuck.
  • Recommendations and in-app engagement
    A support assistant that knows the customer’s order history and the product catalogue can recommend the right product, accessory or plan in the same conversation where it answers a question. Assistants can also live inside a product: for a telecommunication company we built an interactive chatbot with an animated avatar that talks with customers inside the company’s mobile app. Our chatbot development services cover both support and recommendation scenarios.

How a grounded support assistant is put together

Most production support assistants follow the same pattern, called retrieval-augmented generation:

  1. Index the knowledge. Help-center articles, policies, product manuals and resolved tickets are split into passages and indexed for search.
  2. Retrieve for each question. When a customer writes, the system finds the passages most relevant to the question and to the customer’s account context.
  3. Answer from what was retrieved. The language model writes a reply using only those passages, ideally with a link to the source article.
  4. Decide whether to hand over. If nothing relevant was found, the request touches money or account security, or the customer asks for a person, the conversation goes to an agent with a summary attached.
  5. Log everything. Questions, retrieved sources, answers and outcomes are stored so the team can review failures and improve the content.

The design point that matters most is step 4. A bot that tries to answer everything will eventually invent a policy. The legal risk is real: in Moffatt v. Air Canada (2024 BCCRT 149), a Canadian tribunal held the airline responsible for a bereavement-fare answer its website chatbot got wrong, rejecting the argument that the chatbot was a separate entity. More on that failure type is in our glossary entry on AI hallucination.

How to measure whether it works

Containment rate alone rewards a bot for keeping customers away from agents, even when they leave without an answer. A more honest scorecard combines several measures, reviewed on samples of real conversations:

MetricWhat it tells youWatch out for
Resolution rateShare of conversations where the customer’s issue was actually solvedNeeds human review or a follow-up check, not just “no escalation”
Answer accuracyShare of sampled answers that match current policy and sourcesPolicies change; re-test after every content update
Escalation qualityWhether handovers happen at the right moments and carry useful contextToo few handovers is as bad as too many
Repeat contactCustomers coming back about the same issue within a few daysThe clearest sign of a confident but wrong answer
Customer satisfactionPost-chat ratingLow response rates; read the comments, not just the score
Agent time savedHandling time on tickets where AI triaged or draftedCount the time spent fixing bad drafts

Rules to plan for

Disclosure is now a legal requirement in the EU. Under Article 50(1) of the EU AI Act, applicable from 2 August 2026, providers must design AI systems that interact with people so that users are informed they are dealing with an AI, unless it is obvious from the context. Data protection rules apply to the conversation logs as well, so retention periods and access controls for transcripts belong in the design from the start. A human-in-the-loop step for refunds, complaints and account changes covers both the legal and the customer-experience side.

After the Call & Contact Centre Expo

Want to talk more about our cases or AI customer experience automation in general? The expo took place in November 2024, and we are happy to continue the conversation online at any convenient time. If your team is weighing a knowledge assistant for support or internal use, our RAG and enterprise knowledge assistant page describes how we approach them.

What we heard at the expo still shapes how we build support assistants: answers grounded in the company’s own content, a clear handover to people, and quality measured on real conversations rather than on containment alone.

Let's talk about your project

We'll help you harness AI and ML technologies to create innovative software solutions

File requirements: pdf, doc, docx, rtf, ppt, pptx — max 15 MB

No file selected