← All products

PRODUCT · NODEDEALEARLY ACCESS

From first message to paid order. The AI never sets the price.

NodeDeal carries a buying conversation from “hi” to a paid deposit — configuring the offer, negotiating within the authority you publish, sending a versioned quote, and creating the order. The model proposes; your pricing engine and policies decide.

  • Prices from one deterministic engine
  • Discounts within the authority you set
  • Owner approvals with the economics
  • Paid only on a verified webhook

Inbox ·Maya Chen

WhatsApp

Maya Chen

Hi! I need a dress for a gala on Oct 16. Can the Mira be done in emerald silk blend with custom sleeves?

Deal recordcalculated

Customer
Maya Chen · WhatsApp
Offer
Mira dress · emerald
Options
—
Lead time
—
List price
—
Ask
—
Policy
—
Approved
—
Quote
—
Deposit
—
Order
—

Sample deal: Maya Chen asks for the Mira dress · emerald in silk blend with custom sleeves for Oct 16. The engine prices it at C$1,180.00 with rush. She asks for 8% off; the policy routes it to the owner, who approves 5% and a garment bag. Quote Q-2026-1048 version 2 totals C$1,266.73 with tax; the C$633.37 deposit is confirmed by webhook and order O-2026-812 is created, with pickup on Oct 16.

THE PROBLEM

Your customers don’t buy from a cart.

They send a photo, ask for changes, negotiate, wait for a quote, and pay a deposit — all in chat. The seller pieces it together from memory, spreadsheets, and payment links, and value leaks at every handoff.

  • Can you make this in another colour?
  • What would it cost for 45 people?
  • Can you deliver Saturday morning?
  • Can I pay 50% now?
  • What’s the best package under C$600?
  • Can you match this reference photo?
  • What happens if the class is cancelled?

Where the value leaks

  • Slow replies

    Inquiries arrive while you’re busy or asleep, and buyers move on.

  • Inconsistent prices

    Two people, two numbers — assembled from memory and spreadsheets.

  • Excessive discounts

    Concessions given to save a sale, with no view of the margin.

  • Missing details

    Sizes, dates, or guest counts discovered after the quote went out.

  • Promises you can’t keep

    A date agreed in chat before anyone checked capacity.

  • Abandoned quotes

    Quotes sent, never opened, never followed up.

  • Forgotten follow-ups

    Balance reminders and review requests that never go out.

  • No attribution

    No record of which conversations actually became money.

HOW IT WORKS

Three lanes.
One deal.

Every step of a sale has an owner. The AI Seller runs the conversation; engines and your published policies decide anything with a number in it; your team approves the exceptions and delivers. Hover a step to see what happens there.

  1. 01ConversationAI Seller

    A message arrives from your website widget or WhatsApp. One customer timeline, whichever channel they use.

  2. 02IntentAI Seller

    The AI Seller works out what they want, by when, and for how many — one high-value question at a time.

  3. 03ConfigureEngines & policies

    The offer engine checks options, dependencies, lead time, capacity, and price. Nothing is estimated by the model.

  4. 04RecommendAI Seller

    At most three explainable choices — recommended, lower-cost, premium. An upsell has to improve the fit, not just the price.

  5. 05NegotiateEngines & policies

    Your policies trade scope, timing, and add-ons before discount. The engine builds valid counter-offers; the model phrases them.

  6. 06ApproveYour team

    Anything beyond the AI’s authority becomes an approval with the ask, the economics, and the customer’s own words.

  7. 07QuoteEngines & policies

    A versioned quote on a secure mobile page. A sent version never changes — a revision is a new version.

  8. 08PaymentEngines & policies

    Stripe-hosted checkout collects the deposit. It counts as paid only when the provider’s signed webhook confirms it.

  9. 09FulfilYour team

    The order, bookings, and production brief come from the accepted quote — never from an AI summary that could drift.

  10. 10Follow upAI Seller

    Consent-aware reminders, balance requests, and review asks — with quiet hours, opt-outs, and stop-on-reply.

  11. 11LearnEngines & policies

    Objections, unmet demand, loss reasons, and attribution from conversation to collected revenue.

channels — web and WhatsApp
2channels — web and WhatsApp
AI tools, each with limits
9AI tools, each with limits
workspace modules
13workspace modules
prices written by the model
0prices written by the model

THE PRINCIPLE

The model proposes.
Your policies decide.

Write your rules in plain words; the engine runs them exactly, on every message, with a trace. Try the sample policy — pick a dress, set a rush date, and drag the discount the customer asks for.

  1. 01AI Seller

    May offer up to 5% on its own — if margin stays at 42% or more.

  2. 02Sales Manager

    Approves 5–10%, while margin holds at 42%.

  3. 03Owner

    Approves over 10%, over 5% on a rush date, or any margin under 42%.

