I am Asım Serdar Yılmaz (Asim Serdar Yilmaz), a senior legal and regulatory affairs leader based in Istanbul. Across more than twenty years of legal and regulatory work, most of it in-house, I have advised global technology, consumer and regulated businesses across Türkiye and the Middle East: retail and marketplace, streaming and media, advertising, payments, data and IP. My work tends to sit where a business wants to move quickly and a regulator wants to move carefully.
I started in private practice at White & Case in Ankara from 2005 to 2008, on cross-border M&A, privatisations and IP in regulated sectors. From there I moved in-house: first to Philip Morris International from 2008 to 2012, a strict, high-scrutiny compliance environment, across the Middle East and then in global compliance from its Lausanne headquarters; then into e-commerce at eBay’s Turkey and MENA business from 2012 to 2019, and from 2019 to 2026 as Legal Director for Amazon Turkey. In 2026 I moved from legal into regulatory affairs, and today I lead the regulatory agenda within the corporate affairs function of a global group in a highly regulated industry.
The thread through all of it is regulated markets. I have spent years turning regulatory ambiguity into paths a business can actually take, and building what I think of as “yes, if” guardrails: controls that let a team move at speed while protecting the enterprise over the long term. The fields have been competition, consumer protection, payments, data protection, intermediary liability and corporate governance. Much of that has meant sitting across the table from competition authorities, data protection regulators, media regulators and ministries.
The other current chapter is LEXSAS, the AI legal-tech venture I founded in 2026, focused on legal workflow automation and advisory. It blends legal directorship, regulatory affairs, corporate affairs and public policy, now paired with hands-on AI governance. It grew out of years of building generative AI tooling for legal work, on AWS Bedrock among other platforms: tools meant to simplify intake and access to information and to reduce dependence on outside counsel, and years of seeing where adoption succeeds and where it quietly fails. LEXSAS designs workflows for law firms and in-house teams, builds them with your team and runs the first pilot, with the guards in place before the first run. This site is where I write about that intersection of law, regulation and AI, as technology and process consultancy rather than legal representation.
I spent more than twenty years watching good legal work fail to leave a trace. The advice was sound, the decision was defensible, and six months later nobody could reconstruct who had decided it or on what basis. Then generative AI arrived in legal teams and made that older problem urgent, because now something other than a person was producing the material, and the question of what it had read became a question somebody would eventually have to answer under oath or under audit. Building an AI tool inside a legal team showed me the same thing from the other side: the hard part was not the model, or even the tool, but who decides what the tool may do, whose desk the output lands on, and how that is shown six months later. Those three questions are never asked in a product demo.
The market makes two kinds of offer. One sells a tool and leaves the workflow to you; the other sells advice and ends in a presentation. I started LEXSAS in 2026 for the gap between them, to build the missing half. Not better drafting, which the market is solving quickly and well, but the path the work takes: a named file, a written mandate for the software, a person who signs, and a record that survives everyone who touched it. That method is now written down in full, as the LEXSAS operating standard. I do this separately for law firms and in-house legal teams, because the same machine is set up differently at the two desks: the department accounts for volume, the firm for the file.
It starts with one workflow and a call of about an hour: the matter type, the volume, the systems it touches, who will sign, and when you need it. If I do not think a workflow is the right answer, I say so on that call, and there is no charge for it.
What follows is a written scope before anything is built, then a supervised pilot with one measure you set in advance, a named reviewer and a date on which it passes or fails. Design settles the mandate, the record fields and an evaluation set drawn from your own files; in the pilot every output passes a lawyer’s approval and the measure is read weekly; at the evaluation the set is rerun, the result is a pass or a fail, and the decision is yours. The numbered method runs to twelve sections on the operating standard page. Pricing follows that scope rather than a per-seat licence list, and the fee is agreed in writing before work begins. Everything the pilot produces is exportable and yours from the first day, and the exit terms are written into the pilot rather than negotiated at the end.
I work in English and Turkish, remotely and in Istanbul.
LEXSAS is my venture and it is not my only work. I hold a regulatory affairs role in a global group, and LEXSAS runs alongside it, outside that work and in an unrelated field. The number of engagements I run at once is small, and I do not hide it: a pilot starts with a timetable, the timetable is agreed for each engagement, and what I can take on, and when, is agreed for each engagement rather than promised in advance. If I see a possible conflict of interest, I do not take the work, and I say why. If your timetable needs more than that, I will tell you at the scoping call rather than three weeks in.
All views on this site are my own and do not represent any current or former employer.
The eight workflows say what gets done; these six areas say with what legal knowledge. Both are held to the same standard of review.
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. The repetitive part goes to the workflow and the part that needs judgment stays with the lawyer; which is which is settled in writing before the pilot.
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.
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.
Cross-border data transfer and localisation questions under KVKK and GDPR, including undertakings and the practical controls needed when regulatory expectations change.
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.
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.
EXPERIENCE
Those years include statutory board roles and leading legal functions through investigations, licensing questions and market entry. What I say about AI stands on this practitioner ground.
LEXSAS, Istanbul
A global group in a highly regulated consumer industry, Istanbul
Amazon Turkey, Istanbul
eBay Turkey (GittiGidiyor), Istanbul
Philip Morris International, Dubai and Lausanne
White & Case LLP, Ankara (Çakmak Avukatlık Bürosu)
I write about legal workflows, AI governance and regulated digital markets. Five of the essays I publish on this site in English and Turkish are below; shorter notes I first post on LinkedIn are kept on the LinkedIn posts page, and the full index is on the insights page.
A law degree from Ankara University, an MSc in Internet Business from Mercy College in New York, and executive education at Harvard Business School; more recently, AI Law, Policy and Governance at the London School of Economics.
Ankara Bar Association. Languages: Turkish (native), English (fluent).
If any of this is close to a question you are working on, write to me.
APPROACH
I built and ran these workflows inside a legal team before I built them for anyone else. The design starts from the queue as it actually runs on a Tuesday, not from a roadmap.
Well-designed controls let a team move quickly while protecting the business over the long term. A control that slows the work is in the wrong place; in the right place it makes doing the same work twice unnecessary. With the sources fixed, research is done once. With the mandate written, the same argument is not reopened at every pilot. With the record kept, nobody has to reconstruct three months when an audit arrives.
Every AI deployment should begin as a pilot with a defined measure of success and a person owning the output; no output goes out without a lawyer’s review. If it cannot show what it changed, it has not earned a wider rollout.
“The useful answer is rarely a flat yes or no; it is yes, if.”
Asım Serdar Yılmaz, Founder