AI and profit glossary
The terms below are the ones we use with clients every week. Most are not technical - they are the vocabulary of margin, follow-up speed and automation. If you have ever nodded along to one of them without being sure what it meant, this page is for you.
Money and marginLeads and follow-upAI and automationOperations and compliance
Money and margin
Profit leak
- Money already inside your business that escapes before it reaches the bottom line - underpriced work, leads that go cold, manual hours, unattributed ad spend. A profit leak is not a missed opportunity. It is money you already earned and are not keeping.
Gross margin
- Revenue minus the direct cost of delivering the work, as a percentage. At 11% you are busy. At 30% you are profitable. Most businesses at $500K+ are running closer to the first number than the second.
Contribution margin
- What a job, product or channel contributes after its own variable costs, before overhead. The number that tells you whether a channel is worth scaling or quietly losing money.
Margin by client
- Profit per client, not revenue per client. Businesses regularly discover their largest client is their least profitable one.
Cost to acquire a customer (CAC)
- Total sales and marketing spend divided by new customers won. Without it, budget gets reallocated on instinct and the losing channels keep getting funded.
Customer lifetime value (LTV)
- The total profit a customer produces over the whole relationship. The ratio between LTV and CAC decides how much you can afford to spend to win one.
Ghost software
- Subscriptions you pay for that nobody uses or remembers buying. The average business at $500K+ revenue carries $500 to $2,000 a month of it.
Pricing architecture
- How prices are structured across tiers, bundles, thresholds and discounts. Small changes here often move profit more than a new campaign, because the work is already being delivered.
Scope creep
- Work quietly expanding past what was priced. The most common reason margin falls while revenue rises.
Baseline
- A documented record of revenue, profit, spend and conversion rates taken before work starts. The baseline is what settles any argument about what existed before and what was created.
Leads and follow-up
Speed to lead
- How quickly a business responds to a new inquiry. Responding inside 60 seconds can multiply contact rates several times over compared with responding in an hour.
Lead response time
- The clock from an inquiry arriving to the first reply. Most businesses treat it as a feeling. Measure it in seconds.
Missed-call text-back
- An automatic text to anyone whose call goes unanswered, usually inside 30 seconds. In high-value service industries this is often the highest-return automation there is.
Follow-up sequence
- A scheduled series of touches after an inquiry or quote. Most businesses stop after one attempt, and most sales happen after several.
Aged lead
- An old inquiry or quote that never converted. Cheap to work because you already paid to acquire it.
Book reactivation
- A one-time campaign against existing clients and aged leads. Usually run at the start of an engagement because it produces cash fastest.
Appointment setter
- The person or automation responsible for turning a lead into a booked slot.
No-show recovery
- Reminders before an appointment and an easy way to reschedule after a missed one.
Conversion rate
- The share of leads or quotes that become customers. After follow-up speed, the fastest lever most businesses have.
Pipeline stage
- A defined step a lead moves through. Undefined stages mean the pipeline cannot be measured, and what cannot be measured cannot be fixed.
AI and automation
AI agent
- A system that takes a sequence of actions toward a goal - checking a calendar, drafting a reply, booking a slot - rather than answering one question.
Workflow automation
- Connecting tools so a trigger in one produces an action in another. The backbone of almost every useful AI system in a small business.
Prompt
- The instruction given to an AI model. Specific prompts carrying business context outperform clever ones.
Context
- The background an AI model is given: your pricing, your tone, your process. Five minutes of context beats years of prompt tinkering.
LLM (large language model)
- The class of model behind ChatGPT, Claude and Gemini. Strong at language and reasoning, unreliable on facts it has not been given.
Hallucination
- When a model states something confidently and wrongly. The reason a person should review anything that reaches a customer.
Retrieval-augmented generation (RAG)
- Feeding a model your own documents at the moment of the question, so answers come from your material rather than its training data.
Human in the loop
- A person checking the work before it goes out. The difference between AI that helps and AI that embarrasses you.
AI search visibility
- How likely AI assistants and AI overviews are to mention your business when someone asks a relevant question. Also called AEO, for answer engine optimisation.
AI crawler
- A bot that reads your site to train or ground a model - GPTBot, ClaudeBot, PerplexityBot and others. What you allow in robots.txt decides whether you are visible to them at all.
Structured data
- Machine-readable markup that states plainly what a page is, what it costs and who it is for. The most reliable way to be quoted correctly.
llms.txt
- A plain-text summary of a site aimed at AI systems: what the business does, what it charges, which pages matter.
Operations and compliance
Systems thinking
- Designing a process so it produces the same result no matter who runs it. The opposite of relying on memory and goodwill.
SOP (standard operating procedure)
- A written, step-by-step description of how a task is done. What makes work delegable rather than personal.
Automation audit
- A ranked list of repeated tasks by hours multiplied by wage, so the most expensive manual work gets automated first.
Weekly numbers
- A short fixed set of metrics reviewed on the same day each week - leads, close rate, revenue, spend, margin. Visibility is what stops the other leaks hiding.
TCPA
- The US law governing automated calls and texts. Requires prior express consent, quiet hours and working opt-out handling.
10DLC
- The carrier registration system that lets a business send application-to-person texts in the US. Approval takes days to weeks, so start before you need it.