Trained on your archive, not a ready-made template

An agent for every part of your company

We sit with each of your departments to understand how it actually works, record and index its data inside your own account, then train a small model that belongs to it: replying to your customers in your team's dialect, triaging your inbox, forwarding only what matters to your personal number, verifying your transfers, and reviewing your developers' code.

Book a demo on your data See the agents
Agent — Al Diyar Developmentsreplies in 4 seconds WhatsApp
Hi — is the apartment from the New Cairo listing still available?
Hello Mahmoud 👋 Yes, still available — 142 m², third floor, fully finished.
Price was your concern last time, so let me lead with it: 6.2M with 25% down over 6 years. Shall I book you a viewing tomorrow at 5?

Why did it answer that way?

What the agent pulled from the company archive before writing a word

  • Customer historyAsked about two units in March and never closed
  • His recurring objectionPrice — so it opened with the payment plan
  • Team tonePolite, short sentences, a question at the end
  • Before confirmingChecked the unit in your system, so it promised nothing unavailable
Segment: hot lead

Six agents, each fluent in its own job

We don't sell you one chatbot. We build specialised agents that work inside your company, each with its own archive, permissions and limits — all in a single dashboard.

Customer agent

A customer who waits an hour goes to your competitor. This agent replies in seconds in your team's own dialect, having already read that customer's history before writing a word.

Incoming chat — WhatsAppreplied in 6s
Can you check if order 7731 has shipped?
It shipped yesterday and it's with the courier today. Want the tracking link?
It pulled the order number from your system, and knew from his message history that this customer prefers short replies.
  • Answers messages and comments on every channel, day and night
  • Moves serious buyers to DM and hides abusive comments before other customers see them
  • Knows who bought, who stalled, and who is worth chasing today
  • Hands over to a human at anger, or any request beyond its authority

Inbox agent

A manager opens 142 emails every morning and reads nothing but subject lines. The agent reads all of them and hands you the seven worth your time.

Morning brief — Sunday 9:00142 mails · 7 need you
  • UrgentMain supplier threatens to hold tomorrow's shipment unless the payment is approved
  • 4 daysAWS subscription auto-renews at $480 — renew or cancel
  • No reply, 3 daysCustomer asked for a quote · Owner: Ahmed — Procurement
  • Answered for you23 routine mails: invoices, appointment confirmations, file requests
Slowest replies this week: Procurement, averaging 19 hours · Outside your hours, only "urgent" reaches you.
  • Connects every company mailbox and reads thousands of mails a day
  • Warns you about subscriptions and contracts before they renew or lapse
  • A report on which staff replied late and how long they took
  • Answers on your behalf the message types you define

Personal number agent

Your number is printed on every card and invoice, so by now everyone has it. Make it your agent's number: it receives everything, and only what we agreed deserves to interrupt your day reaches you.

Your personal number — today68 messages · 3 forwarded
  • ForwardedMohamed El Far — key supplier: "contract ready to sign"
  • ForwardedBank alert: incoming transfer of 2.4M
  • ForwardedTwo missed calls from a number tagged as a major client
  • Blocked41 marketing messages · 9 donation requests · 15 expired one-time codes
Share your number without fear — it is the agent's number, and you sit behind it.
  • Blocks marketing, donation requests and spam
  • Forwards the critical call or message to you within seconds
  • Learns what matters to you from your archive, our meeting, and your dashboard
  • Answers everything else politely in your name, without you ever seeing it

Finance agent

An accountant opens the bank statement and reconciles transfers one by one, by hand. The agent reconciles all of them before he reaches his desk.

Payment check — invoice INV-2291InstaPay · 09:41
  • MatchReference 884219077 found in the bank SMS
  • MatchSender name matches the customer on the invoice
  • ShortReceived 48,000 against 52,000 due — 4,000 missing
  • ActionNot approved · routed to the account's accountant with the SMS attached
3 short payments and 1 duplicate receipt stopped this month before anyone approved them.
  • Verifies InstaPay and bank transfers via reference number and SMS alerts
  • Catches short payments, duplicate receipts and anything that doesn't match the invoice
  • Links every transfer to the customer and invoice in your system
  • Assigns work to accountants and tracks what hasn't been closed
  • Never approves a payment on its own — the financial decision stays human

Engineering agent — your own CTO

Code review eats your team's hours, and the vulnerability slips through because nobody looked in time. The agent reads every change before merge and arrives with a patch, not a comment.

