What is this course about?
Small businesses compete on relationship, so support automation that sounds generic destroys the advantage it was meant to protect. This course covers capturing your actual voice from real conversations, grounding on your real policies, and designing escalation for the common case where the human is you, on a job site.
Who is this course for?
- Owner-operators handling their own support
- Small support teams of one to five
- Service business operators
- E-commerce and retail owners
What do I need before starting?
- Access to your past customer conversations
- No technical background required
What will I be able to do afterwards?
- Classify support questions as safe to automate or relationship-critical
- Capture your actual voice rather than a generic support register
- Ground responses in your real policies rather than plausible ones
- Design escalation when the only human available is you
- Measure resolution rather than deflection
What does each module cover?
Which questions are safe to automate?
35 minThe triage that separates routine information from the conversations that keep customers.
Objectives
- Classify your real support volume by automation safety
- Identify the relationship-critical conversations
- Set the automation boundary explicitly
Topics
Activity. Sort your last 100 support conversations into automate, assist, and never.
How do you capture your actual voice?
45 minDeriving tone from your own past responses instead of accepting a generic support persona.
Objectives
- Extract voice characteristics from your real responses
- Encode voice so it survives across topics
- Test whether customers can tell the difference
Topics
Activity. Build a voice profile from 50 of your own past responses and test it blind.
How do you ground on your real policies?
45 minReturns, warranty, hours, and service area — the facts an agent must never invent.
Objectives
- Document policies precisely enough to ground on
- Prevent the agent inventing plausible policy
- Handle the questions your policy does not cover
Topics
Activity. Document your policies and test the agent on twenty edge-case questions.
How do you escalate when the human is you, on a roof?
40 minEscalation design for a business where the expert is not at a desk.
Objectives
- Design escalation that respects your actual availability
- Set honest expectations with the customer
- Batch escalations so they do not interrupt work
Topics
Activity. Design your escalation path around a real working day.
How do you cover after hours honestly?
35 minOut-of-hours coverage that helps without implying a response that will not come.
Objectives
- Set after-hours expectations honestly
- Handle genuine emergencies differently
- Queue non-urgent items for the morning
Topics
Activity. Build the after-hours flow and test it with an emergency and a routine question.
What do you do with an angry customer?
40 minThe case for immediate human routing, and why an agent making it worse is expensive.
Objectives
- Detect frustration reliably
- Route to a human immediately rather than attempting recovery
- Prevent the agent escalating the emotion
Topics
Activity. Test your agent with five genuinely angry messages and check the routing.
How do you measure resolution, not deflection?
35 minThe metric distinction that separates a support agent from a customer-avoidance system.
Objectives
- Measure whether the customer's problem was solved
- Detect deflection dressed as resolution
- Track repeat contact as a quality signal
Topics
Activity. Instrument resolution and repeat contact, then review a week of data.
Training the agent on your real conversations
50 minThe build module: an agent grounded in your last 200 conversations, tested blind.
Objectives
- Ground the agent in real historical conversations
- Run a blind test with real customers or staff
- Tune based on what the blind test reveals
Topics
Activity. Build the agent and run a blind test where a customer cannot tell which responses were yours.
What is the capstone project?
Support agent with a blind voice test
Build a support agent grounded in your real conversations and policies, with availability-aware escalation and frustration routing, then run a blind test where reviewers judge which responses were written by you.
Deliverable: A deployed support agent plus blind test results and resolution measurements.
How are learners assessed?
- Blind voice test — reviewers should struggle to identify the agent's responses
- Policy edge cases must produce abstention rather than invention
- Angry-message routing verified across five real examples
What ships with the course?
Facilitator guide
Session-by-session running order, discussion prompts, and the questions that reliably derail a room.
Learner workbook
Exercises, checklists, and the templates each module's activity produces.
Hands-on lab environment
A sandboxed ibl.ai deployment so exercises run against real agents, not screenshots.
Assessment bank
Scenario questions and rubric criteria mapped to each stated learning outcome.
Source bibliography
Every primary regulation and standard cited on this page, linked and dated.
Which AI agents does this course use?
The hands-on modules run against agents already deployable on the ibl.ai platform for small business.
Where does the course material come from?
Every module is grounded in primary sources — the regulation, standard, or research itself, not a summary of it. Each was resolved at authoring time.
- U.S. Small Business Administration
SBA
Customer service and operations guidance for small firms.
- Business guidance
Federal Trade Commission
Disclosure obligations when a customer is interacting with automation.
- Small Business Cybersecurity Corner
NIST
Handling customer data safely in an automated support flow.
- California Consumer Privacy Act
California Attorney General
State privacy obligations that reach small businesses handling customer records.
Delivery notes
Binding guidance for anyone preparing and delivering this course.
- The blind voice test in Module 8 is the course's signature assessment. Run it with real customers if the owner is willing — staff are too familiar with the owner's writing to be a fair test.
- Module 3's abstention behavior matters more than coverage. An agent that says 'let me check with the owner' is fine; one that invents a return policy creates a legal problem.
- Module 6 should be conservative. For a small business, one badly handled angry customer becomes a public review, and the cost asymmetry justifies routing aggressively.
- Keep disclosure honest. Whether to tell customers they are talking to AI is a genuine judgment call; present both positions and the FTC guidance rather than prescribing.
- Coordinate with SB-7 — support quality and review outcomes are the same system, and the two courses should share the reputation measurement.
Why run AI training on a platform you own?
You own the course, not a licence to it
Course content, learner data, and the platform run inside your perimeter — you own all the code and the data.
Model-agnostic delivery
Run the course's AI components on any LLM — Claude, GPT, Llama, Gemini, Command — and switch anytime.
No per-seat training licences
Usage-based or self-hosted, so cost tracks actual use rather than headcount.
Deploy anywhere
Cloud, private VPC, on-premise, or fully air-gapped — including for cohorts that cannot use public AI tools.
Frequently asked questions
What does the AI Customer Support That Does Not Sound Like a Robot course cover?
Small businesses compete on relationship, so support automation that sounds generic destroys the advantage it was meant to protect. This course covers capturing your actual voice from real conversations, grounding on your real policies, and designing escalation for the common case where the human is you, on a job site. It runs 4 hours across 8 modules across 8 modules, at foundational level, and closes with a capstone: Support agent with a blind voice test.
Who should take AI Customer Support That Does Not Sound Like a Robot?
It is written for Owner-operators handling their own support, Small support teams of one to five, Service business operators, E-commerce and retail owners. Prerequisites: Access to your past customer conversations; No technical background required.
Can we run this course on our own infrastructure?
Yes. ibl.ai is model-agnostic and deploy-anywhere — cloud, private VPC, on-premise, or fully air-gapped — and you own all the code and the data. Cohort data, submissions, and any material learners upload stay inside your perimeter, which matters for small business teams that cannot send work to a public AI tool.
How do we get access to AI Customer Support That Does Not Sound Like a Robot?
Request access and we will set it up for your cohort — hosted by ibl.ai, or running against your own deployment. Tell us the group size and timing you need, and whether it should run inside your own perimeter.
How much does AI training for small business cost on ibl.ai?
There is no per-seat pricing — you pay for usage or self-host and pay only for the infrastructure, so a 5,000-person rollout does not cost 5,000 licences. 1.6M+ users across 400+ organizations run the platform this way, including NVIDIA, MIT, and Syracuse University.