Not an industry knowledge document — a configuration asset you import. Scenario skeletons, agent design patterns, evaluation sets and document structure, so you are not starting from “what should a support agent even look like?”
You supply the industry facts. We deliver structure and patterns; you fill in the business content. That is not a disclaimer — it is why this actually works in production.
Five parts. Each one exists to save you the weeks of guessing.
Every industry gets two agents: an advisory agent (no ticketing — answers and guides) and a service agent (with ticketing — handles exceptions and follow-up).
That “1 advisory + 1 service” shape is identical across all nine industries. It is a repeatable method, not nine unrelated examples.
Every agent defines four things explicitly:
Each agent ships with suggested questions and expected outcomes — 18 groups across the nine industries.
This is the part customers find hardest to build themselves: you have to decide what “a good answer” means before you can write a reusable test for it. After go-live, it becomes your quality check.
Three document types per industry: guide / policy / record snapshot.
Keeping “rules” and “snapshots” apart is the point — it tells the agent which content is standing policy and which is only sample data, so it never answers with a demo record as though it were live state.
From importing the configuration and binding a model profile, through connecting the knowledge base, to the first conversation — every step in order, so someone unfamiliar with the platform can still get through it.
We do not invent your business rules.
The reason is in the next section.
How many days a refund takes, what qualifies for an exchange, who can approve it — all of it differs per company and changes over time. A generic “industry template” stuffed into an agent just makes it answer the wrong thing confidently.
Your after-sales rules, SLAs and internal processes are business assets in their own right. In this architecture they live in your own knowledge base and are never sent to a third party.
What you receive is a structure. Swap the fact layer once and the pack becomes yours; when a policy changes you edit one place rather than buying anything again.
Seen the other way round, this is the value: a support agent structure you can maintain yourself beats an industry knowledge document that is out of date within six months.
These skeletons are ready — you can inspect and chat with them in the console.
| Industry | Advisory agent (no ticketing) | Service agent (with ticketing) | Skeleton |
|---|---|---|---|
| Retail | Product advisor | Order after-sales | Ready |
| Enterprise software (SaaS) | Product consulting | Technical support | Ready |
| Manufacturing | Equipment selection | Maintenance and repair | Ready |
| Education | Course consulting | Student services | Ready |
| Hospitality and travel | Trip planning | Booking and changes | Ready |
| Property | Listing consulting | Facility repairs | Ready |
| Logistics | Shipping and rates | Delivery exceptions | Ready |
| HR | Recruiting consulting | Employee services | Ready |
| Gaming | Game guides | Player support | Ready |
About the preloaded content — this needs saying plainly: the industry content currently shipped in the repository is fictional material written to demonstrate the structure and the patterns. Brand names, order numbers, prices and rules in it do not represent real business.
Its purpose is to show you what an industry pack looks like — not to supply rules you can put into production unchanged.
A production-ready version requires replacing the fact layer with your real material. You can do that yourself, or we can help.
Which industry, which classes of question, and whether ticketing and human handoff are involved. If one of the nine skeletons is close, adapting it is faster.
Swap the demonstration content for your real policies, processes and FAQs. You can prepare and import it yourself, or hand us existing material to organise — the latter is delivery work billed per person-day.
Import the agent configuration and knowledge base per the playbook, bind a model profile, and configure the business tools the agent needs to call.
Run the bundled evaluation set and see which questions are answered well and where material is missing. This step usually exposes gaps in the knowledge base and is one of the most valuable parts of the engagement.
Confirm behaviour in the console, then connect real channels using the production deployment.
Yes. Integrator licensing includes redistribution rights — deploy for clients within the licensed scope, present it under their brand, and start each vertical project from a pack rather than a blank page.
A pack earns its keep by shortening your first project in a vertical, not by becoming another document to maintain.
So treat it as a starting point to adapt, not a finished product to resell. Once adapted, that configuration is your own asset.
Quarterly updates to scenarios, evaluation sets and design patterns, kept in step with product releases. Because one update serves every customer in that industry, it can be priced far below bespoke work.
⚠️ Still being planned and validated. It depends on same-industry customers having rules similar enough that one update serves them all. We will offer it once that holds up.
Yes. Packs are organised by industry, and you can take only the service agent, or only the evaluation set and document structure. There is no all-or-nothing requirement.
Yes, under integrator licensing it can be presented under the client's brand.
Either. Doing it yourself keeps it accurate and editable at any time; handing it to us is delivery work billed per person-day — but even then, you still sign off the final rules.
The “1 advisory + 1 service” structure is general, and applies directly to a new industry. We can also build a new skeleton in the same pattern if you want one.
Agent configuration is tied to a product version — after a platform upgrade, older configuration may need adapting. A one-off purchase covers adaptation for the current version; later versions are covered by a content subscription or maintenance.
After import you can run it against the agent in bulk and check whether each answer meets expectations. It is for pre-launch acceptance and post-launch regression checks — not a benchmark for scoring models.
Tell us your industry and scenarios and we will send a matching sample — or bring all nine skeletons up yourself with one command and look at them directly.