SmartQflow
A vertical application platform that adapts across sectors: the same modules take the shape of a new one without being rebuilt.
SCOPE
One team. End-to-end ownership. Our own products.
3SERVICEAREAS
2024PROJECTYEAR
İSTANBUL / KOCAELİ
What we built
- YEAR
- 2024
SmartQflow is not a single product but a base that verticals are built on. The form engine, the permission layer and the personnel record are attached to a shared data model rather than to each other, and each stands up without the others. Adapting it to a sector is not a new release; it is the same modules arranged into a new installation.
The shapes it has taken so far: occupational health and safety, food and hygiene, facility management, security and patrol, vehicles and fleet, retail stores, construction, energy and maintenance, healthcare and cleaning, quality, and on the hiring side a careers portal carrying the posting, the application and the candidate pool. What differs between them is the form, the question types and the scoring weights; the code stays the same. A new vertical is a new installation, not a new version.
The form comes second, because the core was built as a flow rather than as a form. Forms are designed on a visual workbench and questions branch on the answer: saying no at a step opens a photo and a comment, saying yes never shows that step at all. Photo, signature, location, barcode and tag scan are question types too, twenty-seven in all. Ready-made templates arrive with the scoring weights their sector expects, and a template can be drafted by AI. Opening a new kind of check needs a form, not code.
The work happens where there is no connection. A warehouse, a basement, a lift shaft, a construction site, a plant. An app that waits for a connection makes the person on site fill the form later from memory, which corrupts the record itself. So the mobile app keeps records on the device and syncs itself when the connection returns.
In the security and facility verticals the real question is whether the person was actually there. In patrol mode the form does not open until the NFC or QR tag at the point is scanned. Skipped points are reported as missing. The proof comes from the condition under which the form opens, not from what is in it.
Whether a completed record was really carried out is a separate question, and the platform answers it with rules. Implausible completion speed, answers copied from one another, scores that come out perfect every time, consecutive submissions from the same location: the reason behind every flag is written on the supervisor screen. It is not a black-box model, because a system that questions an employee record and cannot show its reasoning is not an acceptable thing to build.
AI lightens the work in two places, and in neither does it have the last word. It reviews photos taken on site and proposes a non-conformity; accepted, it becomes a finding in one click. Someone working in gloves or up a ladder speaks, and the model maps the answers to the questions, but nothing is saved until it is approved. An enterprise customer can supply its own AI key, and then the processing stays in its own account.
When a non-conformity is marked, an action opens by itself and lands on whoever owns it. It is closed with photographic evidence, but it does not count as closed until a manager verifies it. Weighted scoring, trends and root-cause analysis sit on top of that loop, and the map makes location-stamped submissions readable region by region. What a manager sees is not an archive of forms but a list of open findings.
What decides who a task can be assigned to is the personnel record. Certificates, training records and expiry dates live there, so a job that requires a given competency does not open for someone who lacks it or whose certificate has lapsed, and a certificate approaching expiry reaches the manager in advance.
The hiring side is a vertical built on the same logic. The careers portal carries the posting, the application that comes in and the candidate pool. The application form is built on the visual workbench like every other form, so it asks what the sector actually asks: which certificate, which training, valid until when. Because certificates and expiry dates are already fields on the personnel record, what finds a candidate in the pool is the competency itself rather than a free-text search. From healthcare staff to field crews, a filter that runs on the sector’s own fields gives a precision a general-purpose job site cannot. The same installation stands up as an organisation’s own careers page or as a standalone listing and candidate-pool site opened to a sector, and it extends as the scope grows through custom development integrated into the existing modules. And whoever is hired does not move to another system: their onboarding runs on the platform’s own form infrastructure, because an onboarding checklist is already a form.
The enterprise layer is the condition for the product being usable inside an organisation. Single sign-on, location-based permissions, an immutable audit log, and on the KVKK side retention periods, full data export and anonymisation. The panel and the forms run in three languages, and in the right-to-left one the interface turns with them. For customers whose data cannot leave the organisation, the whole platform can be installed on their own infrastructure.