Review PR-418 — payments/webhook.tsGitHub · pre-merge
  • VulnerabilityThe request executes before the payment provider's signature is verified
  • PatchOne-line fix, no change to successful payment behaviour
  • DirectionMatches your team's decision to centralise verification in middleware
- if (body.event === 'paid') markPaid(body.id)
+ if (!verifySignature(req)) return res.sendStatus(401) + if (body.event === 'paid') markPaid(body.id)
40 minutes of review saved on this request alone · [approve] [edit] [reject]
  • Reviews every pull request with a ready patch instead of a vague comment
  • We track security advisories daily, so it closes what opens up in your dependencies
  • Understands your company's direction, so it won't propose what contradicts the plan
  • Suggests interface and user-experience improvements
  • No access to secrets, and it merges nothing without human approval

Marketing agent

The discount ends Thursday and half your customers have forgotten. The agent reminds only those worth reminding, at the hour they actually reply.

Campaign: end-of-season discountreminder 2 of 3
  • Segment342 customers who asked but didn't buy within 60 days
  • TimingTuesday 7pm — the highest-response window in your archive
  • CopyIn your team's voice, reshaped for each platform
  • Expected9–12% reply rate, based on your last four campaigns
It never messages someone who already bought, or anyone who asked to stop.
  • A weekly offer reminder to the right segment, not the whole list
  • The year's seasons and holidays scheduled in one pass
  • Write the post once and it goes out in each platform's format, on time
  • Measures which message actually produced a sale, not just an open

A ready-made template, or a model that knows your company?

Most of what's sold today is a prebuilt workflow wired to a generic model. It performs in the demo and collapses at the first real customer.

The common approach

  • −A prebuilt template, identical for every company
  • −A general model answering in language nothing like your team's
  • −Every conversation starts from zero — no memory, no history
  • −One chat bot and nothing at all for your other departments
  • −Cost climbs with every message because everything runs through a huge model

The nqdr approach

  • The customer never feels they are talking to a machine, because the voice is your team's
  • You stop losing deals to an objection your archive answered a thousand times
  • Departments automation never reaches — inbox, finance, code — come with you
  • Every agent has a narrow permission, and every serious decision stays with a person
  • Running cost stays nearly flat, because the model is small and specialised, not universal

So why isn't ChatGPT enough?

ChatGPT and its peers are excellent tools, and we use them ourselves every day. But they are an assistant sitting in a window, waiting for you to open it and explain who you are all over again. nqdr is something else: a system that runs inside your channels, knows your customers, and acts on your behalf.

A general chat assistantnqdr
Knows your prices, policies and limits
A general chat assistantNo — unless you retype them in every single chat
nqdrTrained on them, and won't cross them even if the customer pushes
Knows the history of the customer it's talking to
A general chat assistantHas no idea who they are
nqdrReads their full record before writing a word
Who opens the conversation, and where
A general chat assistantYou do, in a separate window you copy and paste from
nqdrYour customer does — inside WhatsApp, Instagram and your inbox, around the clock
Acts, or only writes
A general chat assistantWrites text that you then execute by hand
nqdrBooks the appointment, routes to a rep, holds the short payment
Who fixes it when it gets something wrong
A general chat assistantYou do, once you notice — after it reached the customer
nqdrWe do: a test bank built on your cases, a monthly review, a log of every reply
Where your customer data ends up
A general chat assistantYour staff paste it into a tool you don't control
nqdrInside your account alone, with narrow permissions and a full log
We are not a replacement for ChatGPT, and it is not one for us. It helps you when you ask it; we work when you are not there.

How we build your agents

Five stages that start with a meeting, not with code. Your first agent is live within two weeks.

1

We sit with every department

A separate session with sales, support, finance, engineering and marketing. We listen to how the department actually works, what eats its time, which decisions must stay human, and what must never be said or published.

2

We record and index your data

Customer messages, company mail, support tickets, transcribed sales calls, invoices and bank statements, and your code repository if you want it — only what you choose to share, under permissions you grant. It is recorded and indexed inside your account alone: years of it, not a week's sample.

3

We train a model per agent

A small model that belongs to your company and carries its tone, prices and policies. We test it on real past cases and compare its answer to your team's before it touches a single customer.

