Hashinclude / Scarlett

What if AI could run your business?

The first agentic ERP solution from the region, customised 100% for your business.

Scarlett is a business platform with an assistant that works directly with live company data. It prepares the work and moves it forward. The decisions that matter stay with people.

One connected system

Not another AI tool layered on top of the software you already have.

69 modules across 10 areas, one system underneath. Each module keeps its own tables and reaches everything else by reference: one record per party, one per item, one ledger. Choose an area.

Sales and customers

Customers, quotes, orders and invoices, credit notes and the ageing that follows. Sales knows which accounts are worth the effort: the assistant plots margin against days to pay, names the biggest account that is also the least profitable, and drafts the renewal brief before the date.

Parties
Customers, prospects and contacts, one record each, shared with every module.
Quotes
Commercial proposals with versions, sent and accepted from the record.
Sales orders
Orders from accepted quotes, through to delivery and invoice.
Shipments
Delivery notes: what left the warehouse against which order.
Contracts and subscriptions
Recurring services billed on their own schedule.
Interventions
On-site work recorded against the customer and the contract.
Tickets
Issues and requests with an owner, a stage and a history.
Knowledge base
Answers the team and the assistant read from the same place.
Partnerships
Partner agreements and their terms.
Customer profitability: corporate accounts plotted by margin and days to pay, with a renewal brief drafted
Sales and customers

Customers, quotes, orders and invoices, credit notes and the ageing that follows. Sales knows which accounts are worth the effort: the assistant plots margin against days to pay, names the biggest account that is also the least profitable, and drafts the renewal brief before the date.

Parties
Customers, prospects and contacts, one record each, shared with every module.
Quotes
Commercial proposals with versions, sent and accepted from the record.
Sales orders
Orders from accepted quotes, through to delivery and invoice.
Shipments
Delivery notes: what left the warehouse against which order.
Contracts and subscriptions
Recurring services billed on their own schedule.
Interventions
On-site work recorded against the customer and the contract.
Tickets
Issues and requests with an owner, a stage and a history.
Knowledge base
Answers the team and the assistant read from the same place.
Partnerships
Partner agreements and their terms.
Purchasing and suppliers

Suppliers and their invoices, read straight from the PDF into a draft with every field marked with where it came from. Approval before posting, payables ageing, and a payment batch prepared for a human checker to review and release.

Suppliers
Supplier records with terms, bank details on file and a log of every change to them.
Supplier quotes
Request and compare prices before ordering.
Purchase orders
Raised under the agreed limits, approved before they go.
Receptions
Goods received against the order; short and over deliveries flagged.
Supplier invoices
Read from the PDF, matched to order and receipt, approved before posting.
Incoterms
Delivery terms carried on every order and shipment.
Stock and items

Items, locations and movements, with stock on hand derived from the movements rather than stored. A transfer is proposed before a store runs out, a reorder is raised under the agreed limit, and the inventory control account agrees with the count on the shelf.

Items and services
The catalogue: prices, descriptions, units and tax codes.
Stock
Locations and movements, with on-hand derived from the movements.
Lots and serial numbers
Identity and expiry per unit, where it matters.
Variants
Size, colour and other axes on one item.
Barcodes
Barcode and QR labels, scanned on receipt and issue.
Manufacturing

Bills of materials and work orders. Components are issued from stock, finished goods come back into it, and the cost travels with them into the ledger, so the margin on a finished item is a fact rather than an estimate.

Bills of materials
Components and labour per finished item.
Manufacturing orders
Issue components, receive finished goods, carry the cost through.
Workstations
Capacity, and where the work happens.
Finance

Double-entry books with every rule enforced in the database, UAE VAT, multi-currency, fixed assets and budgets. Cash is forecast week by week, and a supplier payment is timed as the terms allow, so the balance stays above its floor in the week that matters.

Invoices and credit notes
Customer and supplier, gapless numbers, ready for e-invoicing.
Accounting
Double-entry ledger, journals, periods, year-end close and exports.
Banks and cash
Accounts, statements imported and matched to open documents.
Payment batches
Supplier and salary batches with the bank file (WPS for salaries), released by a checker.
Direct debits
Collection orders against customer mandates.
Taxes and charges
VAT returns, corporate tax, social and statutory charges.
Salaries
Payroll postings and payslips.
Loans
Loan schedules and their interest.
Donations
For associations and non-profits.
Margins
Margin by item, customer and period.
Multi-currency
Foreign-currency documents, realised and unrealised exchange differences.
Fixed assets
The register and its depreciation.
Budgets
Budget against actual by account and dimension.
Reports
Turnover, journals and ageing read straight from the documents, at any date.
Projects

Cost, revenue and time by project, against the budget it was won on. Purchases, expenses and hours are charged to the project they belong to, and the margin on each is read while it can still be changed rather than after the final invoice.