  4. 04Never

    Below a 30% margin. The engine blocks it — no one can approve it.

Policy simulator

Sample studio · sample policy

Configured offer

Mira dress
C$810.00
Silk blend
C$140.00
Custom sleeves
C$90.00
Rush production
C$140.00
List price
C$1,180.00
8% off

Needs the Owner’s approval

A rush order with 8% off (above 5%) needs the Owner.

Discount
−C$94.40
Net price
C$1,085.60
Est. cost
C$649.00
Margin
45.0% → 40.2%

Why — every rule, every time

  • Discount authorityA rush order with 8% off (above 5%) needs the Owner.Owner
  • Margin floorMargin would fall to 40.2%, under the 42% floor.Owner

What the AI Seller tells the customer

“Because this is a rush timeline, the owner needs to confirm the pricing. I’ll let you know as soon as I hear back.”

Offers, costs, and rules come from the product’s fictional sample studio. Each business sets its own tiers, floors, approvers, and non-cash concessions — and can test a change in the simulator before publishing it.

CAPABILITIES

Everything between hello
and paid.

Move your pointer across each figure. Configuration, negotiation, quotes, payments, and follow-up — one system, one record, one set of rules.

01

Offers that know their options

Variants, modifiers, requires-and-excludes rules, lead times, rush fees, and capacity — modelled once, so a custom request becomes a valid configuration.

02

Web and WhatsApp, one thread

A website widget and the WhatsApp Business Platform feed one customer timeline. WhatsApp’s 24-hour window and template rules are enforced.

03

Negotiation within bounds

Counter-offers that trade scope, timing, and add-ons before discount. The engine builds options that pass policy; the model only words them.

04

Approvals with the economics

The ask, the policy and its version, margin before and after, the customer’s words, and alternatives — approve, add conditions, decline, or delegate.

05

Quotes customers can act on

Versioned quotes on a secure mobile page: view, ask a question, request changes, accept by name, and pay the deposit.

06

Paid means confirmed

Stripe-hosted checkout, card numbers redacted from chat, and payments marked paid only by a signed, de-duplicated webhook. One order per quote.

07

Follow-ups that respect consent

Sequences for quote nudges, balance reminders, and reviews — gated by a consent record, with quiet hours and stop-on-reply.

08

Demand you can act on

Objections, requested-but-unavailable options, loss reasons, and attribution — each metric with its definition and comparison period.

09

Agents released like software

Test suites, versioned releases, canary rollout, rollback, a trace for every AI turn, and a kill switch.

WHO IT’S FOR

Built for purchases
a cart can’t handle.

Owner-led businesses whose customers configure, negotiate, and pay a deposit in the conversation. One platform underneath; a starting template for each kind of business.

First focus

Custom fashion & made-to-order

Bridal and occasion wear, tailoring, jewellery, personalised gifts, print and embroidery.

What gets configured

  • Fabric and colour
  • Sleeves and fit
  • Measurements
  • Fittings
  • Rush dates

“Can you make the Mira in emerald silk by Oct 16?”

Next

Catering, cakes & events

Event catering, custom cakes, meal plans, and private chefs.

What gets configured

  • Guest count
  • Menu
  • Dietary needs
  • Staffing
  • Delivery window

“What would dinner for 45 cost, with a vegetarian menu?”

Next

Services, workshops & classes

Photographers, decorators, beauty studios, workshops, and classes.

What gets configured

  • Packages
  • Add-ons
  • Availability
  • Deposits
  • Rescheduling

“Can I pay 50% now and the rest on the day?”

THE WORKSPACE

One record,
from hello to review.

Inboxes track threads, CRMs track deals, payment tools track money. NodeDeal keeps one versioned commercial record across all of it — so the production brief comes from the accepted quote, and revenue traces back to the conversation that earned it.

  1. Customer
  2. Conversation
  3. Configuration
  4. Policy decision
  5. Negotiation
  6. Quote version
  7. Payment
  8. Order
  9. Fulfilment
  10. Review

Inbox

Web and WhatsApp in three panes, with AI drafts, takeover, and the deal beside the thread.

Offers

Catalogue, variants, modifiers, constraints, and a live configuration workspace.

Policies & approvals

Plain-language rules, a simulator, versions, and an approval queue.

Quotes & payments

Versioned quotes, payment links, invoices, refunds, and reconciliation.

Orders & bookings

Orders, fulfilment boards, bookings, and a capacity planner.

Sequences & reviews

Consent-gated follow-ups, private feedback, and review requests.

Revenue & intelligence

Funnel, opportunities, attribution, objections, and demand.

Agents

Behaviour, tools, test lab, releases, traces, and usage.

Knowledge & media

Approved sources with precedence, conflict checks, and reference images.

CRM

Contacts, identities, consent, customer memory, and duplicates.

Integrations

Widget designer, WhatsApp, Stripe, API keys, webhooks, and logs.

