Free Homework Helper Chatbot: Triage Questions, Then Tutor

The Homework Helper is a subject-triage and escalation flow for anyone who answers student questions at scale: a tutoring company, a school's learning support team, an edtech product, or a university help desk. Its job is to take a vague "I'm stuck on my homework" and turn it into a structured, answerable request — subject, topic, and the specific problem — before either answering it or putting a human tutor on the case.

That framing matters, because the template is honest about where automation stops. Two nodes in the shipped flow carry explicit placeholders: the answer node and the study-tips node. They are marked for you to fill with your own content or wire to an AI model. The template supplies the structure, the branching and the escalation logic; you supply the subject-matter answers. What you get out of the box is the part that is genuinely tedious to build — eighteen connected nodes covering five subjects, a satisfaction check, a tutor handover, and a self-service menu that loops rather than dead-ends.

The flow is built in the chatbot builder, so every question, branch and message is editable text rather than code.

1. Welcome. "Welcome to the Homework Helper! I'm here to assist you with your assignments and answer any questions you might have." Short, warm, and aimed at a student rather than a buyer.

2. Subject choice. A five-way branch — Math, Science, English, History, Other — stored in selected_subject. This is the one decision that changes the rest of the conversation, and it is deliberately first.

3. Topic narrowing. Each subject leads to its own free-text question with a subject-appropriate prompt and its own variable: Math asks "What specific math topic are you working on? (e.g., algebra, geometry, calculus)" into math_topic; Science asks about biology, chemistry or physics into science_topic; English asks whether it is an essay, grammar or literature into english_topic; History asks which period or event into history_topic; and Other simply asks the student to name the subject into other_subject. Five branches, five variables — which means your logs tell you exactly which topics generate demand.

4. The actual question. All five branches converge on one free-text node: "What specific question or problem do you need help with?", stored as student_question. By this point you have a triaged, well-formed request rather than a one-line cry for help.

5. Acknowledgement and answer. A holding message ("I'm processing your question and will provide an answer or guidance shortly") followed by the answer node, which ships with a placeholder. This is the node you connect to an AI model or replace with your own curated responses.

6. The satisfaction check. "Was this answer helpful?" — yes or no. This is the most valuable node in the template, because it is the difference between a bot that looks helpful and one that actually is.

7a. Helpful. The bot says it is glad it could help and moves to a four-option menu: ask another question (loops back to the subject choice), get study tips, find additional resources, or end the session. The tips and resources nodes ship as placeholders and return to the same menu, so a student can browse without restarting.

7b. Not helpful. The bot apologises, offers to approach the problem differently, and then hands over to a human tutor. The handover node is configured with medium priority, automatic agent assignment, a five-minute maximum wait, a waiting message ("A tutor will be with you shortly"), and a fallback that tells the student to try again shortly or email support if every tutor is busy.

8. Close. Both the handover path and the end-session option land on the same sign-off: "Thank you for using the Homework Helper! Good luck with your studies, and don't hesitate to return if you need more assistance."

The clearest fit is a tutoring business that wants to answer the easy questions automatically and route the hard ones to a paid tutor. The satisfaction check does the sorting for you, and the handover node is already tagged so those conversations are identifiable in your inbox.

Schools and colleges can run it as a first line for learning support, absorbing the volume of routine questions that otherwise queue at a teacher's desk. Edtech products can bolt it on as in-app help, using the subject and topic variables to understand which parts of a curriculum generate the most confusion. Libraries and community learning centres get an after-hours helper that never claims to be more than it is.

It is not a fit if you need graded assessment or progress tracking across sessions — the flow treats each conversation independently. For assessment, look at the Skill Assessment Quiz. For longer-horizon planning, the Study Schedule Planner and Exam Preparation Guide cover scheduling and exam readiness, and the tutoring service chatbot covers booking and enquiry capture.

1. Decide how the answer node answers

This is the whole project. You have three realistic options.

The AI route connects the answer node to a model, passing the subject, topic and question variables as context. Conferbot supports 21 AI providers including OpenAI GPT, Anthropic Claude, Google Gemini and DeepSeek, so you are not locked to one vendor. Write the system prompt to explain rather than solve — "guide the student toward the answer with a worked method, do not just state the result" — because a bot that does the homework is a liability, not a feature.