4

We launch with narrow permissions

It starts by watching, then suggesting, then replying. You decide what each agent does alone and what needs your approval.

5

We review every month

A report on the hours and money it saved, what it learned from your new archive, and what needs tuning. Every correction you make feeds the next round of training.

The strongest model is rarely the right one

Every task in your company has a model that fits it: one needs speed at a trivial cost, another needs deep comprehension of language, and a third needs no model at all — just strict programmatic verification. We choose, we test, we switch. You never read a single AI news item.

The taskWhat we runWhy
Inbox triage and message sortinga small, fast modelThousands of mails a day; a huge model here is cost without benefit and a delay your customer feels
Drafting a reply to an angry customera stronger model, with phrasing proven on your archiveA mistake here costs you a customer, so we pay for quality on purpose
Matching a transfer to its referencestrict programmatic checks, not a generative modelAccounting cannot tolerate a guess, however clever the model
Reviewing code before mergea code-specialised model with long contextIt reads the whole file and its history, not the changed lines alone
Transcribing a call in local dialecta transcription model trained on dialectsGeneral models collapse on Egyptian and Gulf Arabic and return text with no meaning

What changed for you without you lifting a finger

  • This weekA lighter model appeared for inbox triage; it passed your test bank at the same accuracy for 62% less, and was swapped overnight.
  • This monthWhatsApp changed its message template rules; we updated the connection two days before the cut-off.
  • Last monthA reworded second reminder lifted replies from 7% to 11% after a two-segment test.
  • Two months agoA vulnerability in a library your service depends on; the patch reached your team within 4 hours.
Four changes, and not one message from you.

We follow it, so you don't

New models, new tools, security advisories, and changes to platform APIs. Tracking all of it is part of the subscription, and you see the result in your monthly report rather than in a newsletter.

No upgrade before it's tested on your cases

We build a test bank from your own conversations and transactions. A new model reaches production only if it beats the current one on those specific cases — not on global leaderboards that look nothing like your market.

Switching changes nothing on your side

Same dashboard, same permissions, same tone in the replies. Only what sits under the hood changes: lower cost, higher quality, or both — with nothing going down.

Your agent knows your customers better than you do

Indexing the archive doesn't only improve replies — it shows you what the market is actually thinking.

  • The moment a customer goes quiet, and what was said just before it
  • The top questions in your field and how their share shifts month to month
  • The objection that kills your deals, and the sentence that has actually saved them
  • A weekly report: what your customers ask for that you don't offer

What your customers ask about most — last 30 days

Payment plans and down payments34%
Delivery date21%
Available unit sizes17%
Is there a cash discount?12%
Bought before 8% Comparing offers 39% Price-hesitant 33% Not serious 20%

Real replies, sector by sector

We don't sell the same agent to six sectors. Pick yours and read what your agent actually says — and why it said it.

Real estate Automotive Clinics E-commerce Education Field services

It knows the available units and payment plans, separates serious buyers from browsers, and books viewings straight into the rep's calendar.

A discount request beyond its authority WhatsApp
How much do you take off if I pay cash?
A cash discount isn't mine to promise, and I'd rather not quote a number that turns out wrong.
What I can do now: hold the unit for 48 hours with no deposit until the sales manager calls you with the cash offer, and send you the current payment plans to compare. Shall I place the hold?
Why did it answer that way?
  • Your written policy: no discount without the sales manager's approval
  • A 48-hour hold is within its authority, so it used that instead of a promise
  • 31% of discount requests in your archive end in a sale on a longer plan
Passed to sales
Budget below the entry price Instagram
Looking for a villa in Sheikh Zayed, budget 8M. What do you have?
Welcome 🙌 Our Zayed villas start at 11.5M, so 8M won't stretch to one, honestly.
But I have a 240 m² townhouse in the same compound at 8.3M, 10% down over 8 years — only 300K above your budget. Want the master plan?
Why did it answer that way?
  • Never promises what isn't available or negotiates an unapproved price
  • Picked the closest alternative inside the same compound
  • Hands over to a rep the moment the lead agrees
Alternative offered

Projects we have shipped

Real examples from companies running nqdr agents today, with results measured rather than promised.

