Asks before it risks
You set the limits. Above them, it waits for a yes.
human-in-the-loop approval
The office is empty.The work isn't.
AI systems built for businesses. An agent that lives inside your company and keeps working when your team goes home.
Scroll to see it work
Worked example
A day inside an operation
Pick a moment. Watch the agent plan, act and ask. Press approve.
Trigger
At 10:40, orders arrive at three times the usual pace. By noon there are 1,200.
Plan
- 1.Read each order into clean fields. structured output
- 2.Send it to the nearest branch that has every item in stock.
- 3.Balance drivers by load.
- 4.Use a small, cheap model for the routine orders and the strong one for the odd ones. cost-aware model routing
Systems it touched
- the order system
- branch stock
- the driver dispatch board
- the customer messages
Needs a person
38 orders cannot be filled at any branch. The operations manager sees the list and chooses substitute, delay or cancel for each.
Result
…
Illustrative. The names and numbers are made up to show the shape of the work. They are not a client result.
Built into how you work
Not a chatbot in a browser tab. A colleague with a memory, your tools and your rules.
We call the whole thing AIOS, an AI operating system for a business.
- 01Context
Remembers your business
Who you are, your rules, your customers, your prices.
- 02Data
Reaches your systems
Your ERP, CRM, store, inbox and messages.
- 03Intelligence
Thinks it through
The AI model, reasoning over what it remembers and what it can see.
- 04Automation
Acts on real events
A new order, a low stock level, a signed contract.
- 05Output
Reports where you look
WhatsApp, email, a dashboard.
What's inside
The skills behind the work
In plain words first. The technical name sits beside each one.
Remembers last week
What it learned in one conversation is there in the next.
memory across conversations
Spends wisely
Small jobs go to small, cheap models; hard ones get the strong model.
cost-aware model routing
Starts by itself
A new order, an email or a low stock level begins the work without anyone asking.
event-triggered automation
Argues the hard calls
Several agents take opposite sides before one decides.
multi-agent debate
Visible at every step
Every step, tool call and cost is logged and can be traced.
production observability
Open underneath
Built on open frameworks your engineers already know, such as LangGraph and CrewAI.
open frameworks
Tone and dialect
Speaks your customers' language
Gulf Arabic, Egyptian, formal English or a quick WhatsApp reply: the same agent answers each customer the way they wrote.
Example
هلا، متى بيوصل طلبي؟
هلا والله! طلبك رقم 4821 طلع مع المندوب، وبيوصلك اليوم قبل الساعة 8 المسا. أول ما يقرّب منك أعطيك خبر.
Grows like an employee
It starts small. Then gets promoted.
Like a good new hire, an agent starts in one role and earns bigger ones.
Your company
Customers
answers customers
Stock
keeps stock in step
Finance
reads invoices
Management
writes the morning report
- Accuracy
- 97%
- On time
- 99%
- Asked when unsure
- 100%
- Cost per task
- $0.04
Promoted: Operations, 3 months in
01
Teach it by talking
Your team explains the new job in plain words, the way they would brief a new hire.
02
Let your developers extend it
It is built on open frameworks, in your own accounts. Your engineers add skills and tools.
03
Have us do it
Small changes come with the monthly plan. A bigger role is quoted as its own project.
No promotion without a report card. Before an agent takes on more, you see its record.
Agent settings
- Customer desk
- Stock watcher
- Invoice reader
Model
Tools
Rules
Ask a manager before any refund above $200.
Make it yours
Build it your way
Every agent has its own settings, written in plain words, so your team can see how it works and change it.
Pick the model for each job
A leading paid model, or an open one that runs on your own server. Choose by cost, speed and accuracy, for each agent.
Set the rules
Limits, approvals, tone, and what it must never do. It checks them before it acts, and asks when unsure.
Connect the tools
Switch systems on and off for each agent. Your developers can add new ones.
Works with what you run
If it has an API, the agent can use it.
- SAP
- Microsoft Dynamics 365
- NetSuite
- Odoo
- Salesforce
- HubSpot
- Zoho
- Shopify
- Stripe
- QuickBooks
- Xero
- Slack
- Microsoft Teams
- WhatsApp
- Outlook
- Gmail
- SharePoint
- Notion
- Jira
- PostgreSQL
- Twilio
- Google Sheets
- Google Drive
- Google Calendar
- Telegram
- n8n
- Airtable
- MySQL
- MongoDB
- Supabase
- WooCommerce
- PayPal
- Square
- Sage
- Asana
- Trello
- ClickUp
- Zendesk
- Intercom
- MailChimp
- Calendly
- Dropbox
- GitHub
- Instagram
- Facebook
- WordPress
- OpenAI
- Claude
- Google Gemini
{ Your in-house system }
Works with these and more. Logos belong to their owners. Listing a system is not a partnership or an endorsement.
The proof is us
We run on it ourselves.
Flowlayers is one founder and 16 AI agents. One of them is in charge of operations.
