What's inside a vertical pack, and why generic prompts lose demos

Every Lumina Sales demo runs on a vertical pack: the services, price bands, objections and booking language of one trade. Here is what a pack contains, how it changes a demo, and what it does not do for you.

Lumina8 min read
A fan of thin outlined cards on a near-black background, each etched with a different geometric line pattern, one card lifted forward and lit by warm amber light

A vertical pack is the trade-specific script that sits behind every Lumina Sales demo. It carries the services that trade sells, the price bands it quotes, the objections its customers raise, and the booking language its front desk uses. Pick "HVAC company" and the receptionist triages emergency versus maintenance. Pick "med spa" and it handles price questions using treatment bands without discounting. There are 100+ of them, and they are the reason a demo sounds like the prospect's business instead of a chatbot with the company name pasted in.

That last part is the whole game. A prospect decides in the first thirty seconds of a demo call whether the thing understands their business. Generic prompts lose that thirty seconds.

What is actually inside a vertical pack?

Four things, all written for one trade.

Services. The list of what that kind of business sells, in the words their customers use. A dental caller asks about a cleaning or a crown. A roofing caller asks about a leak, an inspection, or storm damage. The agent needs those nouns ready before the first question lands.

Price bands. Not a single made-up price, a range that fits how that trade quotes. When the caller on our own live demo asks how much PRP injection costs for a single joint, the agent answers in the regenerative clinic's language instead of stalling.

Objections. The pushback a real front desk hears every day in that trade. "Do you take my insurance." "Can you come today." "Why is it so much more than the other guy." A pack prepares the agent for the questions that decide whether a caller books.

Booking language. How that trade actually schedules. A law firm does intake. A plumber dispatches. A salon books a chair. The same "book an appointment" instruction sounds wrong in two of those three.

On top of that, each pack is written for the person who buys. The HVAC pack assumes an owner in the van who just missed a call. The law firm pack assumes a partner or intake lead and keeps the tone sober, not chirpy. The real estate pack assumes an agent who needs every portal enquiry answered in seconds and qualified on budget, timeline and pre-approval before the callback. You can see the four worked examples in the verticals section of the site.

Why not write one good prompt and swap the business name?

Because the voice vendors themselves tell you that is not how agents behave.

OpenAI's Realtime prompting guide recommends a system prompt with labeled sections for role, personality and tone, context, reference pronunciations, tools, rules, conversation flow, and escalation. It also says to "add domain-specific sections (e.g., Compliance, Brand Policy)". The same guide warns that the model follows sample phrases so closely it can overuse them and sound robotic. So the example phrases in a prompt are load-bearing. If they come from a telecom support script, your dental receptionist will sound like telecom support.

Vapi's Voice AI prompting guide is blunter. It lists six required sections, including company context and step-by-step playbooks for each scenario, and it tells the agent never to infer or make up prices, schedules, policies or discounts. That rule is the reason price bands have to live in the prompt. An agent with no pricing context either refuses every price question or guesses. Both lose the call.

Retell's situation guide goes down to how to say a phone number, an email address, a website and a time out loud. That is the level of detail it takes for a voice agent to sound competent. Now multiply it by the specific way a septic company, a fertility clinic, and a towing company each talk.

You can write all of that yourself. People do. It takes real hours per trade to get right, and you do not know it is right until you have run a lot of test calls. Vapi's own advice is to validate prompt changes "against a representative test set, not single calls." That is the work a pack saves you, repeated across 100+ trades.

How do you use a pack when you build a demo?

The build flow is short. In the demo builder you pick the trade, type in the prospect's business name and website, and arrange the nodes the way you pitch. The pack loads with the trade. From that point the receptionist, the chat agent and the follow-up all speak that trade's language.

A few things change when you switch trades:

  • The voice
  • The services it knows about
  • The objections it expects
  • The price bands it quotes

The rest of the page, the website audit, missed-call recovery, and the ROI slider seeded with the prospect's ticket and close rate, works the same way across trades. If you want the node-by-node breakdown, we covered it in inside the live demo site.

One practical habit. Before you send a demo to a prospect in a trade you have not sold before, call it yourself and ask the three questions you think their customers ask most. If the answers sound like their front desk, send it. If one sounds off, you have learned something about the trade before the prospect did, which is a good position for a sales call.

