The automation layer is missing - not the system.
Your SAP, Netsis, Canias, local CRM or Microsoft stack already works. Someone still pulls the report by hand. We hand that work to a layer built on top, without changing the system.
Not a proposal, a conversation. We start from whichever report you pull by hand today.
You are not behind. Nobody got around to writing the layer.
Your team prepares the same report by hand every week. The person who knows the ERP pulls the data, the person who is good at presentations tidies it up, and they are not the same person. The data gets moved again every month and none of it becomes a lasting process.
Teams living with this usually think they have fallen behind. You have not. You have the system and you have the data. Nobody was simply positioned to write the layer in between.
No company should have to call three people to reach its own data.
What is an automation layer?
An automation layer is a piece of software that sits on top of your existing business system - whether that is SAP or a Microsoft stack - and runs the repetitive work a person does by hand, with the same permissions and without changing the system itself.
- No upgrade project: the layer works with your current version.
- Your authorization model stays intact: the layer sees nothing beyond what the person doing the job today can see.
- Reversible: switch the layer off and the system is exactly as it was on day one.
Systems we work on
We work on processes that run by hand not because they cannot be automated, but because nobody wrote the code.
SAP
- reporting, order and stock processes, through official interfaces (BAPI, OData, RFC) or scheduled output.SAP/Netsis
- reports moved into Excel and reconciliation work at companies running SAP/Netsis.SAP/Canias
- repetitive production and stock reports on SAP/Canias.Local CRM systems
- processes where records are carried into reports and other systems by hand.Microsoft 365 and Power Platform
- Copilot, Copilot Studio and Power Automate; making the licences you already pay for work on your team's real jobs.
Three ways we work
Implementation training
Not a generic walkthrough. We teach your team to build the automation layer on their own jobs, in your own stack, and the training ends with something working.
Four programs for companies on Microsoft 365, one for every role
- AI for leaders - executives and department heads: agent strategy, governance, KPIs, choosing use cases.
- Copilot every day - all employees: prompting principles, Researcher and Analyst, saved and scheduled prompts.
- Agent builders - department champions: Copilot Studio, Power Automate, connectors, testing and publishing in Teams.
- Agent engineers - product owners: Microsoft Foundry, Agent Framework, M365 Agents SDK, Copilot Control System.
Hands-on, in your own Microsoft environment, with your own scenarios. Aligned with Microsoft Learn paths and certifications.
One-to-one collaboration on a process
One process, a clear definition, a measurable result. We prepare the report end to end with the person who does the job, extract the unwritten rules, and build the layer.
Example scenarios: hours a finance team spends on SAP exports, how long production waits for the morning numbers, average resolution time in a support team. Before anything is built, every scenario gets one KPI and one owner, and results are measured against your own before-and-after numbers, not vendor averages.
Product development partnerships
Where a repetition deserves a product, we build a custom one.
Example: Marka Araçları, a product for trademark agents that automates TÜRKPATENT's bulletin and ePats checks.
Products already in use
Marka Araçları
TÜRKPATENT publishes a new trademark bulletin every week, and someone read the PDF line by line to find applications with no legal representative. Tracking objections means logging into e-Devlet and checking every application number one by one on ePats - a process one office has to run roughly 8,000 times a month.
Marka Araçları reads the bulletin and pulls out unrepresented applicants on its own. Give it a list of application numbers, log into e-Devlet yourself, and it runs the rest: checking hundreds of numbers on ePats one by one and saving the results. If you want, it also searches the web to figure out how to reach those people.
The e-Devlet login cannot be automated; the login stays with you, the repetition goes to the machine.
Who it is for: Trademark agents and offices scanning bulletins by hand for unrepresented applicants; paralegal teams tracking objections manually every week.
Bakımcı
At a manufacturing plant, maintenance documents were spread across folders, files and people. Finding the right information to fix a fault took a very long time, and sometimes it wasn't possible at all. Teams occasionally mixed up the maintenance schedules of two machines.
Bakımcı brought all of those documents into a single chat assistant. When a technician is unsure, they type the question and get the answer from the plant's own documents. Name the broken machine and its spare parts list comes back directly. The assistant is reachable from anywhere, so nobody has to walk back to the office to look something up.