The knowledge base route grounds answers in your own material: worked examples, course notes, past FAQ answers. This is slower to set up and far more defensible, because every answer traces back to something your educators wrote.

The curated route replaces the single answer node with a branch of prewritten responses for your twenty most common questions, and sends everything else straight to handover. Unglamorous, but it is the fastest path to a bot that is never wrong.

2. Match the subjects to your actual curriculum

Math / Science / English / History / Other is a generic starting point. A GCSE revision service, a coding bootcamp and a nursing school need entirely different lists. Edit the subject choice node and the five topic questions behind it together — the topic prompt's examples ("e.g., algebra, geometry, calculus") should match the subject you renamed it to.

3. Staff the handover, or change what it promises

The handover node is set to a five-minute maximum wait. Live chat handover and the unified inbox are included on every Conferbot plan, including Free, so the capability is there — but a queue with nobody in it is worse than no queue. Either roster tutors against your busy hours, or lengthen the wait and rewrite the fallback message to set an honest expectation. Replace the placeholder support email while you are in there.

4. Fill the tips and resources nodes

Both ship as placeholders. Resources are the easier win: link your own worked-example library, revision guides or recommended reading. Because they sit behind the post-answer menu, they reach students at the exact moment they are receptive.

5. Log the questions somewhere

Every conversation produces a subject, a topic and a question. That is a curriculum diagnostic. Push it to Google Sheets or Airtable for weekly review, into Notion if your teaching team works there, or to your own systems via webhook (paid plans). After a term you will know precisely which topics your teaching under-serves.

Ask for the work, not just the question. Add a file-upload node after the question so students can attach a photo of the problem or their working. It removes most of the ambiguity that makes text-only homework help frustrating.

Capture the year group. One choice node asking which year or level the student is in lets you calibrate every answer. "Explain quadratics" means different things to a fourteen-year-old and an undergraduate.

Escalate on repeat, not just on dissatisfaction. Count how many times a student loops through the menu and offer a tutor after the third question, before they get frustrated enough to say no.

Add an academic-integrity line. A short message after the subject choice — that the helper explains methods and does not complete assignments — protects your institution and sets the right tone. Some schools will not deploy this kind of tool without it.

Book instead of queueing. Replace or supplement the handover with a calendar booking step so a student who needs real teaching gets a slot rather than a wait. This is also the natural conversion point for a paid tutoring business.

Offer a follow-up by email. An optional email capture at the end lets you send the worked solution and a related practice set, turning a one-off interaction into a returning user.

Split the "Other" branch. Look at what students actually type into other_subject after a month and promote the top two answers into real subject options.

Conferbot does not gate channels by plan, so this runs wherever your students are. The website widget is the default — put it on your learning portal, resource pages and course pages rather than only the homepage, so it appears where confusion happens.

WhatsApp is the strongest second channel for homework help in most regions, because it is where students already message about school and it works on a low-end phone. Discord is worth serious consideration for university and bootcamp audiences whose study communities already live there. Microsoft Teams fits institutions running Microsoft 365, and Telegram, Instagram, Messenger, LINE and the mobile SDK are all available on any plan.

The free plan includes 600 conversations a month and one chatbot, which is enough for a department-level pilot across a term. Paid plans start at $19/month and add webhooks, API access and removal of Conferbot branding; A/B testing arrives on Pro, which is how you would compare two answer strategies properly. See pricing for the full comparison.

Three limits worth designing around in 2026.

It is not a grader. The flow has no marking logic and no gradebook. It answers questions and escalates; it does not assess. Claiming otherwise in your copy will disappoint the first teacher who tries it.

It is only as accurate as what you connect. The answer node ships empty. If you point it at a general-purpose model with no grounding, it will occasionally be confidently wrong about a maths step, and students are poorly placed to notice. Ground it, or curate it, or keep the handover prominent.

It should not do the assignment. Build the prompt and the copy around explanation, not completion. The satisfaction check gives you a live signal for whether that line is working: a rising rate of "no, I need more help" on one subject usually means the explanations are too thin, not that students want answers handed over.