Merkys.ai

Your institution's procedures and the law, in one chat.

Who is responsible for what, what to do when something happens, which procedure applies: the answers are usually already written down in internal rules, descriptions and orders. Staff just can't find them, so they ask colleagues, guess, or each do it their own way. Meanwhile the law changes, and internal procedures fall behind.

Merkys brings the institution's internal documents together with healthcare law and answers in plain language, with a source for every claim. Agents follow the legislation the Ministry of Health is preparing and check whether the institution's documents and procedures are in line with it. All on your own servers.

Merkys answering a question about a contract notice period, with two numbered sources and a note that the answer is grounded in 2 sources
A simple question, an answer with its sources. From merkys.ai.

How it works

From question to checked answer.

  1. 01

    Connect sources

    Upload documents, or connect SharePoint, Confluence, Google Drive, Teams, email or a website. Preview and confirm exactly what gets indexed. Access rights carry over.

  2. 02

    Ask

    In Lithuanian or English, in the web app, Slack or an embedded widget. Switch to Legal Review or Complaint Resolver from the same box.

  3. 03

    Retrieve and verify

    Keyword and semantic search find the passages. A second model checks the answer against them, and contradictions between sources are flagged.

  4. 04

    Answer or abstain

    Every claim links to its source. When the evidence is too weak, Merkys says so instead of guessing.

Problems it solves

What it solves.

  • 01

    The answers are written down, but out of reach.

    What happens today

    Internal rules, descriptions, orders and instructions sit in shared folders, email and the intranet. Finding the right one takes hours, if it's found at all.

    What it does

    One chat over all the institution's documents. A plain-language answer with a link to the exact place in the document.

  • 02

    It's unclear who is responsible for what.

    What happens today

    When something happens, staff don't know which procedure applies or whose responsibility it is, so everyone does it their own way.

    What it does

    Merkys answers which procedure applies and who is responsible, from the institution's own documents, and flags it when documents contradict each other.

  • 03

    The law changes; internal procedures fall behind.

    What happens today

    The Ministry prepares new legislation and amendments, and the institution finds out once its procedures no longer comply.

    What it does

    Agents follow legislation being prepared and coming into force, check whether the institution's documents and procedures are in line with it, and send alerts by email.

  • 04

    Questions about what a patient is entitled to.

    What happens today

    When staff can't quickly say what a patient is entitled to, the answer becomes a liability and a source of complaints.

    What it does

    Answers on procedures, reimbursement and the Ministry of Health legal base, each traceable to its source.

  • 05

    Contracts and complaints take a lawyer's time.

    What happens today

    Checking a contract against the law and mandated templates, or assessing a patient complaint, takes days.

    What it does

    In the Healthcare edition, the Law Agent checks contracts clause by clause and the Complaint Resolver assesses a complaint from both sides and drafts a reply.

  • 06

    Generic AI tools aren't safe here.

    What happens today

    General chatbots invent answers, know little Lithuanian medicine, and put data outside the institution's control.

    What it does

    Answers only from your sources and abstains on weak evidence. Local-only mode keeps every AI call inside your environment.

Capabilities

What Merkys does.

Answers from your documents

Staff ask in plain language; Merkys finds the answer in the institution's documents and shows the source.

  • A source on every answer
  • Warns when documents contradict each other
  • A second model checks the answer against the sources
  • Abstains when the evidence is too weak
  • Separate assistants for different departments

Healthcare edition

On top: contracts and complaints.

The Healthcare edition adds two specialised agents for legal and quality staff.

Example

A contract checked against the law, clause by clause.

Paste the clauses of a data-processing agreement. The Law Agent finds the statutes and the state-mandated template, then returns a verdict with its reasoning and a finding for each clause.

Here it finds the agreement non-compliant: it lets the processor change the scope of patient data, bring in sub-processors without consent and keep data indefinitely. Results are labelled as an internal, non-final review.

Merkys Legal Review in Lithuanian: five data-processing clauses checked, 24 excerpts found, verdict non-compliant, with clause-by-clause findings
Legal Review of five data-processing clauses. Lithuanian interface.

Example

A patient complaint, assessed from both sides.

Paste the complaint or attach the file. If key facts are missing, Merkys asks for them instead of guessing. It argues the patient's side and the clinic's side separately, then derives the verdict from those findings, so the verdict can't contradict its own reasoning.

Every argument names the law it relies on, and every assumption is stated. Staff review the assessment and edit the draft reply before anything is sent.

Merkys Complaint Resolver in Lithuanian: verdict that the patient's complaint about a dental hygienist filling a tooth has legal basis, with arguments for the patient and for the clinic and the assumptions made
Complaint Resolver, citing the dental hygienist norm MN 35:2019. Lithuanian interface.

Talk to us.

Tell us which product interests you, and we'll get back to you.