Highlights
Three separable modules on one shared data model: form engine, permission layer, personnel record
A visual flow workbench with twenty-seven question types and conditional branching, opening a new vertical without code
A mobile app that keeps records on the device and syncs itself when the connection returns
Patrol mode, where the form opens only once the NFC or QR tag at the point is scanned
Rule-based, explainable verification that shows the supervisor why a record was flagged
An AI layer that proposes findings from photos, gates voice filling behind approval, and can run on the organisation’s own key
Competency logic that blocks assignment to a lapsed certificate and continues from hiring through onboarding on the same record
A candidate pool and application flow filtered on competency and certificate validity, standing up either as an organisation’s careers page or as a standalone single-sector listing site
An enterprise layer in three languages with RTL support, in the cloud or on the organisation’s own servers
Technologies used
- A form engine that holds conditional flow as data rather than as a screen
- An offline architecture that treats the device as a temporary record authority and resolves conflict at sync time
- An on-site verification layer that makes a physical tag scan the condition for the form opening
- Separable modules attached to a shared data model rather than to each other, which is what every vertical is assembled from
- A verification layer that produces its flags from rules and never hides the reasoning behind one
- An AI layer that produces the suggestion and leaves the decision to a person, able to run on the organisation’s own key
- A hiring layer that filters the candidate pool on certificate, training and validity fields instead of free text
- A multi-language structure where the interface, the form content and the writing direction are managed separately
Frequently asked questions
What is SmartQflow, exactly?
A vertical application platform that adapts across sectors. The form engine, the permission layer and the personnel record are three separate modules on a shared data model, so adapting it to a sector is not a new release but the same modules arranged into a new installation. Today it carries ten verticals, from occupational safety and food to facility management, fleet, retail and construction, plus a careers portal on the hiring side.
Does adding a new kind of check require development?
It does not. Forms are designed on a visual workbench and questions branch on the answer: no opens a photo and a comment, yes never shows that step. Photo, signature, location, barcode and tag scan are question types too, twenty-seven in all. There are ready-made templates with the scoring weights their sector expects, and a template can be drafted by AI.
How does it work where there is no internet?
The mobile app does not wait for a connection. It keeps records on the device and syncs itself when the connection returns. The reason is simple: the work happens where there is no connection. A warehouse, a basement, a lift shaft, a construction site. An app that waits makes the person on site fill the form later from memory, which makes the record worthless.
How is it proved that the record was made at the point itself?
In patrol mode the form does not open until the physical NFC or QR tag at that point is scanned, and skipped points are reported as missing. In the security and facility verticals the question is not whether the form was filled in but whether the person went there, and the proof comes from the condition under which the form opens rather than from its contents.
How do you know a completed record was really carried out?
The platform measures it with rules: implausible completion speed, answers copied from one another, scores that come out perfect every time, consecutive submissions from the same location. The reason behind every flag is written on the supervisor screen. No black-box model is used, because a system that questions an employee record and cannot show its reasoning is not acceptable.
Where does AI come in, and does it decide anything?
In two places, and it decides nothing. It reviews photos taken on site and proposes a non-conformity, which becomes a finding in one click if accepted. Someone working in gloves or up a ladder speaks, and the model maps the answers to the questions, but nothing is saved until it is approved. When an enterprise supplies its own AI key, the processing stays in its own account.
What happens when a non-conformity is marked?
An action opens by itself and lands on whoever owns it. It is closed with photographic evidence, but it does not count as closed until a manager verifies it. Weighted scoring, trends and root-cause analysis sit on top of that loop, and the map makes location-stamped submissions readable region by region. What a manager sees is a list of open findings, not an archive of forms.
What decides who a task can be assigned to?
The personnel record. Certificates, training records and expiry dates live there, so a job requiring a given competency does not open for someone who lacks it or whose certificate has lapsed, and one approaching expiry reaches the manager in advance. The same record continues on the hiring side, where a new joiner’s onboarding runs on the platform’s own form infrastructure.
Can the careers portal run a sector-specific job site?
It does. The careers portal carries the posting, the application and the candidate pool; the application form is built on the visual workbench, so it asks what the sector actually asks, and what finds a candidate is not free text but certificates, training and expiry dates. From healthcare staff to field crews, a filter that runs on the sector’s own fields gives a precision a general-purpose job site cannot. The same installation can stand up as an organisation’s careers page or as a standalone listing and candidate-pool site opened to a sector. As the scope grows, from a single organisation to a single sector and on to a marketplace open to every sector, the same base extends through custom development integrated into the existing modules.
Our data cannot leave the organisation. Is that possible?
It is. The platform can be used in the cloud or installed on the organisation’s own infrastructure. In both, retention periods, full data export, anonymisation and an immutable audit log are part of the enterprise layer, working alongside single sign-on and location-based permissions.