Integrating AI Into Existing Customer Support Frameworks

Generative AI has been making waves in the industry for over two years, revolutionizing how businesses engage with customers. In this blog, the Engineering team at Noom shares how they developed their AI-powered customer support solution. Noom is a digital health company offering a subscription-based mobile app that helps users achieve their wellness goals, and it relies heavily on its chatbot for customer interactions. While directly leveraging ChatGPT-4 for customer chats was a promising first step, the team identified several challenges: issues with hallucinations, a lack of customization to user needs, and a mismatch with Noom's unique communication style. To address these challenges, the team developed a customized solution. They started by using Prompt Instruction with GPT-4 to form the foundation of their AI assistant. Next, they implemented Prompt Augmentation with Noom's Knowledge Base (RAG), Dynamic Prompts based on user data, and JSON Format Responses. These elements enabled the system to accurately process user messages, understand their needs, and deliver tailored responses. Furthermore, recognizing the importance of human connection, the team integrated classification models with LLMs to identify when a human touch was needed, ensuring users felt understood and valued. This approach is a great example of companies leveraging generative AI to create customized solutions that address their unique challenges. #datascience #machinelearning #generative #LLM #chatGPT #customer #chatbot – – –  Check out the "Snacks Weekly on Data Science" podcast and subscribe, where I explain in more detail the concepts discussed in this and future posts:    -- Spotify: https://lnkd.in/gKgaMvbh   -- Apple Podcast: https://lnkd.in/gj6aPBBY    -- Youtube: https://lnkd.in/gcwPeBmR https://lnkd.in/gvJg5tMK

The AI-infused workforce is here. You can't just implement a chatbot, reduce your team size, and expect to keep meeting customer expectations. You have to reimagine the whole customer experience team. You need to think about both humans + technology and how they work together. What does this mean? ⭐ 𝐕𝐨𝐢𝐜𝐞 𝐨𝐟 𝐭𝐡𝐞 𝐂𝐮𝐬𝐭𝐨𝐦𝐞𝐫 𝐭𝐞𝐜𝐡𝐧𝐨𝐥𝐨𝐠𝐲 𝐜𝐚𝐧 𝐫𝐚𝐭𝐞 𝐜𝐮𝐬𝐭𝐨𝐦𝐞𝐫 𝐬𝐞𝐧𝐭𝐢𝐦𝐞𝐧𝐭 𝐀𝐍𝐃 𝐪𝐮𝐚𝐥𝐢𝐭𝐲 𝐢𝐧 𝐫𝐞𝐚𝐥-𝐭𝐢𝐦𝐞. These platforms can monitor 100% of tickets in real time - far superior than the current method of combining CSAT & QA. But they have to be managed properly. You can't just set it and expect the insights to flow in. Which leads us to.... ⭐ 𝐍𝐞𝐰 𝐫𝐨𝐥𝐞𝐬 𝐚𝐫𝐞 𝐞𝐦𝐞𝐫𝐠𝐢𝐧𝐠. These include: 𝘊𝘰𝘯𝘷𝘦𝘳𝘴𝘢𝘵𝘪𝘰𝘯 𝘔𝘢𝘯𝘢𝘨𝘦𝘳. This is a role that Intercom has created on their own internal support team. The focus is managing the whole customer experience journey, as customers flow from the bot to human agents. 𝘒𝘯𝘰𝘸𝘭𝘦𝘥𝘨𝘦 𝘔𝘢𝘯𝘢𝘨𝘦𝘳. Another role pioneered by Intercom, this person is responsible for keeping the knowledge base up to date and making sure it's written not just for humans to understand, but for computers to understand. 𝘝𝘰𝘪𝘤𝘦 𝘰𝘧 𝘵𝘩𝘦 𝘊𝘶𝘴𝘵𝘰𝘮𝘦𝘳 𝘈𝘯𝘢𝘭𝘺𝘴𝘵. VOC analysts will replace QA analysts. Instead of completing audits they'll be monitoring VOC systems for trends and insights that will be used to drive trainings and process improvements. 𝘈𝘐 𝘊𝘢𝘭𝘪𝘣𝘳𝘢𝘵𝘪𝘰𝘯 𝘚𝘱𝘦𝘤𝘪𝘢𝘭𝘪𝘴𝘵. Someone who will be in charge of continually monitoring performance of the AI bot, as well as other automations, to ensure they are doing their job. ⭐ 𝐏𝐨𝐝𝐬 𝐨𝐟 𝐬𝐮𝐛𝐣𝐞𝐜𝐭 𝐦𝐚𝐭𝐭𝐞𝐫 𝐞𝐱𝐩𝐞𝐫𝐭𝐬 𝐰𝐢𝐥𝐥 𝐫𝐞𝐩𝐥𝐚𝐜𝐞 𝐭𝐞𝐚𝐦𝐬 𝐨𝐟 𝐠𝐞𝐧𝐞𝐫𝐚𝐥𝐢𝐬𝐭𝐬. AI will take on the simpler tickets, and what's left will be increasingly complex. According to Declan I., the VP of Support at Intercom, the best way to manage this complexity is to upskill agents so they have particular areas of expertise, which is what Intercom has done. ⭐ 𝐘𝐨𝐮 𝐧𝐞𝐞𝐝 𝐧𝐞𝐰 𝐊𝐏𝐈𝐬 Your AI and automation tools need their own KPIs, like AI Resolution Rate. More on this in a future post. ⭐ ... 𝐀𝐧𝐝 𝐦𝐮𝐜𝐡 𝐦𝐨𝐫𝐞 Agents will need different schedules and breaks to avoid burnout. More and better training is needed for agents. And so on. ⭐ ⭐ ⭐ ⭐ 𝐈𝐍 𝐒𝐔𝐌𝐌𝐀𝐑𝐘 ⭐ ⭐ ⭐ ⭐ AI tools aren't "set it and forget it." You need to reshape your team, and then you need multiple people on your team who can monitor your AI KPIs and recalibrate the technology when KPIs start to slip. You can do this internally, or partner with a #bpo who can do it all for you. But not all BPOs can or will. If you're already working with a BPO partner, ask them about it. We have an upcoming podcast episode with Declan I., all about the lessons Intercom has learned deploying its AI technology on its internal support team. Stay tuned!  

