Everything your IP practice needs, structured to perform.

Trademark Prosecution and Enforcement

One continuous workflow from filing through to enforcement. Legwork carries a trademark across examination, opposition, registration and every action that follows, holding the deadlines, the precedent and the correspondence on a single matter record instead of across five inboxes.

• Examination reports. Objection-by-objection analysis under Sections 9 and 11, with each ground mapped to the response it needs.
• Written submissions. Hearing-ready submissions drafted in your firm's format, citing your own earlier arguments.
• Opposition workflows. Notices, counter-statements and Rule 45 to 47 affidavits tracked as one sequence, not separate files.
• Infringement analysis. Structured comparison of marks, goods, trade channels and consumer overlap.
• Enforcement workflows. Cease and desist letters, replies, settlement drafts and the follow-through after them.
• Matter-level intelligence. Every deadline, filing and communication on the mark, readable in one place.

Dilution Searches

A dilution search that reasons the way an examiner does. Legwork surfaces phonetically, visually and conceptually proximate marks across the register, then explains where the real risk sits rather than handing back an undifferentiated list.

• Similarity across three axes. Phonetic, visual and conceptual proximity assessed separately, then read together.
• Cross-class coverage. Allied and cognate goods pulled in, so the conflict outside your class is not missed.
• Market context. Trade channels, actual use and proprietor history around each conflicting mark.
• Full register status. Registered, pending, opposed and abandoned marks in a single view.
• Risk rating with reasoning. Every rating carries the argument behind it, ready to be challenged.
• Client-ready output. A search report in your firm's format, not a raw export.

Dilution Search Matrix

Where the dilution search argues, the matrix counts. Legwork distils raw Registry exports into a class-by-class count of every third-party mark carried on the register — registered, pending, opposed and invalid — with the client's own group filings recognised and set aside. Every figure is counted from the register itself, never generated, so every figure can be defended.

• Class × status grid. Every class searched — all 45 if need be — counted across four register states, with totals down both axes.
• Counted, not generated. The arithmetic is deterministic code over the Registry's own exports. AI never touches a number, so every cell traces back to a register row.
• Group-aware exclusion. Entity resolution spots the client's subsidiaries, national arms and spelling variants among proprietors, and sets their marks aside — the matrix shows genuine third parties only.
• Two views, one toggle. The matrix excluding the client's own portfolio, and including it — the dilution picture and the full-register picture, side by side.
• Registry-fluent parsing. Twenty-odd raw statuses normalised to four, multi-class placeholder rows resolved, duplicates across per-class exports collapsed — no manual clean-up.
• A workbook, not a screenshot. Client-ready Excel: both summary matrices plus a drill-down sheet per status, every row carrying application number, mark, proprietor and status.

1. Search the register The word element is run across every class that matters — live through Legwork's Registry search, or from the Registry's own Excel exports, uploaded one per class. Either way, the matrix starts from the public record.
2. Read every export Workbooks are parsed as the Registry wrote them: headers found wherever they sit, columns resolved by name, multi-class placeholder rows attributed to the class actually searched. An application appearing in two exports is counted once.
3. Sort the lifecycle The register speaks in twenty-odd statuses — Formalities Chk Pass, Marked for Exam, Accepted & Advertised. The matrix speaks in four: Registered, Pending, Opposed, Invalid.
4. Recognise the client One narrow AI pass answers one narrow question: which proprietors on the register are the client's own group? Red Bull GmbH, Red Bull India Pvt Ltd and REDBULL are one company; an unrelated business that merely shares a dictionary word is not.
5. Count and deliver Third-party marks are tallied class by class into the matrix, and the analysis ships as an Excel workbook — two summary views, a detail sheet per status — filed against the same matter as the search that produced it.

How crowded is the register? sits behind half of trademark strategy. A word carried by three hundred third-party marks is common to the trade — hard to enforce, easy to design around, and a ready answer to a citation objection. A word carried by three is worth protecting hard. The matrix is that answer, counted: per class, per state, on one page.

Search Strings from TMO Website

The TMO public search, driven for you. Queue the strings — word, class and match type, row by row — and Legwork logs into the portal with your credentials, runs each one in sequence, and brings home every “Export to Excel ALL” file, labelled and filed to the matter. You answer the captchas; Legwork handles the navigation, the waiting, the exporting and the filing.

• Queued, not retyped. Any number of strings in one run, each with its own word, class and match type — executed in sequence on a single login.
• The portal's own match types. Start With, Contains, Match With and Phonetic — the TMO's real Wordmark and Phonetic searches, not an offline approximation.
• Human gates stay human. Captchas and OTPs are never solved by the machine — they are lifted into the app for you, and the prompt follows you as a notification while the run carries on.
• Filed, not downloaded. Every export auto-saves to My Work → Search Strings, stamped with date and time, labelled word · class · type. A nil return is recorded too.
• One-click consolidation. A finished run merges into a single de-duplicated workbook — live marks split from dead ones — saved beside the originals.
• A dilution sweep built in. Flip one toggle to run a word across a whole class range, one export per class — then feed the lot straight into the Dilution Search Matrix.