Projects and leads
Opportunities and projects; any document can be assigned to one.
Tasks and time
Hours charged to the project they belong to.
Events and agenda
Meetings, calls and automatic events on the record.
Resources
Rooms, vehicles and equipment booked against events.
Event organisation
Events with attendees and bookings.
People

Employees are parties the ledger already knows, so nobody is entered twice. Leave with a decision, expense claims approved by somebody other than the claimant, and documents that expire, such as visas and labour cards, raised before they do.

Users and groups
Who can sign in, and what each group may do.
Employees
Employment records, contracts, departments, skills and interviews.
Leave
Requests, balances and decisions.
Expense claims
Submitted by one person, approved by another.
Recruitment
Positions and candidates.
Members
Association members and their memberships.
Expiring documents
Visas, labour cards and passports, raised before they expire.
Documents

Every document lives on the record it belongs to, with a checksum and a retention date. The copy the customer was sent is kept exactly as sent, the supplier’s PDF stays attached to the invoice it became, and the assistant reads from the same place people do.

Document management
Every generated or stored document filed on its record, shared when needed.
Attachments
A checksum and a retention date on each file.
Tags and categories
On items, parties, contacts and members.
Document sections
Title and subtotal lines on quotes, orders and invoices.
Rich text
Formatted descriptions and notes on documents.
Network drive
Documents reachable as a WebDAV share.
Front desk and portals

Where customers, suppliers and partners meet the system without an account on it: a touchscreen sale in a shop, a portal that shows a customer their own invoices, a payment page linked to an order, a survey sent after the job.

Point of sale
Touchscreen sale for shops, cafés and restaurants.
Web portal
Customers, suppliers and partners see their own documents.
Online payments
A card payment page linked to an invoice or an order.
Surveys and polls
Votes, polls and surveys.
Websites
A public site served from the same system, edited in place.
News feeds
Feeds shown on the home page.
System

The machinery every module shares: workflow between them, imports and exports, scheduled work, an unalterable log of what happened, data privacy under the UAE law, and the assistant under the organisation’s own permissions.

Workflow
Automatic creation and status changes between modules.
Imports and exports
Assisted, with bad rows handed back.
Scheduled jobs
Recurring work with a record of each run.
Unalterable log
A chained, read-only log of business events, exportable.
Data privacy
Retention, deletion and consent under the UAE data protection law.
Assistant
The AI features, under the organisation’s permissions and rules.
Shortcuts
Bookmarks to the pages a person uses most, always to hand.
Module builder
A low-code tool for building a module of your own.
01 · The assistant

From answering questions to doing the work.

The assistant works with the company’s live data, inside the permissions each user already has. Ask it a question, and it answers from the records. Ask it to prepare something, and it prepares it for the person who is allowed to decide.

Ask in plain language
Which customers are worth the effort? Why did margin fall in September? Which store runs out first? The answer comes from the ledger, the stock and the sales, with the figures beside it.
Surface what needs attention
Overdue invoices, low stock, a delivery short of what was invoiced, cash running below its floor in a given week. Each is raised with the reason and the number behind it.
Prepare the routine work
A stock transfer between stores, a reorder under the agreed limit, an outbound payment batch, a query to a supplier. Prepared and queued for review, within the permissions the user holds, and never released on its own.
Inventory view proposing a transfer of 180 linen throws from Deira to Jumeirah, with the margin it keeps and a map of the two stores
Stock. A transfer proposed before one store runs out, with the markdown it avoids.
Operations view tracing a problem to its cause across stores and suppliers
Operations. Not only what went wrong, but where it started.
02 · Control

AI prepares the work. People remain responsible for the decisions.

That distinction is the centre of the platform. The assistant cannot bypass user permissions, credit limits, minimum margins or required approvals. Payments, credit notes, deletions and permission changes need a person to confirm them.

Runs automatically

  • Reorder a stocked item under the agreed limit
  • Hold a payment on an invoice mismatch
  • Flag unusual refunds and voids

Needs approval

  • Move stock between stores
  • Change a payment date
  • Change a price or promotion

Never automated

  • Record a payment
  • Post to the ledger
  • Delete anything

The lines between the three columns are the organisation’s own rules, approved by a named person on a named date, and every action the assistant takes is logged with its reason.

Automation rules: what runs automatically, what needs approval and what is never automated, with an activity log of the assistant's actions and the people who approved or overrode them
Automation rules and the activity log. A payment held on a mismatch, within the rules. Only a person can release it.