Settings & onboarding

Roles, permissions, audit log, retention, and a guided launch.

Customer pages

Quote, payment, order status, booking, and feedback — no login needed.

Thirteen modules, one record underneath. Every material change is an event — written to the audit log and delivered to your own systems as a signed webhook.

UNDER THE HOOD

Built for money, not just messages.

A wrong price or an unkept promise costs real money. So NodeDeal keeps commercial truth in deterministic services, checks what the model says, and treats every payment as unconfirmed until the provider proves otherwise.

The engine, in plain words

  • One pricing engineEvery price, margin, and lead time computed in one place
  • Money in integer centsNo floating-point drift in a quote or a total
  • Versioned policiesPlain words for owners, typed rules for the engine
  • Guarded state changesQuotes, payments, and orders move only through legal steps
  • Verified repliesEach AI reply checked for amounts no tool returned
  • Hybrid retrievalVector and keyword search over approved knowledge
  • Events and outboxEvery change is an audit entry and a signed webhook
  • Idempotent by designA retry or a duplicate webhook never charges twice
  • Tenant isolationEvery query scoped to the signed-in workspace

No card data in chat

Payments run on Stripe-hosted pages. A card number typed into a message is detected and redacted before anyone sees it.

Nothing paid on a redirect

The customer’s return page says “confirming” until the provider’s signed webhook lands. Only then does the order exist.

Margins by role

Cost and margin are removed on the server for roles without access. A seller sees a lock and a reason — never the number.

Consent before contact

Follow-ups need a consent record for the purpose and channel, and they respect quiet hours and opt-outs.

ROADMAP

What’s live,
and what’s next.

The full path from conversation to paid order works today on the web and WhatsApp. More channels and connections to the tools you already use come next.

Now

Available in early access
  • Website widget and WhatsApp inbox with an AI Seller
  • Offers, variants, modifiers, and constraints
  • Policy engine, simulator, and approvals
  • Versioned quotes with a secure customer page
  • Stripe-hosted deposits confirmed by webhook
  • Orders, bookings, capacity, and fulfilment
  • Follow-up sequences, reviews, and demand intelligence
  • Agent test lab, canary releases, and traces

Next

Up next
  • Instagram messaging
  • Shopify, Square, and Toast connectors
  • QuickBooks and Xero sync
  • Calendar sync
  • A portfolio view for agencies

Later

On the roadmap
  • AI calling for opted-in callbacks
  • Custom reports
  • Vertical templates from partners

COMMON QUESTIONS

Before you let AI sell.

Can the AI give a discount on its own?

Only inside the authority you publish. In the sample policy, the AI Seller may offer up to 5% when the margin stays at 42% or more. Anything beyond that becomes an approval for the right person, showing the ask, the margin before and after, and the customer’s own words. The model never decides the number.

Where do the prices come from?

From one pricing engine that reads your offers, options, costs, lead times, and rush rules. The AI can mention a price only after a tool returns it, and every reply is checked for amounts no tool produced before it is sent.

How do deposits and payments work?

Customers pay on Stripe-hosted checkout, so card details never pass through the chat — a card number typed into a message is detected and redacted. A payment is marked paid only when Stripe’s signed webhook confirms it, never on a redirect, and each accepted quote creates exactly one order.

Which channels does it support?

An embeddable widget for your website and the WhatsApp Business Platform, with one customer timeline across both. WhatsApp’s 24-hour service window is enforced: outside it, only approved templates can be sent. Instagram messaging is on the roadmap.

What happens when a customer wants a person?

They can ask at any time. The conversation is handed off with a summary, the AI pauses on that thread, and your team picks it up in the inbox with the configuration, quote, and history beside it. You can hand it back to the AI when you’re done.

How do we know the AI is ready to go live?

Every agent version runs against a saved test suite before release — including adversarial cases like a large discount on a rush order. Releases go through review, a canary percentage, and live, with one-click rollback, a trace of every AI turn, and a kill switch.

Who can see costs and margins?

Only roles you allow. Cost, margin, sensitive customer notes, and transcripts are removed on the server for everyone else — a seller sees a lock and a short explanation, never the number.

How is it different from NodeConnect?

NodeConnect answers product questions from your catalogue and turns interest into leads. NodeDeal is for purchases that have to be configured, negotiated, quoted, and paid for in the conversation — it carries the deal to a paid order under your policies. Some businesses need one; some need both.

Can NodeFlow Lab build something like this for us?

Yes. NodeDeal is built the way we build for clients: AI that proposes while deterministic services decide, every decision traced, and money handled only through verified events. Book a call and tell us what you need.

NODEDEAL · EARLY ACCESS

Every conversation,
a governed path to paid.

Book a 45-minute call to walk the sample deal end to end — or to map your own offers, policies, and channels onto it.

Book a demo