Intake notes from a first meeting

Why this is the right place to start

Everything downstream of intake — the research, the drafting, the strategy call with the client — runs better if the intake memo is structured. Most small and mid-sized firms in Nepal do not do this consistently. The notes from the first meeting sit in a notebook or a Word file in flowing prose, the junior who picks up the file has to re-read everything before doing anything, and half the time some important detail — the exact date a notice was served, the name of the second heir, the fact that the client already paid a fifty-thousand-rupee retainer to another lawyer — gets lost in the prose and surfaces uncomfortably a month later.

A structured intake memo solves this. The shape of the memo is the same whether the matter is a property dispute, a partnership breakdown, a divorce, or a banking complaint. The headings do not change. Only the contents do. That is exactly the shape AI is good at — take unstructured input, produce structured output that follows a fixed template you supplied.

The intake memo, eight headings

A useful intake memo for a Nepali practice has eight headings and almost never needs more. Parties (each side, role, the relationship between them, contact channels). Facts (a neutral, chronological statement of what happened, in the client’s framing — the model should flag where the client’s account contains an assumption rather than a fact). Timeline (a dated chronology pulled from the facts, gaps marked explicitly). Likely causes of action (what claims the client could plausibly bring, or face — phrased as hypotheses, not conclusions). Statutes likely implicated (the Muluki Civil Code chapters, the Contract Act, the Labour Act, the Banks and Financial Institutions Act, etc. — again, hypotheses to verify). Documents to collect (a checklist — sale deeds, citizenship copies, bank statements, board minutes, WhatsApp screenshots). Conflicts check items (the names and identifiers you must run through the firm’s conflicts log before opening the file). Fee-arrangement starter (the rough shape of the engagement — hourly, fixed, milestone — and the realistic range for a matter of this size in your market).

Eight headings. Same eight headings every time. That consistency is what lets the model help you, and what lets a junior pick up the file two weeks later and know exactly where to look.

The prompt pattern, paste-ready

Once the meeting is over and you have a quiet ten minutes, type your rough notes into a single block — bullet points, half-sentences, whatever the meeting produced — and feed them to the model with a prompt that names the headings. A working pattern, with every detail redacted before you paste:

Below are my rough notes from a first client meeting. Turn them into a structured intake memo with exactly these headings, in this order: Parties, Facts, Timeline, Likely causes of action, Statutes likely implicated, Documents to collect, Conflicts check items, Fee-arrangement starter.

Rules: Use only the facts in the notes. Do not invent any date, name, amount, or document not in the notes. Where the notes contain an assumption rather than a verified fact, flag it in square brackets. Under “Likely causes of action” and “Statutes likely implicated”, phrase every entry as a hypothesis to be verified, not a conclusion. The memo is for internal use, not for the client.

Notes: [paste rough notes here, with full names, citizenship numbers, contact numbers, and the names of any employer or opposing party removed]

The output is a clean one- or two-page memo you spend three minutes reviewing. The full intake-to-memo cycle that used to take forty minutes is now perhaps eight. The structure is consistent across every matter the firm opens, which makes everything downstream — research, drafting, supervision, billing — easier.

Confidentiality first, before any of this works

None of this is safe if you paste the client’s identifiers into a public chatbot. The free tiers of these tools may use your inputs for training. The paid tiers route data through servers outside Nepal that you cannot audit. The professional confidentiality obligation under the Nepal Bar Council Rules does not pause because the tool is convenient, and the client whose citizenship number, full name, employer, and opposing-party details you pasted into a foreign chatbot has not consented to that disclosure.

The practical discipline is redaction at the prompt. Before pasting, replace full names with roles — “the client, a fifty-eight-year-old widow” — replace citizenship numbers with the phrase “citizenship number redacted”, replace bank account numbers and PAN numbers the same way, and replace the name of the opposing party with “Party B, a sole proprietor in the same ward.” The model needs the shape of the situation to produce the memo. It does not need any identifier that would let a stranger reconstruct who the client is.

Four real Nepali matters, one structure

The same eight-heading memo handles every kind of matter a small Nepali practice sees. The contents differ; the structure does not.

A property dispute in Patan with multiple heirs. A widow has come to you because her late husband’s three brothers are claiming shares of an ancestral house and a Lalitpur shop. The Parties section names the widow, the three brothers, and the two minor children whose interests must be protected. Timeline starts at the registration of the original parental deed and runs through the husband’s death and the first informal family meeting. Causes of action are hypotheses — partition under the Muluki Civil Code, declaration of the widow’s coparcenary share, possible injunction against alienation. Documents to collect lists the parental sale deed, the husband’s death registration, the widow’s marriage registration, the citizenship copies of all heirs, the latest land-revenue receipts, and the property tax bills.