Sales research, content, accounts, engineering and quality checks each have an agent with a role, tools and a boss. They report to Horus. Ahmed reads one summary a day. Write to Horus on WhatsApp. Ask what we build, how it works, or for a call with Ahmed.
Talk to Horus, our AI CEO, now
Horus is an AI. It will tell you so.
Horus · Flowlayers
AI
Hi. We run 6 clinics and the phones never stop. Can an agent help?
Hello, I'm Horus, Flowlayers' AI. Yes, that is a common first job. What are most of the calls about: bookings, prices, or results?
Mostly bookings and changing them.
Then a voice agent that books into your calendar and passes hard calls to your team would fit. Would a free 30-minute audit with Ahmed, our founder, help? I can book it now.
The fleet at work
- Horus assigned a task to Thoth1 min ago
- Thoth researched a prospect3 min ago
- Ra drafted a post for review6 min ago
- Ma'at checked a post against the rules9 min ago
- Ma'at approved a post14 min ago
- Ptah fixed a failing check18 min ago
Real operations, real numbers
Three businesses. Names withheld. Only figures we can stand behind.
A pharmacy chain in Kuwait, 39 branches
About 1,100 to 1,450 real deliveries a day
Order dispatching automated.
Our own measurement.
The same chain
About 400 invoices a month
Supplier invoices read by AI and entered into the ERP.
A trading group, two companies, one owner
600 to 800 invoices a month
Invoice processing.
A software company
Its whole client onboarding automated and orchestrated on n8n. We built it and handed it over to their team. It saved them time.
We keep client names private. We publish only the figures listed here.
Pricing
Prices, in the open
One price list, wherever your business is. Ahmed confirms the exact figure in writing after a short call.
US dollars. Invoicing in local currency is available on request.
One workflow
One problem, one automation.
One-off setup
$750 – $1,500
About two weeks
Goes up with: a third system, messy data, approval steps, Arabic plus English.
Talk to Horus about thisAn agent
One channel, answers from your own information, hands hard cases to a person.
One-off setup
$1,000 – $2,000
About two weeks
Goes up with: extra channels or systems; more call types or lines.
Talk to Horus about thisThe full AI operating system
All five layers, installed across the business, with your team trained to run it.
One-off setup
$3,000 – $8,000
About 2 to 4 weeks
Goes up with: the number of systems, branches and exceptions; company size; custom forecasting.
Talk to Horus about thisA custom platform or app
Web, mobile and backend for a process nothing on the market handles. A first working version (MVP) within 2 weeks, scoped by Ahmed.
from$3,000
Talk to Horus about thisAfter launch, optional
Run and maintain
Monitoring, fixes, changes and updates, so it keeps working as your business changes.
$250 – $600 a month
Prices do not include outside tool costs: AI model usage, phone numbers and call minutes, WhatsApp conversation fees charged by Meta, and software subscriptions. Your written quote lists each one. For voice agents on an Egyptian phone line we confirm the connection route before we quote.
Ahmed confirms the exact figure in writing after a short call.
From first call to running
Four steps, no slides.
- 1
Talk to Horus.
Tell it what slows you down; it asks the right questions.
- 2
A free 30-minute audit with Ahmed.
Horus books it once it understands the work. You leave with a written, fixed-scope quote.
- 3
We build it into your operation,
piece by piece, with regular demos.
- 4
It runs inside your business.
We install it in your accounts under a licence, with your data staying yours. Keep us on to run and improve it, or have your team take over the day-to-day.
Want us to keep running it? That is the monthly plan above.
Questions people ask
Short answers. Horus has the long ones.
Is this a chatbot?
No. A chatbot answers. An agent also acts: it uses your systems, follows your rules, and asks a person before it takes a risk.
Will it replace my team?
It takes the repetitive work, and anything above your limits goes to a person. What your team does with the time is your call.
What systems does it connect to?
Anything with an API. We check each system's access first, because some vendors restrict who may connect, and we promise nothing until we have confirmed it.
What if it gets something wrong?
You set the limits, it asks before anything above them, and every step is logged. A bigger role comes only after its report card.
Who owns it?
Your data and your accounts are yours. The agent framework is ours and is licensed to you for as long as you use it. Your team can extend it; the licence covers that.
How long does it take?
About two weeks for one workflow or one agent, two to four weeks for the full operating system, and a first working version of a custom platform within two weeks. We never promise a date before scoping.
How is the price set?
By product and scale, with one price list for every country. The ranges are above, and outside tool costs such as AI model usage come on top. Ahmed confirms the figure in writing after a short call.
Do you work in Arabic?
Yes, Arabic and English, natively.
Do you do voice?
Yes, we build voice agents that answer your calls. We do not build agents that cold-call strangers. For an Egyptian phone line we confirm the route first.
Do you work outside the Middle East?
Yes, remotely, anywhere. Ahmed is in Cairo.
Is Horus a person?
No. Horus is an AI, and the call is with Ahmed Elrayes, the founder.
Can I see a demo?
Ahmed walks you through real systems on the call.