Digital transformation in municipalities: start inside

When people talk about digital transformation in local government, they usually mean the side citizens see: online payments, e-services, mobile apps. Those matter. But while moving the internal operations of a municipality onto software over the past years, I saw something else: the real bottleneck is not on the citizen-facing screens. It is inside the building, in paperwork travelling from desk to desk.

In this article I share the experience from the municipal management system we took live in a municipality in Turkey, written as a roadmap. I will cover the order in which to move, where time gets lost, and where AI actually helps.

The problem is not missing software, it is disconnection

Most municipalities do not lack software; if anything, they have too much. In Turkey, civil servants and contract staff are held in SAYSİS, the state's central personnel information system for local government. Permanent workers sit in an ERP, and employees of the municipality's affiliated companies in a separate accounting package. Each system does its own job well, but none of them knows about the others.

The result is familiar:

• Scattered staff data. Finding an employee's leave balance, department and payroll means checking three different systems.
• Approvals on paper. Leave, transfer and personnel-record forms travel from signature to signature, and nobody knows whose desk a request is waiting on.
• Lost work. Tasks and fault reports are discussed by phone or in messaging groups and then forgotten.
• No overall picture. The mayor's office cannot see the current state of the departments on one screen; reports arrive weeks later.

That is why digital transformation should not be thought of as "buying a new system", but as connecting the pieces that already exist.

1. Data first: one record per employee

Approval workflows, the mobile app, reporting: all of it depends on correct staff data. So the first step is not to replace existing systems, but to pull data from them and merge it into a single record.

Our setup has four sources: SAYSİS, KamuTech ERP (an ERP used by Turkish public bodies), Zirve Muhasebe (an accounting package) and the information the institution enters itself. Records are matched on the Turkish national ID number, and each employee ends up with a single personnel record. Synchronisation runs automatically twice a day, at 06:00 in the morning and at 16:00 in the afternoon.

There is one rule that is easy to overlook: corrections made by the institution must be preserved. When staff or HR correct a piece of information, the next sync must not overwrite that correction with the outdated value from the external source. If you do not set this rule from the start, users stop trusting the system within a few weeks.

2. The fastest visible win: approval workflows

Once the single record is in place, the thing that makes a difference fastest is turning paper approvals into a tracked digital workflow. This is where staff, department heads and the mayor's office feel the change every day.

The key is to model the workflow on the institution's real signing order, not on what is convenient for the software. The leave request, for example, follows this path:

1. The employee picks the type of leave and the dates; working days and the remaining balance are calculated instantly.
2. The department head approves, or rejects with a written reason.
3. HR does a preliminary review, checking regulations and entitlement.
4. The deputy mayor approves. For departments reporting directly to the mayor's office, this step is skipped.
5. With HR's final approval, the leave is confirmed.

Transfers are more complex: the heads of both the releasing and the receiving department sign, as do the deputy mayors each side reports to, and HR. If both sides report to the same deputy mayor, that signature is requested only once. When all signatures are in, the official document is generated, and the transfer is applied to the personnel record automatically on its effective date.

These details look small, but a system that does not match the workflow on paper exactly will not be used. In the analysis phase, most of the time should go into mapping these signing orders department by department.

3. Task tracking and support requests

The third step is to put on record the work that gets lost on the phone and in messaging groups. When a fault report or request is opened, a defined rule based on institution and category should send it to the right team, and until it is closed it should always be clear who is working on it.

This is not only about tidiness. Once requests are recorded, it becomes measurable how long each team takes to resolve them. That is what lets management discuss impressions like "IT can't keep up" or "this department is slow to respond" on the basis of data.

4. Where does AI actually help?

In the public sector, AI is usually discussed at the extremes: either a tool that will solve everything or a risk to stay away from. In practice, the clearest benefit I have seen is much plainer: summarising reports.

When departments enter their weekly and monthly activity reports into the system, AI turns them into a one-page executive summary for the mayor's office. Which reports arrived on time, which department has approvals piling up, what stands out: all of it is visible in a few sentences.

A few principles matter here:

• AI summarises, it does not decide. The summary does not replace the report; it shows where to look.
• Send no more data than necessary. If a summary does not need personal data, that data should not go to the model. In our setup, the system and its data run on the municipality's own servers; only the departments' report text goes to the AI service, no personal data. Under KVKK, Turkey's data protection law, the safest data is the data you never process.
• Permission boundaries hold. Only senior management should see the executive summary; AI output is subject to the same role and permission rules as every other screen.

There are other use cases too, such as classifying citizen requests or answering frequently asked questions. But before moving on to those, internal data has to be in order. AI working on scattered data only produces the mess faster.

5. Security and data protection must be designed in from the start

Municipal systems hold sensitive personal data: identity details, payroll, leave based on medical reports. Security therefore cannot be a layer added later. In our system, these foundations were there from day one:

• Role and permission management: with 7 roles and 69 separate permissions, everyone sees only what their job requires.
• Audit log: every create, delete and update is recorded with who did it and when.
• Strong sign-in: one-time codes by SMS and passkey support with fingerprint or face recognition.
• KVKK consents: privacy notices and user consents are recorded and documented.

A step-by-step roadmap

If you are starting this work in a municipality, this is the order I recommend:

1. Take inventory. Which software is in use, where does staff data live, how do the approval chains work?
2. Build the single record. Connect to the existing systems and merge records on the national ID number.
3. Pilot with one approval workflow. Leave requests are a good starting point.
4. Open mobile access early. Field staff reaching the system from their phones speeds up adoption.
5. Add task tracking and requests. Put tasks and fault reports on record.
6. Add reporting and AI last. Once the data is in order, the summaries become meaningful.

Frequently asked questions

Do we have to replace our existing software?

No. The right approach is to pull data from existing sources such as SAYSİS, the ERP and accounting software, and bring it together in one place. The systems stay where they are; the gaps between them are closed.

How long does implementation take?

It depends on the chosen modules and the systems to be connected. In our first deployment, the first phase went live within three months.

Does the AI process personal data?

It should not where it does not need to. Uses such as summarising activity reports do not require personal data. In our system, too, only the report text is sent to the AI service; staff information never leaves the municipality's own servers.

Conclusion

Digital transformation in municipalities is less about adding new screens and more about connecting existing systems and carrying the real processes on paper over into software accurately. The order matters: data first, then approvals, then task tracking, and AI last.

If you would like to talk about how to set this up in your own institution, you can explore the modules and workflows in detail on the municipal software page or get in touch with me.