Which trades are covered?

More than 100, grouped roughly like this:

  • Health and clinics. Dental, oral surgery, med spa, chiropractic, physical therapy, veterinary, urgent care, optometry, dermatology, fertility, hearing, GLP-1 clinics and more.
  • Beauty and fitness. Salons, barbers, nail, lash and brow, day spas, massage, tattoo, yoga, martial arts, gyms.
  • Home services. HVAC, plumbing, electrical, roofing, landscaping, pest control, solar, garage doors, locksmiths, restoration, septic, foundation, tree service, pools, moving, junk removal and more.
  • Auto. Repair, body, tire, towing, dealerships, detailing.
  • Professional and property. Real estate teams, mortgage, insurance, property management, personal injury and family law, accounting, staffing, IT managed services.
  • Hospitality. Restaurants, catering, hotels, event venues, bakeries.
  • Industrial and logistics. Freight brokerage, trucking, heavy equipment rental, welding and fabrication, manufacturing plants, ports, rail terminals, warehouses and others.

The full list is on the site. It is long enough that you should scroll it before assuming your niche is missing.

What a pack does not do

This is the part worth reading twice.

A pack knows the trade, not the client. It gives the agent the services and price ranges that fit a typical business in that trade. It does not know that your prospect charges $189 for a diagnostic, closes at 4pm on Fridays, or refuses to service a certain brand. For a demo that is fine, since the point is to show how the system handles their callers. For a live build, those facts come from the client. That is what onboarding collects: hours and holiday cover, services with the price band for each, calendar, phone number, and who gets the leads. We walk through that sequence in what happens after your client signs.

A pack does not make a regulated trade simple. The dental pack sounds like a dental front desk. It does not make you compliant with anything. A dental practice is a HIPAA covered entity, and that changes your obligations once the build goes live. We covered what that means in how to sell AI systems to dental practices. The same caution applies to law firms, where the pack screens for facts without giving advice, and to any clinic.

The generic pack is a real fallback, not a disguise. Anything appointment-based that is not on the list gets a generic pack written to read as industry-neutral on purpose, rather than a template with the trade left blank. It works. It will not sound as native as a trade pack, and a sharp prospect in a niche field may notice. If you sell into that niche more than once, ask us for a pack. Niche packs are written by hand, and subscriber requests get built first.

Some trades on the list are harder sells than the pack makes them look. A chemical plant or a power-generation facility has a pack, and the pack is good. But those buyers do not decide on a phone demo the way a plumber does. The pack gets you a credible demo. It does not shorten a procurement cycle.

Who gets the most out of the packs?

Reps and small agencies who sell across several trades. If you only ever sell to roofers, you could hand-tune one prompt and be fine. The packs pay off when this week's calls include a med spa, a garage door company and a property manager, and you need all three demos to sound like they were built for them. That is most people selling AI into local service businesses, because you go where the replies come from.

FAQ

Can I edit what a pack says? The pack sets the trade defaults, and your build adds the prospect's business name and website. Treat the pack as the starting point for a demo, and treat the client's real details as intake for the live build, which runs through onboarding rather than through a prompt edit at midnight.

What if my prospect's trade is not listed? Use the generic pack, which is written for any appointment-based business. If the trade comes up again, request a dedicated pack. Subscriber requests get built first.

Do packs change the voice? Yes. When you switch trades on the demo, the voice, services and objections change with it. Try it on the live demo by switching between the five trades on the page.

Does using more packs cost more? No. All 100+ are included in the $49 a month plan, along with 100 live voice minutes, 500 chat messages and 50 website audits a month. You can keep unlimited drafts and one published demo at a time.

Will the prospect see that it is a template? They see their business name, your branding, and a trade-specific conversation. The one thing always visible is the sandbox disclosure that says the AI is a demo and bookings in it are simulated.

Try the pack for your next call

Pick the trade of the next prospect on your list. Open the live demo, switch to the closest trade, and ask it the question that prospect's customers ask most. If the answer sounds right, start a workspace and build their demo before the call. The site says the price goes up on October 1.