Real estate · ٢٠٢٦

A sales agent for a property developer

Four years of WhatsApp and Instagram conversations, and a model that separates serious buyers from browsers and books viewings into the reps' calendar.

Reply time from 4 hours to 9 secondsViewings booked +38%
Trade & distribution · ٢٠٢٦

Inbox triage for a trading group

Nine mailboxes connected, one morning brief for the owner, and a weekly report on which departments replied late.

142 daily mails reduced to 7No contract lapses unnoticed since
Technology · ٢٠٢٦

Code review for a software company

Connected to the team's repositories, reviewing every pull request with a ready patch, plus daily tracking of security advisories.

Review hours −55%6 vulnerabilities closed pre-production

One screen that shows you everything

For a whole company — or for a single business owner — one dashboard: every agent and what it did, everything that needs you today, and every number that matters. And you can give each department only the view that belongs to it.

8,420conversations this month +18%
9saverage reply time −94%
316work hours saved +41
147opportunities passed to sales +23
406transfers verified 4 mismatches
89code reviews 6 vulnerabilities

Needs you today

  • Warehouse lease expires in 9 days and hasn't been renewed
  • 5 customers unanswered for two days — Sales
  • 12 hot leads no rep has contacted this week
  • Second reminder campaign goes out in 4 hours — review the copy

Your agents right now

Customer 1,204 replies today live
Inbox 142 mails · 7 need you live
Personal number 68 messages · 3 forwarded live
Finance 19 transfers · 1 held needs review
Engineering 4 reviews ready awaiting approval
Marketing campaign scheduled Tuesday live

It works with what you already have

We don't ask you to change your systems. We connect what you use today and read the customer's history, number and purchase record from it. If your system is in-house or old, we connect it over an API or a scheduled export.

Channels

WhatsApp BusinessMessengerInstagramTikTokXFacebook commentsWebsite chatCalls

Mail & office

Gmail & Google WorkspaceOutlook & Microsoft 365Domain mailCalendarDrive & SharePoint

Customer records

HubSpotSalesforceZohoOdooBitrix24In-house CRM via API

Finance

InstaPayBank SMS alertsE-walletsInvoicing systemsERP

Engineering

GitHubGitLabBitbucketJiraSlack

Commerce

ShopifyWooCommerceSallaZidIn-house inventory

Your data is yours alone

We never train one company's model on another company's data, and we never use your archive outside your own agents. Private hosting on request.

A narrow permission for every agent

Each agent sees only what it needs. The finance agent doesn't read code, and the engineering agent has no access to secrets and pushes nothing without human approval.

A full log of every action

Every reply and every decision is stored with its reason and its source inside your archive, so you always know why the agent did what it did — and who approved it.

Clear monthly pricing

The department sessions, the first training and channel setup are included in every plan. No hidden setup fee.

Starter

4,900 EGP / month

For a company starting with one agent that wants proof on the ground.

  • One agent on two channels of your choice
  • Up to 3,000 conversations a month
  • Trained on 6 months of archive
  • Dashboard and live conversation monitoring
  • Email support
Start with Starter
Most popular

Growth

14,900 EGP / month

For a company with several departments and a real archive worth mining.

  • Up to 3 agents and every channel, uncapped
  • Inbox agent and personal-number agent
  • Trained on the full archive, with CRM connected
  • Customer scoring and a weekly report
  • Reminder campaigns and scheduled publishing
  • Account manager and a monthly review
Request Growth

Enterprise

On request

For an organisation that wants every agent, engineering included, on private hosting.

  • Every agent uncapped, including engineering and finance
  • Custom model and private hosting
  • CRM, ERP and code repository integration
  • Roles, multiple teams and a service level agreement
  • Training for your team on the dashboard
Talk to us

What changed for them

“The first automated replies none of our customers could tell were automated. It talks the way our sales team talks.”

A Ahmed SalemSales director — property developer

“I open my phone to 3 messages instead of 60. And every one of them genuinely deserved to reach me.”

N Nourhan AbdallahFounder — trading group

“Code review takes minutes now. A ready before-and-after lands with us, the developer approves or edits, done.”

T Tarek MansourCTO — software company

Try it on your own data, not on a slide deck

We meet your team for an hour, you pick 500 conversations or 500 emails, and we show you what the agent would have done within 48 hours — before any commitment or payment.

Your details are used only to contact you. We don't touch your archive until there's a written agreement.