Answers come from the plant's own maintenance documents, not from the internet.
Who it is for: Maintenance teams and technicians at manufacturing plants whose maintenance documents are scattered.
Three steps to start
Discovery call
Together we pin down which report or task you run by hand today and which system it sits on. No proposal, a conversation.
Pick one process and write it down
Not the biggest one, the one that repeats most often. We go through it end to end with the person who does it and surface the unwritten rules.
Build the layer and hand it to your team
The layer runs on top of your existing system with the same permissions. It tells you when it fails, and your team knows what runs and why.
Starting company-wide: learn, choose, build together
A three-stage roadmap for companies that want to start with several departments rather than a single process.
Stage 1 - Learn
AI in everyday work
A short survey of AI knowledge with leaders and selected people, then training shaped by the results: what AI is, what it can do today, where it helps.
Result: everyone starts from the same level.
Stage 2 - Choose
Choose your agents
Each department proposes 2-3 agents for itself or another team. We help sharpen every idea by value, data needed and effort. Your leadership chooses what gets built.
Result: a shortlist chosen by your leadership.
Stage 3 - Build
Build together
We build the chosen agents side by side with your team. Your team learns by building, and the agents go live in the tools you already use.
Result: working agents and a team that can build the next ones.
Four rules we follow on every project
People first
We start with training, so ideas come from people who know what AI can do.
Your ideas, your decision
Departments propose the agents, and your leadership decides which ones get built.
Together with your team
Your team builds with us, so the know-how stays with you after we leave.
Safe from day one
Everything runs in your environment, with your permissions, and nothing goes live without approval.
About
Akyalabs builds the automation layer on top of SAP, Netsis, Canias and Microsoft-based systems. It uses process knowledge from several industries to take work off teams without changing the system underneath.
Can Kuşaklı
Can reads a problem from inside the process first, not from the system, and then builds the answer hands-on, next to the team that does the work. Can's background spans industrial applications in mechanical and automation, open banking solutions in the UK and across Europe, and then crypto and finance. Today that experience comes together with AI to move the work companies still run by hand into a layer that sits on top of their existing systems.
Bengisu Gültekin
Bengisu has spent many years building scalable backend systems and payment integrations in the financial sector. Bengisu has developed high-performance, regulation-compliant APIs with teams from different cultures in fast-paced environments, and has recently turned to AI automation and intelligent business process solutions.
Which report do you pull by hand today?
Let's start the conversation there. Reach us through the channels below or the discovery call link.
Frequently asked questions
What is an automation layer?
An automation layer is a piece of software that sits on top of an existing business system such as SAP and runs the repetitive work a person does by hand, with the same permissions. It adds nothing to the system; it uses the existing usage surface the way a person does. Switch it off and the system is exactly as it was, so no upgrade or permission change is needed.
Can you automate SAP reporting without replacing SAP?
Yes. The person who pulls the report by hand today already uses an interface SAP provides, and the automation layer uses that same interface with the same permissions. The connection is made through SAP's official interfaces (BAPI, OData, RFC), scheduled output, or as a last resort the screen itself. The age of your version matters less than which access path is open.
We are moving to S/4HANA. Should we wait before automating?
No, waiting usually costs you twice. The layer needs a review after the migration, but its lasting value - the process written down - is untouched by it. A company with documented processes also enters the migration better prepared. Starting today gives you back today's time and means you go into the migration with documentation.
What does Akyalabs actually do?
We work in three ways: implementation training that teaches your team to build the automation layer, one-to-one collaboration on specific processes, and custom product development partnerships where it makes sense. Our focus is processes running by hand on SAP, Netsis, Canias, local CRM systems and Microsoft stacks. The starting point is usually a single report or process.
Does every automation need AI?
No. If the steps are fixed, the filters are known and the format does not change, it is a scripting job, and adding AI only raises cost and uncertainty. AI comes in where steps are not fixed, such as reading an order that arrives by email, matching product codes or catching exceptions. A well-built layer decides at each step which one is needed.
How long does it take and what does it cost?
Cost depends on the number of processes and how many exceptions each one has. A single report with fixed steps is very different from a matching process with dozens of exceptions. That is why the first step is narrowing the scope: one process, a clear definition, a measurable result. Timeline and pricing are discussed once that scope is clear.