← Back to the deck

Try EngineIn

EngineIn is a chatbot that runs a structured intake conversation, then generates a summary report, routes the conversation to the right person, drafts a follow-up email, and offers a meeting time. Everything it does is shaped by a single config file — there's no hard-coded behavior. This page is the demo's front door.

Quick start

1. Open the demo. The link below carries an access token; treat it as the way in.
2. Have a chat. Pretend you have a project you need help with. The bot will ask the things a real intake would ask — name, company, project type, location, scope, timeline, budget. Be specific or vague; both are interesting.
3. Finish. When you've answered the core questions, the bot wraps up: it picks the right team member from the directory, scores the lead, drafts a follow-up email to you, and offers a meeting slot.

Open the demo →

Now switch hats.

See the admin view

The same app has an internal panel — it's how the people receiving these intakes triage, configure, and review them.

1. Click the lock icon at the top-right of the chat window.
2. Enter PIN 6107 (super-admin — you'll see everything).
3. The panel opens on the right.

Worth poking at

• Sessions tab — every chat that's run through the demo, including yours. Click one to see the full conversation transcript, the generated .docx lead summary, the .ics calendar invite, and the routing decision the bot made.
• Prompt tab — the live system prompt the bot is using right now. You can see exactly how it's been instructed.
• Workflow tab — the report sections that get generated, plus the notification routing.
• Builder tab — read on.

How the builder works

The builder is the heart of the system. It's a structured editor over a single config file, and that config file is the bot's behavior — there's no separate codebase to change when you want a different intake experience.

A few things you can configure, and what changes when you do:

• Company identity. Name, tagline, intro greeting, contact info, brand colors. These show up in the chat UI, the system prompt the AI sees, and on every generated document.
• Project types and questions. A list of disciplines (e.g. structural, civil, building envelope), each with its own follow-up questions. When the bot detects which discipline the conversation is about, it asks the matching questions. Adding a new discipline = adding an entry, no code change.
• Team directory. Names, roles, regions, specialties, and per-person keyword lists. The keywords drive routing: when the conversation contains words a team member's keyword list mentions, that member becomes the recommended contact. The bot can revise the recommendation as the conversation evolves.
• Offices and regions. Used to match a project location to the office that covers it. Multi-office orgs get correct routing without writing logic for it.
• Scheduling. Meeting types (phone call, video call, on-site visit) with durations, plus business-hours rules. The "offer a meeting" step at the end of the chat reads from this directly.
• Lead scoring. A weighted set of factors (project size, urgency, budget clarity, fit, client profile) that produce the lead score on the summary card.
• Notification routing. Which channels (Teams, email, SMS, Slack, generic webhook) fire when a lead lands, with priority thresholds and digest schedules.
• Theme. Colors, fonts, bubble shape. The chat UI re-skins live as you edit.

The trick: the system prompt the AI sees is generated from this config every time the bot starts a conversation. That means a config change isn't just metadata — it changes what the AI knows and how it talks. Add a team member with a new specialty, and within seconds the bot is offering them as a routing target. Change the intro greeting, and the very next visitor sees the new one. There's no "deploy" step.

In the demo, the Builder tab is read-only — you can browse the structure but the save buttons return an error. Look around to see how it's organized; that organization is the product surface.

What's deliberately disabled

This is a public demo, so anything that mutates real config or fires real outbound notifications returns a friendly 403. You can read everything; you can't change anything. If a button in the admin doesn't work, that's by design.

Privacy & reset

Anything you type briefly passes through Anthropic's API and is stored only for the demo session itself:

• Your session wipes itself 30 minutes after your last message.
• The whole demo (sessions, files, usage stats) wipes daily at 23:00 ET.
• The link itself is honor-system — don't share it outside this deck.

When you're done

Close the tab. If you want to talk about something like this for your own intake flow, the contact button is on the deck page you came from.

Open the demo →