Business hours Reps first
After hours · holidays Full team
1·800·558·0119
For 20+ years, I've helped White River — makers of premium hardwood moldings, tambour, and hand‑carved millwork — sell to the luxury interior hospitality, remodeling, and new construction segments. Everything digital runs through me: eCommerce, the internal CRM, sales enablement, marketing automation. A department of one. And every system you're about to see, I built hands‑on with AI.
You get all of it.
Put the phone down — nothing here needs a screenshot.
Every slide exactly as presented — the numbers, the agent roster, the full technology stack and the economics. PowerPoint and PDF.
This walkthrough, narrated by me. Click into any section, hear the real customer calls, and replay any part at your own pace.
What I actually said on each slide, written out — the reasoning behind every number, not just the headline.
Buying happens at 9 pm, on weekends, mid‑project — and California, our biggest state, runs two time zones behind us. Accessibility was quietly capping growth, and I couldn’t hire my way out of it.
The bet: the companies that use AI to go truly 24/7 will win — and White River gets there first.
Four goals, one rule.
Scaling with AI, not headcount — here is what that had to mean in practice.
TambourTouch runs entirely on agents and Mouldings is in beta; both back up WhiteRiver’s phones and sit live on the site.
Build the quote. Look up the order. Open the trade account. Answer the product question. Send the catalog.
A bilingual agent on the line from day one. No hiring, no scheduling, no second queue.
Picked up in under a minute, every time. In B2B, whoever answers first usually gets the order.
Every quote, lead and trade application the agents create emails the territory rep the moment it happens. The agent does the work — the rep keeps the relationship.
Not website chatbots — they answer the phones. Each runs on a ~9,000-word playbook: seven product lines, trade-discount math, two languages, and a different brain for a phone call than a web chat. Tess joined in August as a second sales overflow — the bench got busy, so we added bench, not logic. All six work hand in hand with the sales team, around the clock, around the customer.
The reps get first refusal.
Business hours, the line rings the reps. Only if nobody picks up does an agent.
We are running two of these at once.
Andy Hoar’s BOT 2 BOT maps this as three stages over years. We ended up living two of them at the same time.
Every routine call goes to a person. AI helps at the edges, if at all.
Reps take the call. Agents cover the overflow and every closed hour.
Agents handle nearly everything. A human takes the exceptions.
An agent to watch the agents.
The same week Tess joined, something new sat above all six of them. Nobody has six hours a week to listen back — so the reviewing job went to an agent too.
Every conversation from all six agents, start to finish — not a sample. Last week: 128 calls, 367 minutes, read before Monday starts.
Pricing ambiguity, failed handoffs, repeat callers stuck in a loop — and calls logged as a success that weren’t one.
Every flag names the agent, the call, and why it might matter. Owen recommends — I decide.
The email Owen sends.
Not a dashboard nobody opens — an email in the inbox before the week starts.
Every sales email sent, by type — plus any that bounced.
Which rep the AI ecosystem fed that week, and how many.
A written read of the week, before the numbers.
The specific calls worth ninety seconds — and why.
Each platform plays one part — voice, phones, commerce, inventory, email, SMS — wired through n8n into a single instrument. The whole system was composed in partnership with Claude Opus: architected, wired, and debugged by one person.
The call you just heard — rendered as a live, photoreal presence. Every frame is generated in real time from Blake's own voice. Nothing about the agent changed.
The stack from earlier — ElevenLabs for voice, Twilio for the lines, Claude Sonnet and Gemini reasoning, Shopify and NetSuite for commerce and stock, n8n wiring all of it together.
+
+
The same team you just heard on the call works right on the website — by chat or by voice. A customer can open a trade account, build a real quote, or check an order without waiting for business hours. One brain, every channel.
Walks the whole application one question at a time. Activated within a business day — catalog ships in the mail.
Builds the order line by line, applies trade pricing, and emails a real Shopify Plus draft order — usually in 3–5 minutes.
Pulls carrier, tracking number, and ETA from the order — and emails the details over on request.
One branch — every call starts with the AI team. Blake sells, Brooke supports, and Blake signs up the new trade accounts. The human sales team is the second line, for escalations and callbacks. They basically run the site for us — kind of crazy, and so far it’s working.
Identical architecture, third property — in build now. June still greets, Brooke still supports, Blake keeps trade — and a new voice joins the team: Riley takes sales. When it ships, the flagship gets the same front line.
Everything in this section is a receipt — generated by the system in production, while everyone slept. Real inboxes, real Shopify admin, real safety nets. Three exhibits:
n8n
Seconds after that call, the right rep — matched by territory — gets an email and a text, and a tagged draft order is already sitting in Shopify with trade pricing applied. The rep just reviews and sends the invoice.
Rep email + SMS fire the instant a lead is captured — BCC to Richard on every one.
The rep hears about it first.
Every lead and trade application lands in the rep’s inbox, already written up.
Catalog fulfillment for new trade customers used to fall through the cracks — a manual step that got missed. Now the moment a trade application arrives, the system files it, routes it to the territory rep, and creates the complimentary catalog order on its own.
Some web orders were silently failing to sync into NetSuite — and an order NetSuite never sees is an order that never ships. The fix is two scheduled flows: a daily alert that catches misses within 48 hours, and a Monday recap that sweeps the whole week.
An order that misses NetSuite sits unfulfilled. These used to be caught days later — by an upset customer. Now they’re caught the next morning.
Synced orders carry an integrator tag. The flow simply asks Shopify for recent orders missing it — no NetSuite API call needed. Auto‑generated catalog orders are excluded on their own.
Voice Ops, by the numbers
Twelve weeks live — every call and web conversation the AI team handled, June 1 – August 24.
Straight from the dashboard.
Nothing cropped, nothing cleaned up — including the calls that went sideways.
One agent. Every job at once.
Cassidy doesn’t take turns. She holds dozens of conversations at the same time — in two languages — and inside any one of them she can build a draft order, send the right catalog, sign up a trade account, explain the product lines, or hand the call to a human. One mind, many calls.
From a voice call to a quote, while everyone slept
One trade lead came in by phone after hours. By morning the agents had filed the application, activated the account, sent the welcome, and drafted the quote — then handed the relationship to the rep. Real customer. Real emails. One night.
Your trade pricing is live on your existing account — no new login. Your territory rep, Maurice, will reach out to introduce himself.
| Name | Maurice |
| Territory | Mid-Atlantic |
| Direct Line | 1·866·967·2948 |
Hi Marcus, your draft order #DQ-2214 is ready with your trade pricing applied. Stacy will reach out with any questions.
| Draft Order | #DQ-2214 |
| Items | 1 |
| Subtotal | Trade pricing applied |
| Shipping (est. 6%) | Calculated at checkout |
A second shift for ~$350 a week
The voice agents run on three line items — a flat plan, the LLM behind them, and the phone lines. Here’s the whole bill, and what it buys.
Where this all goes, in one curve
The partnership keeps shifting: AI carries more of the build end to end, and the human moves up to the judgment call. The gap is what AI can’t do yet — and it narrows every month.