I · The system · Chapter 1
What PORTWARE is
What an enterprise system built on the world's most widely used open-source business-management core and combined with an office suite and artificial intelligence is, and why it was built that way.
PORTWARE is an enterprise resource planning system. It is built on the world's most widely used open-source business-management core, backed by CapitalG, the growth fund of Google's parent Alphabet — and taken far beyond it with manufacturing, engineering, AI and office layers. The two largest layers: an office suite that edits documents inside the browser (PORTWARE Office) and an artificial intelligence assistant that works with the system's own data (Cowork). Around them are hundreds of modules written to serve the real work of a company.
The system's nameplate
All of these figures were generated by reading the source tree and the production database; none of them was typed in. The module pages show the same data in full, module by module.
Three obstacles to growth
What holds most companies back is not the workload but disorder. Three symptoms appear together:
- No traceability. Who did what, and when, is unknown; information is spread over e-mails, spreadsheets and people's memory.
- No coordination of action and follow-up. A decision taken in a meeting has no owner, follow-up depends on individuals, work falls through the cracks.
- No scalability. As the company grows, processes stay dependent on a few people; adding a team or a branch breaks the order.
PORTWARE builds the order and the system that solve these, and much more: every job, document and decision sits on its record with its history; every decision becomes a task with an owner and a date, and the morning briefing reminds; a new team, branch or application opens in the same system. Growth comes with order.
Why open source and not a closed licence
The decision is deliberate and on record: no closed-licence modules are used. The reasons come down to three points.
- Ownership. The system runs on our own server and its data is on our own disk. Not renewing a licence is not a cost line; it is a decision for independence.
- Changeability. The open-source core can be read and extended. Every missing capability was written as a module instead of being bought as a commercial package — and that is most of the system this book describes.
- Measurement. A closed component cannot be measured. We can see what our own module does and why, and what each test proves.
Three layers
The easiest way to understand the system is to see it as three layers.
1. The core — records and processes
Customer, quotation, order, invoice, product, stock move, employee, project… All of them are records that work to the same pattern: each has a form, a list, a conversation stream (the chatter), attachments and a history. Chapter 4 explains this pattern in detail; once learnt, most of the rest of the system feels familiar.
Processes connect these records: the opportunity leads to the quotation, the quotation to the order, and the order is responsible for delivery and invoice. Chapter 16 follows this chain from end to end.
2. PORTWARE Office — documents
Word, Excel and presentation files are not downloaded and uploaded again; they open inside the browser, on top of the record, and are edited together. ONLYOFFICE Document Server runs underneath, but for the user it is invisible: a .docx attached to a quotation opens when clicked, two people can write at the same time, every save leaves a version and two versions can be compared in colour. Chapters 8 and 9 describe it.
3. Cowork — artificial intelligence
Cowork is not a chat box; it is an assistant that can reach the system's data and call tools. It searches records, reads attachments, writes documents, produces reports and extracts tables from PDFs. It works with the models of eight different providers, and which model does which job, in how much time and with what accuracy, is chosen by measurement. Chapters 12–15.
The company and the real work
PORTWARE did not grow as an abstract product but to serve the work of a real company: an industrial firm that makes servo motors, drives, motion control and CNC machines. That visibly shaped the system:
- Product data is technical. A product's name and price are not enough; measurable values such as torque, voltage class, encoder type, communication protocol and temperature range sit in separate fields on the record — both the website and the selection guide read them.
- Documents are heavy. Datasheet, CAD file, EDS/ESI description file, user manual. That is why attachment management, folders, versioning and the CAD viewer are first-class topics.
- The channel has two tiers. Direct customers and dealers. A dealer sees the same website with their own prices; their request turns into an order. Chapter 28.
A typical request that shaped the system
- Situation
- A dealer wants to see a product on the website at their own price and place an order from there.
- What was done
- No separate dealer website was built. A bridge module was added to the existing corporate website that shows a signed-in dealer their own price list; the basket became a "request", and an approved request became a sales order.
- Result
- One catalogue, one set of product data, one address. The page the dealer sees and the page everyone sees are the same page; the only differences are the price and the order button.
This shows a preference that recurs throughout the book: extend what exists instead of producing a second copy. Two copies drift apart sooner or later, and the question of which one is right appears.
In step with Türkiye's AI Action Plan
The Ministry of Industry and Technology's Türkiye AI Action Plan 2026–2030 (Presidential Circular 2026/9) rests on four axes and sixteen actions: Notice, Benefit, Produce, Govern (Fark Et, İstifade Et, Üret, Yönet). For a company the most concrete axis is Benefit: putting AI into everyday work in the public and private sectors. PORTWARE meets that axis for an SME today, installed:
| Axis | The plan's focus | In PORTWARE |
|---|---|---|
| Notice | AI literacy, talent, the data ecosystem, the legal-ethical framework | An assistant on every screen and this handbook; data in one database, every figure defined |
| Benefit | Compute access, sector pilots, AI transformation | Meeting notes, the morning briefing, the digital team, an assistant that builds dashboards — with no integration project |
| Produce | Domestic models, financing, physical AI and robotics | Built in Türkiye; Turkish interface, voice and e-invoicing; the model on your own server if you wish |
| Govern | AI safety, sandboxes, international cooperation | An assistant that never writes without approval, a trace of every action, a spend ledger, one address for KVKK |
The plan foresees tools that support SMEs' access to AI solutions (for example AI vouchers for SMEs). We review that support with you as its conditions are announced. Source: sanayi.gov.tr/yapayzekaeylemplani.
Frequently asked
Where does our data live? On the server you choose: your own or PORTWARE Cloud. Backup and recovery are written into the contract.
Does AI take our data outside? Only the part the task needs goes out, through the providers' enterprise programming interfaces. For confidential work PORTWARE Local AI is chosen and the data does not leave the server. Web research is one-way: information comes from the web into the company.
Are we locked in? No. The core is open source (LGPL-3), the data is in standard PostgreSQL and the AI provider is a setting.
What about our old data? Customers, products, stock and open balances are imported; data migration is its own step in the implementation project.
Does it work on a phone? Yes: it installs like an app, receives notifications and takes voice.
How do updates arrive? One product, one codebase: every installation takes the next release with its notes and a maintenance window; a customer-specific need lives in its own module and never blocks the update.
The system this book describes is live
Everything you read here runs in production. The behaviour described is not an imagined design; it has been measured, tested and put live. If a feature is not finished yet, the book says so plainly — hiding the difference between "to do" and "done" is the most harmful thing documentation can do.