1. Queue the strings Word, class (1–45) and match type per row — Start With, Contains, Match With or Phonetic. Add as many rows as the matter needs, or switch to Dilution Search for one word across a class range.
2. One login — yours Legwork opens the TMO public-search portal and signs in with your registered email or mobile. The captcha and the OTP come to you inside the app; the machine never answers a human-verification gate. One login then covers the entire run.
3. Every string, in sequence For each row: select the search type, fill the wordmark, set the class, wait for your captcha, run the search, click “Export to Excel ALL”. The portal's pace is Legwork's problem, not yours — switch tabs and the prompts follow you as notifications.
4. Everything filed Each export lands in My Work → Search Strings the moment it arrives — stamped with date and time, labelled word · class · match type. “No records found” is recorded as an answer, not dropped: a nil return is still a result.
5. One workbook, if you want it The consolidator merges a run's exports into a single de-duplicated workbook, live marks split from dead ones. A dilution sweep goes one step further — straight into the Dilution Search Matrix, nothing re-keyed.

A register search is clerical work wrapped around two human moments — the captcha and the OTP. Legwork automates the wrapper and leaves the human moments where the portal insists they belong: with you. What used to be an afternoon of portal babysitting — logging in, re-picking radios, waiting out slow pages, exporting, renaming, filing — becomes a queue you set once and a handful of captchas answered from wherever in the app you happen to be working.

Analytics Tab

The question every partner asks in month two — what is this actually doing for us, and what does it cost? — answered as a ledger, not a slide. Every task Legwork completes logs itself: what ran, what it produced, how long it took, what the API call cost. You add one number — how long the job takes by hand — and the saving multiplier is computed from your estimate, not our marketing.

• A ledger, not a dashboard. Every completed task — drafts, register runs, status fetches — files itself day by day. Nothing is entered twice; the rows come from the same store My Work reads.
• Timed to the second. Each task records the wall-clock time it actually took, from start to finished output — 42 seconds or 1h 05m.
• Costed to the fourth decimal. Per-task API spend priced from each call's real token usage at list rates, input and output counted separately.
• You supply the denominator. Time Expected — how long the task takes by hand — is the one column a human fills in. The saving multiplier is computed from your number.
• Honest dashes. Anything untracked shows a dash, never an estimate. The ledger would rather admit a gap than invent a figure.
• Sliceable by date. One day or any range — today, the last seven days, the billing month — with every total recalculated for the slice.

1. Every run logs itself Finish any task — an examination-report response, a Search Strings run, a status fetch, a dilution matrix — and it appears in the ledger with a timestamp, the tool used, and a description of what was produced.
2. Time is recorded Each run carries its own duration, measured from start to finished output. No timesheets, no reconstruction — the number was taken while the work happened.
3. Cost is metered Every API call's actual token usage — input and output, counted separately — is priced at list rates and attached to the task that spent it. A ₹40,000 question answered per row: $0.83, $0.02, —.
4. You supply the denominator Type one number against any row — Time Expected, how long that task takes by hand. It is the ledger's only manual column, saved per activity, and nothing pre-fills it for you.
5. The multiplier reports itself Time Saving is computed — 29x, 4.2x — and the summary keeps the running totals: activities, time worked, time saved, spend, tokens. Filter any date range and the totals follow.

Legal AI is sold on time saved and almost never audited on it. This tab is the audit. The drafting run that took four minutes six seconds against the associate's two hours is a real row, on a real date, attached to a real deliverable — and the row where nobody has entered an estimate simply shows a dash. Figures like these survive the partners' meeting because each one can be clicked back to the work it describes.

On the roadmap, not in the tab yet: matter-stage analytics, objection patterns by examiner and office, portfolio composition and renewal exposure. Everything above is what ships today.

Email Automation

The correspondence around a matter is high in volume, low in variance and reliably the first thing to slip. Legwork watches the inbox as mail arrives, sorts every email by what it needs — a formal filing, a plain reply, or nothing — and writes the replies it can in the firm's own register. Each one lands in Gmail Drafts, threaded and formatted, waiting for the only click Legwork cannot make: Send.

