An AI team thatdoesn’t sleep

CORE holds the knowledge. On top of it, VELA answers inquiries, MAKE produces content, HUNT finds buyers. Live in 7 days, with the 2am message answered.

XELONO · WorkspaceAnonymized
WorkbenchSnapshot 10/09 16:27
AllAwaitingAI only
24H COVERAGE172 off-hours
0006121823
🇰🇪
Customer 0590Kenya
02:47·Africa/Nairobi (UTC+3)·us 07:47
AI handlingEnglish
AI handling · 4 rounds auto-servedAuto-translate on
Hi, we need cotton jersey single knit, 180 gsm, like last time. 3000 kgs monthly, ship to Mombasa.
→ [ZH 自动译] 你好,我们要纯棉单面针织布,180 克,跟上次一样。每月 3000 公斤,发蒙巴萨。
Customer·02:41
AI extracted 5 fields in one message
Good to hear from you again. 180gsm single jersey in 3000kg monthly — same spec as your June order. I've pulled the previous lot's shade records so we match the hand-feel. One thing to confirm: do you need the same 3 colours or a new palette?
→ [ZH 自动译] 又见面了。180gsm 单面布、每月 3000kg——与你 6 月订单同规格。我已调出上一批的色差记录,保证手感一致。需确认一点:沿用原来 3 个颜色,还是换新色卡?
AI·02:41
AI recalled order history from knowledge base
Same 3 colours. Can you do 4% off since volume is up?
→ [ZH 自动译] 还是那 3 个颜色。量上去了,能给 4% 折扣吗?
Customer·02:43
Price negotiation detected → flagged for human
Volume up 20% year on year — I've noted the 4% request. Pricing at that level needs our sales manager to sign off, so I've passed it to them with the full volume breakdown. Everything else is already quoted: 3000kg, Mombasa, T/T 30% deposit. You'll get the revised sheet before your 9am.
→ [ZH 自动译] 同比量增 20%,4% 的折扣要求我已记录。这个价位的让价需要销售经理签字,我已连同完整的用量明细转给他。其余部分已确认:3000kg、蒙巴萨港、T/T 30% 定金。你当地早上 9 点前会收到修订报价单。
AI·02:43
Escalated: discount beyond AI authority → human takeover
AI is replying in English…Auto-send ON
86
Lead confidenceHigh
Repeat buyer, real spec, verifiable company
CAPTURED BY AI6/6
Product96%
Cotton jersey, single
"cotton jersey, single knit"
Weight94%
180 gsm
"180 gsm, like last time"
Quantity91%
3,000 kg / month
"3000 kgs monthly, recurring"
End use88%
T-shirt production
"for our tee line"
Destination90%
Mombasa, KE
"ship to Mombasa"
Paymenthistory76%
T/T 30% deposit
from 2 prior orders
BACKUP CHECKS
Repeat customer2 orders in last 12 months, no disputes
Real spec givenWeight + quantity + port all specified unprompted
Company verifiableDomain email + registered name match
Price sensitivityAsked for 4% discount vs last order

Drag the title bar to move · drag edges or corners to resize

28
VELA iterations
Versions run against real inquiries
78
Data tables
The operating model behind MAKE
1,000hrs+
Closed-door interviews
Raw material for the knowledge base
7days
To go live
From building the base to answering
Product highlights

An AI team that delivers

Not four tools. One line from knowledge base to inquiry.

Operations dashboard

Knowledge before AI

Drawings, process specs, quoting limits, structured into a knowledge base first. What the AI says must trace back to it.

Workspace conversation

No spec, no quote

No drawing, no tolerance, that’s a vague inquiry. The system hits the red line and steers toward a sample first.

Client library

Answers in their language

English, Spanish, Swahili all running. Fast, and it reads like a trade person, not a translator.

Priority tiers

Three levels of access

Publishable, negotiable, never disclosed. The cost floor sits in the innermost layer, and key actions need approval.

Opportunity pool

The shop floor becomes content

Competitor teardown, shot lists, cross-platform scheduling. 78 data tables, one line.

Built by role, no factory rebuild

One base, different jobs on top.

01

Inquiry handling

A 2am message in English or Spanish gets answered in the buyer’s language. Application and volume first, price after.

  • No drawing, no quote
  • One-page buyer check
  • Auto alert past one hour
Inquiry handling
02

Content production

Shop-floor footage, container loading, spec sheets, cut into a shot list on the competitor’s cadence and scheduled out.

  • 78 data tables
  • 11 agents
  • 8 loops
Priority clients
03

Buyer outreach

Search global bills of lading and procurement data to find who already buys what you make, then write the first email.

  • Bills of lading
  • Target list
  • First outreach email
Client library
Results

What it looks like running

Every figure below traces back to a record in the system.

VELA: the 2am inquiry gets answered

VELA: the 2am inquiry gets answered

1hr
Overdue alert
28
Real iterations
7days
Base to live
MAKE: shop floor to publishable content

MAKE: shop floor to publishable content

78
Data tables
11
Agents
8
Content loops
CORE: one knowledge set, four agents

CORE: one knowledge set, four agents

1,000hrs+
Interviews
2,000
Pain points
3
Access levels
HUNT: find buyers instead of waiting

HUNT: find buyers instead of waiting

6
Trade languages
24/7
Outreach
1
Feeds VELA

Different trades, same base

The product categories differ wildly. Answering an inquiry doesn’t.

Precision hardware & CNCTolerances
Textiles & fabricGSM & MOQ
Home & furniturePanels & finish
Small appliancesCerts & voltage
Auto partsOEM supply
Packaging & printingProcess & layout
LightingOutput & safety
Medical devicesRegistration
Building materialsStandards
Electronic componentsPart numbers
Chemical materialsGrades & MSDS
Outdoor & sportingSeasons
Pet suppliesTooling
Jewelry & accessoriesPlating & weight
Rubber & plasticsMolds & runs
Construction machineryFitment
Service and support

Delivery isn’t handing over a login

Building the base, drawing the red lines, reading the data, we do that.

01

Knowledge build

Drawings, processes, quoting limits, itemized and tagged in three tiers.

02

Red-line rules

Cost floor, factory details, certification wording. Settled before any conversation.

03

Script rehearsal

Run real inquiries through it before it goes live, so it sounds right.

04

Data review

Which content brought inquiries, which line stalled the buyer. Adjusted monthly.

Want to see this working on your shop floor

Leave a contact. We’ll run one of your real inquiries through it.

Common questions

The ones we get most.

Will VELA quote the wrong price?
No. Without a drawing or tolerance it’s classified as a vague inquiry, which triggers the no-tolerance-no-quote rule and steers toward a sample. The cost floor and factory details sit in the innermost access layer, out of reach.
Do we have to build the knowledge base ourselves?
Not entirely. You supply the raw material (drawings, process specs, quoting limits). Structuring, tagging and access tiers are part of delivery. CORE isn’t sold separately.
How long until it’s usable?
Seven days. Account and network on day one, first real inquiry on day seven. First rehearsal on day 15, first data review on day 30.
Which languages?
English, Spanish, Swahili and other trade languages are running. Replies go out in the buyer’s language, not Chinese routed through a translator.
Does it connect to our ERP or WhatsApp?
Yes. WhatsApp is the main entry point, and deal stages map to existing client-record fields. Exact integration scope depends on your stack.
Where does the data live?
Private deployment on a server you designate. Model calls, customer conversations and the knowledge base all stay in that environment.

Every 2am inquiry gets answered

VELA at $567 / month, 3 seats, including private deployment, model usage and a custom knowledge base. Live in 7 days.