Last updated: 3 November 2026
At QD Academy we use AI for three jobs: triaging and summarising email, tracking marketing performance, and following up with warm leads. We keep people in charge of anything involving money, anything a lead has asked a person to handle, and anything the AI does not have enough context to answer. This article walks through what we automated, what changed, and one mistake that got our WhatsApp number restricted.
What have we automated at QD Academy?
These are the workflows we run today. Each one started with a task that repeated every week and took more time than it should.
Email triage and summary
We use Claude with Gmail to sort incoming email and produce a short summary of what needs attention. Instead of opening every message to find out what it is, the team starts the day with a summary and opens the emails that need a decision.
Performance tracking
We use Claude with Google Analytics to turn traffic and conversion data into a readable summary. The weekly question we used to answer by clicking through reports is now answered in plain language, and we spend the saved time acting on it.
Warm lead follow-up with an AI agent
This is the biggest one. We built an AI agent inside our CRM, powered by ChatGPT and a knowledge base we maintain about our courses. It follows up with warm leads by message, raises engagement, answers their questions, and keeps the conversation moving toward sign-up when a lead is clearly ready. If the agent lacks the context to answer, or a lead says they want to speak to our staff, it notifies a human team member who takes over. If you are new to the idea, our guide on what an AI agent is explains the basics.
What results did we see?
We will keep the numbers general, because every business has a different starting point. In our own tracking, the team saves hours of work every week. Reply rate and paid conversion rate both went up by more than 50%. The lead follow-up agent drove most of that change, since leads now get a prompt, relevant reply instead of waiting for a free person.
What do we keep human on purpose?
AI handles the repetitive first pass. People handle the moments where a mistake costs trust or money. In our case that means:
- Important payment notices. Anything that tells a customer what they owe, or confirms a payment, is sent and checked by a person.
- Conversations a lead has asked a person to take. When someone says they want to talk to our staff, the agent steps back.
- Questions the knowledge base cannot answer. The agent hands these over instead of guessing.
- Grant eligibility and claims. Funding rules depend on the company and the participant, so our team checks each case.
- Training delivery. Our courses are taught by working practitioners, and that part is not something we hand to software.
The pattern behind these choices is simple. If the AI is wrong, how bad is it, and can the customer tell? Where the answer is serious, a person does the work.
What did we try that did not work?
Our first approach to follow-up was to message the phone numbers that people left on interest forms at our events. We sent the first message ourselves, and WhatsApp restricted our number for a while. Messaging people who have not contacted you first is the kind of activity messaging platforms limit, and it also feels intrusive to the person receiving it.
We changed the flow. After someone fills in an interest form at one of our events, they now see a link that opens a chat with us, so the lead sends the first message. Once they have done that, our follow-up is a reply to someone who asked, and the AI agent can pick up from there. Reply rates improved, and we stopped risking the number we rely on.
What can your team copy from this?
You do not need a CRM or a custom agent to start. Pick one repeated task, write down the steps, and test an AI tool on it. Our task audit guide shows how to choose, and building your first AI agent without coding walks through a first workflow. To keep it going after the first few weeks, read why SMEs stop using AI tools. If you would like to build your own workflows with a practitioner beside you, see the AI-Powered Productivity course for the latest schedule and funding details.
Frequently asked questions
What AI tools does QD Academy use internally?
We use Claude with Gmail for email triage and summaries, Claude with Google Analytics for performance tracking, and an AI agent in our CRM that uses ChatGPT and a knowledge base for lead follow-up.
Does the AI agent talk to leads without a person checking each message?
Yes, for routine follow-up and questions it can answer from the knowledge base. When it lacks context, or a lead asks for our staff, it notifies a team member who takes over.
Why did WhatsApp restrict your number?
We had been sending first messages to phone numbers collected from interest forms at our events, and the number was restricted for a while. We now give leads a link to start the chat themselves, then follow up.
What results did QD Academy see from using AI?
In our own tracking the team saves hours of work each week, and reply rate and paid conversion rate both rose by more than 50%. Your results will depend on your starting point and process.
Do we need a CRM to automate lead follow-up?
A CRM makes it easier to connect leads, messages and a knowledge base in one place, but you can start smaller. Write down your follow-up steps, test an AI tool on a few real enquiries, and add automation once the process works.