A partnership dispute between a Newari merchant and a foreign investor. A trading partnership has broken down. One partner is a long-standing wholesaler in Asan; the other is a Singapore-based investor who wired in roughly forty lakh rupees against a partnership deed nobody now wants to read literally. Parties names both. Facts captures the original understanding versus the documented terms. Timeline runs from the deed to the disputed transactions. Causes of action are hypotheses around breach of the partnership deed, accounting, possible Foreign Investment and Technology Transfer Act issues. Documents to collect includes the original deed, bank evidence of the wire, the partnership’s tax filings, board minutes if any, and the WhatsApp thread the partners actually relied on.

A divorce with cross-border assets. The client is a Nepali national; the spouse holds Australian permanent residency; the matrimonial property includes a flat in Sydney, a fixed deposit in Kathmandu, and shares in a small Nepali company. Parties, facts, timeline as usual. Causes of action are hypotheses around divorce under the Muluki Civil Code, division of matrimonial property, the jurisdictional question of which forum can effectively reach the Australian flat, child custody if relevant. Documents to collect spans the marriage registration, citizenship copies, evidence of each asset on both sides of the border, and any prenuptial understanding.

A banking dispute with an NRB-supervised institution. A small business client has had a working-capital facility unilaterally classified as non-performing by the bank. Parties names the client and the bank, with the bank’s branch and the relationship manager noted. Facts captures the original sanction, the alleged default events, and what the client says actually happened. Causes of action are hypotheses around breach of the facility agreement, possible improper classification under NRB directives, possible writ jurisdiction depending on the bank’s status. Documents to collect lists the sanction letter, the facility agreement, the security documents, the statements, all written communications, and the bank’s classification notice.

Four very different matters. One memo structure. The AI’s job in each case is identical — arrange the rough notes under the fixed headings, flag the gaps, propose the hypotheses, and stop.

Verification before action

The intake memo is not a legal opinion. Everything under Likely causes of action and Statutes likely implicated is a hypothesis the model generated by pattern-matching against general knowledge of legal systems — including, often, Indian and Anglo-American systems whose doctrines do not map cleanly onto Nepali law. Before you advise the client on any of it, you confirm each hook against the actual source. The Muluki Civil Code section the model cited — does it say what the model says it says, in the version currently in force? The Supreme Court authority the model mentioned — does it exist, in the NKP volume cited, with the holding described? The procedural step the model suggested — is that the right court, the right time limit, the right form? You read the source. You do not advise the client on the model’s summary of the source.

This is the same verification discipline the previous chapter on legal research and the citation problem set out, applied at the intake stage. The reason for repeating it here is that the intake memo, because it is structured and confident-looking, is precisely the document most likely to be relied on without re-checking. A junior reads the memo, picks up the matter, and quietly assumes the statute citations are correct because they look correct. They are not citations until you have verified them. Until then they are hypotheses formatted as citations.

Check your understanding

Quick check

—

An advocate has forty minutes of rough notes from a first meeting with a client whose late husband's brothers are claiming shares of an ancestral property. What is the right way to use AI on these notes?

1. AAsk the model to draft the writ petition directly from the notes, since the facts are fresh.
2. BAsk the model to email the client a summary of next steps, to save time.
3. CAsk the model to turn the redacted notes into a structured intake memo under fixed headings (parties, facts, timeline, likely causes of action, statutes implicated, documents to collect, conflicts check, fee-arrangement starter), with every legal hook phrased as a hypothesis to verify.
4. DPaste the notes verbatim into a public chatbot and ask it to "tell me what the law says".

Check answer→

Quick check

—

An advocate is about to paste rough intake notes containing the client's full name, citizenship number, employer, and the opposing party's name into a free public chatbot to produce a structured memo. What should they do first?

1. ANothing — the chatbot will keep the data private because the conversation looks one-to-one.
2. BRedact the identifiers before pasting (replace the full name with a role such as “the client, a fifty-eight-year-old widow”, redact the citizenship number, replace the opposing party with a generic description), or use an enterprise tool with a no-training data agreement.
3. CPaste everything but add the sentence “please keep this confidential” at the top of the prompt.
4. DEmail the client to ask retrospective permission to share their details with a foreign AI service.

Check answer→

What comes next

A clean intake memo in English is still the wrong document to hand a monolingual Nepali client. The next section is about using AI to translate legal language into plain Nepali the client can actually act on — and the specific reason pure translation, on its own, is almost never enough.