Payments: prepared by the assistant, released by a checker

Scarlett prepares the batch and flags what a careful accounts clerk would notice: an invoice that looks like one already paid, a supplier whose bank details changed since the last run. An authorised human checker then reviews the batch and releases it. The assistant cannot.

The same holds on a phone. Approvals reach the person who owns them, with the reason and the number, and nothing moves until they say so.

Three phone screens: approvals waiting for a manager, the assistant answering how the brands did in September, and a stock transfer waiting to be approved
On a phone. Four approvals waiting, each with why.
03 · Your model

AI, without being locked to one AI company.

The business chooses which model powers the assistant, and can change it later without changing the system, the records or the workflows underneath.

Claude
Anthropic
Gemini
Google
OpenAI
GPT models
Falcon
TII Abu Dhabi, open weights
Other open-weight models
on infrastructure you control

For organisations with stricter data requirements, the assistant can run on controlled infrastructure using an open-weight model, so no record leaves the building. And a business can run Scarlett with no assistant at all.

Organisation settings, where the model behind the assistant is chosen and the organisation's own key can be stored encrypted
Organisation settings. The provider and the key belong to the organisation, not to the product.
04 · The ledger

Underneath, a ledger that cannot be wrong.

Everything the assistant reads and prepares rests on double-entry books where each rule is enforced in the database itself. An entry balances or it does not exist. A posted entry is never rewritten. Numbers are gapless. Money is never a floating-point value.

An issued sales invoice, with its settlement and the journal entry it posted on the same page
One issued invoice. Its settlement and its ledger entry on the same page, so what the customer owes and what the books say never drift apart.
Stock on hand by item, shown beside the inventory control account balance, which agrees
Stock on hand. Derived from movements, and shown beside the control account, agreeing.
A supplier invoice form filled from an uploaded PDF, every field marked Read from the document and awaiting a person's review
A supplier invoice, read from its PDF. Every field says where it came from. Nothing is saved until a person has looked, and the number and the ledger entry are allocated only on approval.
05 · Fit

Built around the way the business actually works.

Scarlett is configured around the organisation’s own processes rather than forcing it into a fixed workflow.

Configured, not customised
Fields, documents, numbering, approval stages, permissions and terminology are set around the organisation. Roles are built from a catalogue of permissions, so what a role can do is a list you can read.
A module where the industry needs one
Where an industry needs more than the standard capabilities, Hashinclude builds it as a dedicated module inside the system, rather than leaving the process to spreadsheets and disconnected tools.
Industries
Construction, manufacturing, trading, field services, professional services and retail.
The role editor, listing permissions by module with some marked as requiring a person

The role editor. Permissions grouped by module. The ones only a person may exercise are marked, and no assistant can hold them.

06 · Your data

Your data, in your jurisdiction, under your control.

Scarlett is a single-tenant system. Each company has its own dedicated system, database, encryption keys and backups, separated from every other customer’s.

A data-centre corridor: rows of server racks with structured cabling, in low light
One dedicated instance per customer. Its database, its keys and its backups, in the jurisdiction it requires.
Deployed where you need it
In the jurisdiction the customer requires, including in-country UAE deployment. Nightly backups, scheduled restore testing, and administrative access that is time-limited and logged.
Never used for training
Customer records are not used to train an AI model. What reaches a model is what the assistant needs for the task in hand, and with an open-weight model on your own infrastructure, nothing leaves at all.
Connects rather than replaces
REST APIs, webhooks, scheduled data transfers, email capture and single sign-on, so existing applications, payment systems, storefronts and logistics platforms stay part of the stack.
07 · Rollout

A gradual shift, not a system overhaul.

Scarlett is introduced progressively. The first operational area typically goes live around the second week, and the rest follow at the pace the business sets.

  1. Week 1

    Your instance, your data

    The dedicated instance is created and existing data is loaded into it.

  2. Week 1 to 2

    Your processes

    Business processes, permissions and approval stages are configured around the organisation.

  3. Around week 2

    First area live

    The first operational area goes live with the people who run it.

  4. Then

    Progressively

    Purchasing, stock, invoicing, accounts and projects are introduced one after another.

08 · Why we built it

The founder on Scarlett

“Businesses don’t need another AI tool sitting beside the systems they already use. They need technology that understands how the business actually operates and can help move that work forward. Scarlett is built around that idea, connecting the business first, putting AI inside the workflow, and making sure automation remains subject to the company’s own permissions, processes and people.”

Anas Abdul Latheef Founder, Hashinclude

See it with your own numbers.

A demonstration takes an hour. Bring a month of invoices and a stock list, and ask Scarlett what it sees.

anas.exe
> _
$
Powered by Hashinclude AI
WhatsApp