I have been working with AI in customer support for a while now. And lately, one thing is becoming clear. This space is getting crowded. Every vendor claims their AI is the magic wand. Just plug it in, and your support problems disappear. But the reality is different. AI isn’t magic. It’s a strategy. It has to be planned, adapted, and rolled out based on: 🔹 Your goals 🔹 Your current challenges 🔹 And your team’s capacity Most support leaders we speak with aren’t confused about the tech. They are confused about where to use it. That’s the real challenge. So we created a simple matrix to help teams make better AI decisions. It’s built on just two questions: 1. What’s the risk if AI gets this wrong 2. How complex is the task When you map support work using this lens, things get clearer: - Use AI fully for low risk, repetitive tasks like tagging, triaging, or summarising. - Use AI as a helper for pattern based tasks like routing, recommending actions, or drafting replies. - Keep humans in control for high risk, complex issues like escalations, complaints, or anything tied to revenue.   And here’s the other mindset shift: Don’t think of support AI as one giant bot. Think of it as a system of specialised agents: 🔹 Analyzers – Understand queries, profiles, logs 🔹 Orchestrators – Manage workflows, routing 🔹 Reasoners – Diagnose problems 🔹 Recommenders – Suggest next steps 🔹 Responders – Write or send replies Each agent plays a specific role, just like your support team does. Done right, AI doesn’t replace humans. It supports them, speeds them up, and helps them focus where it matters most. This approach is also being recognised by the front-runners in the space. At a recent ServiceNow event I attended, many speakers echoed the same thought: AI is not one size fits all. It must be tailored to each organisation’s structure, systems, and bandwidth.   Let’s stop using AI for the sake of it. Let’s start using it where it actually makes a difference.   If you are building or evaluating AI for support and want to walk through the matrix, Feel free to drop me a message.  Always happy to exchange notes.

Today, we're publishing the AI Agent Engine! 🙌 It's a distillation of our learnings from many successful deployments of AI agents at enterprises. Ultimately, this is what's required for a successful implementation of AI agents. Of course, this is specific to our space (customer service & experience), but the themes will carry over to any vertical. 1. First, you have the "AI agent", defined as a software system that can autonomously do the work of a human agent, such as looking up data, taking actions, making complex decisions, and writing personalized responses. This is the holy grail that everyone wants to get to. 2. Around it, the rest of the engine is designed to reinforce the AI agent and allow it to continuously improve. This starts with a mechanism (i.e. "Routing") that determines when the conversation should be escalated to a human in the loop. This is key because it allows you to roll out your AI agent incrementally. 3. Next, you have the AI tooling for your human agent to use that automates away mundane tasks, like drafting an answer, finding relevant information, polishing tone, etc. We call this "Agent Assist", and it's akin to a copilot. 4. Then, the conversations all feed into a central data platform, our "Admin Dashboard", that allows the leaders of the team to use LLMs to analyze the conversations. This will surface themes, trends, and anomalies in the data easily. It'll also identify gaps in your knowledge, for example, and proactively tell you how to fix them. 5. Finally, you need a way for human staff to "QA" the conversations to constantly give feedback. We've built this directly into the product. These components form the AI Agent Engine, a helpful framework for thinking about AI implementations. The full post written by Bihan Jiang, Kaylee George, and Cynthia Chen is below! 👇

Your company needs to start thinking about how to automate the predictable, but “humanate” the pivotal. It seems that most companies want to wire AI into everything without first asking which customer journeys should be run by machines and which deserve a human in the lead. If it’s AI Everywhere, it’s how you get fast, cheap, and broken in the moments that matter most to your customers. There’s a simple decision model to use. Look at four signals on every mission: 1. Emotion 2. Risk 3. Complexity 4. Novelty Then route into four lanes: 1. Automate 2. AI-led with human fallback 3. Human-led with AI assist 4. Human-only. Then measure and tune the framework weekly. Now you can make it real in org design, with job specs are the enforcement layer. Require skills like prompt design, escalation architecture, policy-as-code controls, and handoff choreography. Add new roles that keep the approach honest: Resolution Designer, Escalation Architect, Conversation QA, and Frontline Guides trained on copilots. Here’s a few examples: Password reset goes to automation. A stuck high-dollar wire for a first-time business client gets humanated within 90 seconds, with AI doing prep and wrap-up. Bereavement or foreclosure starts human-led while AI handles paperwork and status. If your job specs don’t name the skills to run them, the frame collapses into bots or phone trees. There’s AI or Human Loops become Doom Loops. In your business, what should be machine-to-machine vs human-to-human? #ai #futureofwork #hiring #customerexperience #journeymapping #customerjourney