AI agent Amy provides round‑the‑clock consultations to clients of the language center via WhatsApp: she identifies their needs, selects a course according to the current schedule, and forwards the already prepared application to the manager.
84% of customers continue the conversation with the agent, and 7% have reached the payment stage.
About company
Headway Institute is a language center in Dubai. The company teaches adults and children, conducts group and individual classes, works with schools and corporate clients, and attracts international students.
Sales involve a lot of incoming inquiries: a person might write in the evening or at night, asking about a specific language, classes for a child, a schedule, or an appropriate level.
The response “we’ll be in touch soon” is not enough. It’s important for the client to immediately understand which course is right for them, when it starts, how much it costs, and whether they can study at a convenient time.
Problem: the client is waiting for the manager to become available.
Before the introduction of AI, the first response took an average of 2,3 hours.
This was particularly noticeable outside of working hours: only 15.6% of nighttime inquiries received a response within an hour.
At the same time, Headway has already tried to automate the first contact using Salesbot.
How Salesbot worked
He asked several questions according to a pre‑set scenario.
- Who are you looking for a course for — for yourself or for someone else?
- Adult or child?
- Online or in person?
- Is it convenient to come to our location?
After that, the application was passed on to the manager.
However, it never came to a full‑fledged consultation: the bot collected basic data but couldn’t select and propose a course to the client on its own.
Therefore, a significant portion of the work still fell to the manager.
It was necessary to read the answers, ask additional questions, figure out the schedule, choose a course, and only after that proceed to the sale.
Solution: instead of an answering machine, we have an AI consultant named Amy.
Headway, together with TextBack, launched an AI agent named Amy.
The goal was specific: to pass on to Amy the first stage of the manager’s work.
The agent was supposed to conduct an initial consultation and prepare an application.
When a new request arrives, the agent immediately begins the consultation: they understand what the client needs, ask clarifying questions, refer to the latest Headway data, and suggest an appropriate option.
Before Amy, Headway tested a voice bot, but it performed worse when dealing with deviations from the script. According to the team’s observations, the text agent confidently handles a wider range of requests.
How the consultation is structured
The scenario depends on the client.
For an adult, eight mandatory questions are provided; for a child, three; for a corporate request, one or two. A separate branch is used for the summer camp.
Amy can clarify the language, the format of the classes, the current level, the learning goal, a convenient time, and other parameters.
The main difference from the old Salesbot becomes apparent further on.
Amy uses the collected responses to select an appropriate course.
For example, the client might not choose an option from a predefined menu, but simply write:
«I want to learn Spanish for travel»
Amy understands the request, clarifies the level and preferences, and then looks for a suitable option among the current programs.
Amy works with the live schedule at Headway.
This was a mandatory requirement of the project.
The Headway schedule is constantly changing: new groups appear, some groups start, are rescheduled, or are closed. Data for adult courses is updated daily.
At the time of the upload, there were approximately 45 relevant courses in the schedule.
Amy refers to this data during the conversation and can provide the client with specific course parameters:
the start date, schedule, teacher, duration of classes, and price.
The difference from a classic chatbot is simple here.
Salesbot knows in advance:
“If you pressed A → ask B.”
Amy is solving a different problem:
“Here’s what the client wants → find something suitable among the current options.”
24/7 and in the client’s language
Headway works with an international audience, so some clients write in languages other than English.
Amy can continue the conversation in the client’s language. The project materials contain examples of switching between English, Russian, and Arabic.
For Headway, this turned out to be an important part of the product: the team specifically highlights the ability to adapt to the user’s language as an advantage of the text‑based agent.
At the same time, the agent is working. 24/7.
Here, we obtained the most notable result of the project.
Average time to first response
It was: 2.3 hours.
It’s been: 6 seconds
Nighttime inquiries with a response within an hour.
It was: 15.6%
It’s been: 96,9%
That is, the client no longer needs to wait for the start of the workday to at least start figuring out which course to choose.
Where does Amy’s work end
We deliberately did not try to automate the entire sales process.
Amy does not accept payment and does not schedule a trial lesson for the client. These actions are still left to the person.
Amy takes the pre‑sale as far as the dialogue allows, and when a person is needed, she passes the gathered context on to the manager.
At the end, Amy collects the client’s responses in summary. The client themselves can see that the agent understood them correctly, and then they receive a message stating that the request has been passed on to the manager. The Headway team notes that this removes uncertainty: the client understands that their request has been received and they will be contacted further.
The manager no longer has to start the conversation from scratch.
He connects with a person who understands the need, format, level, schedule, and other parameters.
Amy is responsible for the quick presale and preparation of the request. The person is responsible for the deal.
What came of it
The conversion of the agent’s dialogues turned out to be as follows:
100% — entered into a dialogue
84% — continued the conversation with Amy
9% — Amy picked the course.
If we look at conversions in terms of successful actions:
17% — transferred to the manager
7% — have reached the payment stage
Here, it’s important not to attribute too much to AI: we didn’t compare the conversion rate to payment before and after implementation, so we can’t claim that the agent increased sales by 7%.
But something else is visible.
AI doesn’t work like a FAQ bot on the periphery of the process. Customers go through a real sales funnel with it and eventually reach the payment stage.
The most noticeable measurable change is at the beginning of the funnel:
2.3 hours → 6 seconds until the first response.
Without magic from day one
After the launch, Amy had to be refined. For example, at the beginning, the agent sometimes couldn’t find a course that was actually in the table. There were problems with determining the level and switching the language.
The Headway team identified such dialogues and passed them on to TextBack. We corrected the errors and gradually stabilized the agent’s performance.
There was also a less serious case. One of the clients realized he was talking to a bot and started making jokes. Amy picked up the conversation and responded in the same style — the Headway team remembered this dialogue separately after the launch.
What has changed for Headway
Previously, the automation ended before the full consultation.
Salesbot could collect several responses. After that, a manager was still needed.
Now, the first stage of the sale begins immediately after the customer contacts us — including at night.
The client receives a response within 6 seconds.
Amy is identifying the need.
Finds a suitable course in the current schedule.
It is collecting the application.
When a person is needed, Amy passes the already assembled context to the manager.
According to the Headway team, managers find working with this scheme more convenient, and the deal cycle has become shorter.
The next logical step is to gradually hand over the following stages of the funnel to the agent: for example, recording the client and sending the payment link. Headway is already considering this possibility, although for now these actions are deliberately left to the manager.
Do you want to launch an AI agent that responds to customers 24/7 and helps convert inquiries into sales?
Get 3 days of access to the platform for free. Or submit an application — we’ll help you launch an AI agent on WhatsApp, Telegram, or MAX and integrate it with CRM, just like at Headway Institute.
