Legal tech and AI workflow automation
Designing and advising on AI-enabled tooling for legal work: intake, document and knowledge workflows, research and compliance operations. The aim is fewer manual steps without a lower standard of review.
AI LEGAL TECH
Legal workflow automation, with the guardrails built in
LEXSAS is the AI legal-tech venture I founded in 2026. I design and advise on tools that take the manual work out of legal and regulatory operations, from intake and document workflows to research and compliance, with privacy, IP and governance guardrails in place from the start.
FOCUS AREAS
Six areas, one operating standard
From intake automation to digital market governance, every engagement ends with controls the team can run day to day.
AI governance and regulatory compliance
Privacy, IP and governance guardrails that ship with the tool rather than after it, alongside model oversight, human review and documentation. I work on the line between what a policy says and what a team can actually operate.
GenAI adoption for legal teams
Legal operations strategy for generative AI: where to start, what to build versus buy, and how responsible adoption actually happens. I have co-developed GenAI legal tooling in-house on AWS Bedrock and run the pilots that followed.
Data protection and data sovereignty
Cross-border data transfer and localisation questions under KVKK and GDPR, including undertakings and the practical controls needed when regulatory expectations change.
E-commerce and platform regulation
Marketplace and platform rules across consumer protection, advertising, intermediary liability and payments touchpoints, including notice and takedown under Law No. 5651. Licensing perimeter questions and the controls that keep a platform compliant at scale are part of the same ground.
Competition and digital markets
Governance, controls and strategy advisory for digital markets, covering parity clauses, ranking practices, pricing algorithms and bundling, with the decision records that show how each was set and who can change it.
APPROACH
Legal workflow management and agent design
Legal workflow management
Intake, documents, research and compliance run as named stages, each with an owner, a clock and a record. That is the difference a managed workflow makes: when a regulator, a client or the board asks afterwards, you can show who touched what, when, under which control.
Workflow agent design
An agent gets its mandate the way a new team member does: which tools it may hold, what it may read, where a person signs. That capability set is fixed before the pilot starts, re-checked on every model, prompt or connector change, and never wider than the matter needs.
What in-house teams and law firms gain
For an in-house team, hours on routine intake and first-pass review drop while every output keeps a named reviewer, so the record defends itself. For a law firm, matters move faster without diluting review, and the same records let you price the work, not the hours.
AGENTIC WORKFLOWS
A named file, a locked mandate, a signed record
This is the run itself: what an agentic legal workflow looks like on the desk, from the first loose instruction to the entry nobody can quietly rewrite.
Intake becomes a named file.
Mail, attachments and notes arrive loose. The run binds them to one matter: an owner, a clock, a type, a place in the queue.
The agent works inside a fixed mandate.
Before a run starts, the capability set is written: which sources it may read, which memory it may keep, which tools it may hold. The agent does not leave that frame.
A person signs; the record stays.
Output stops at a named reviewer. A compliance gate sits on the same line. What ran, who signed, under which control, is written once and not rewritten.
LATEST WRITING
Notes on law, regulation and AI
- July 2026Essay Who keeps the agent? Custody clauses for an AI pilot An old custody rule, read as charter language: a named keeper, care you can prove, and a recourse trail for whatever startled the agent.
- July 2026Essay KVKK standard contracts: what the first full year's 2,497 notifications say The five-business-day filing was never the hard part. The Authority now holds a register of transfers, and it has started reading it.
- July 2026Essay The EU AI Act on 2 August 2026: what actually applies Regulation (EU) 2026/1744, the Digital Omnibus on AI, entered into force on 27 July and moves the high-risk dates to 2027 and 2028. Article 50 duties still apply from 2 August.
- July 2026LinkedIn When an AI agent breaks out of the sandbox, who answers?
- July 2026LinkedIn Open weights as geopolitical insurance
FAQ
Common questions
LEXSAS is an AI legal-tech venture focused on legal workflow automation and advisory. It designs AI-enabled tooling for legal and regulatory teams, from intake and document workflows to research and compliance operations, and advises on responsible adoption with privacy, IP and governance guardrails built in.
Legal technology and AI governance, regulatory affairs and corporate affairs in regulated digital markets, e-commerce and platform regulation, data protection under KVKK and GDPR, and competition questions in digital markets. The work is grounded in more than 20 years of legal and regulatory experience, most of it in-house, across global technology, consumer and marketplace businesses.
Asim Serdar Yilmaz, founder of LEXSAS and a senior legal and regulatory affairs leader with more than 20 years of experience, most of it in-house and including Amazon and eBay, following private practice at White & Case.
Yes. Advisory, writing and speaking are available in both languages, from Istanbul and remotely.
No. LEXSAS provides informational consultancy on legal technology, AI governance and regulatory strategy. It is not a law firm and does not provide legal representation.
CONTACT
Let’s talk.
serdar@lexsas.comAdvisory on AI governance, regulated digital markets and legal operations. Speaking and board conversations are welcome, in English or Turkish. When you write, a short outline of the workflow, team and timing you have in mind helps, without any confidential detail.