• Triage on arrival. Every incoming email is sorted into three lanes — needs a formal document, answerable by email, or noise — with a one-line reason shown against each call.
• Auto-drafted replies. Enquiries on procedure, costs, timing and status are drafted unprompted — analysed first for who is writing, which marks and what deadline — and saved to Gmail Drafts.
• Reporting emails from the documents. A journal advertisement, examination report or registration certificate arriving as a PDF becomes a client update drafted from the document itself — particulars extracted, never invented.
• Deadlines by Indian practice. Opposition four months from journal publication, non-extendible; examination responses one month under Rule 33; renewal windows and non-use dates — computed, and marked calculated or stated.
• The firm's register, not a template. Eight correspondence types in house style — the captioned-matter openings, the measured British-English body, the closing assurances — in the firm's first-person-plural voice.
• Drafts, never sends. Legwork connects to Gmail with read-and-compose permission only — no send scope exists on its token. Replies thread onto the conversation and the send button stays with the attorney.

1. Connect Gmail — least privilege Consent happens on Google's own screen, and Legwork asks for exactly two permissions: read the inbox, create drafts. Send is never requested, so sending is not merely disallowed — it is impossible.
2. Watch the inbox The server checks for new mail every 45 seconds and wakes the app the moment something lands. The backlog that gathered overnight can be triaged in one pass.
3. Classify, with reasons Each email goes into one of three lanes — formal document, email reply, noise — plus a reporting flag for status updates on live matters. The label and a one-sentence reason sit on the thread, so every call can be checked at a glance.
4. Draft what can be drafted Enquiry replies run the same two-pass drafter as the manual button — analyse the chain, then write in house style — so an automatic draft is identical in quality to a requested one. Status updates with TMO documents attached are drafted from the documents, with [ ] standing wherever a particular is absent.
5. Review in Drafts, send yourself The reply is rendered as clean formatted HTML — real tables, headings, bold — threaded onto the conversation with the original subject, and saved to Gmail Drafts. The attorney reviews, edits and sends. A thread that fails to draft stays in Needs review; nothing is lost quietly.

The triage is deliberately biased toward drafting: when a written reply would plausibly do, Legwork writes one — a superfluous draft costs seconds, a missed one costs a client. And the drafts are not fill-in-the-blank templates. An enquiry reply is analysed before it is written — who is writing, about which marks, against what deadline. A reporting email is built from the attached documents themselves, its deadlines worked out by Indian practice — opposition four months from journal publication; an examination response one month under Rule 33 — and each one marked calculated or stated.

On the roadmap, not in the module yet: follow-ups scheduled against deadlines and reopened when no reply arrives, and internal notifications to the docket, attorney and billing owner on the same event.

RIA for Paralegals

Prosecution runs on nudges — the letter that asks the Registrar to advertise the mark, correct the record, or issue the certificate. Legwork keeps the firm's seven Registry follow-ups as fixed templates: pick the letter, type the particulars, and the finished Word file comes back on firm letterhead, ready for a signature. No model in the loop — nothing to review but the facts you typed.

• Seven letters, one flow. Advertisement and hearing requests, re-advertisement, advertisement discrepancies, corrigenda, end-of-opposition certificate requests, certificate discrepancies and post-registration chases.
• Zero AI, on purpose. Fixed firm templates with typed particulars dropped in — deterministic to the byte. Where a template is the right tool, no model is used.
• Asks only what the letter needs. The questions adapt to the letter chosen — journal number and date for a re-advertisement, the error list for a discrepancy, the pending TM-M and TM-R requests for a post-registration chase.
• Your words, verbatim. Each error you list becomes its own paragraph in the letter, worded exactly as you typed it. Nothing is rephrased.
• Blanks stay blanks. Skip a particular and the letter keeps the template's blank underscores for hand-completion — and the covering note says precisely what was left open.
• Letterhead, dated, signature-ready. A real .docx on the firm's letterhead, dated today unless told otherwise, named by letter and number, with space left for the wet signature.

1. Pick the letter Seven Registry follow-ups covering the stretch from examination to post-registration — each with a one-line description of when it is the right one to send.
2. Type the particulars The CMP number and the application or registration number — whichever this letter runs on. The date if it shouldn't be today; day-first formats are read as typed (19/08/2026 works).
3. Answer what this letter needs Journal number and date, the errors one per line, or the requests already filed — asked only when the chosen letter uses them, and skippable. A skipped slot stays as the template's underscores, for the hand that signs to complete.
4. Take the letter The .docx assembles on firm letterhead — your particulars dropped into the fixed text, everything else untouched — and the covering note flags anything left blank. From the chat or from the tool, the output is byte-for-byte the same letter.

A follow-up letter is the cheapest way to move a file and the easiest thing to postpone. Each one is fifteen minutes of someone's attention — finding the template, the reference, the journal number, the letterhead — multiplied by every matter waiting on the Registry. Legwork makes it a minute. And this tab states a design principle the rest of the product lives by: where the right tool is a template, Legwork uses a template. The model is saved for work that needs judgment; the Registrar's nudge needs accuracy and letterhead.

On the roadmap, not in the letters tab yet: matter-grounded answers with the file already in context, cited research pulls, document intelligence on incoming Registry mail, and background docketing off the letters that go out.

Trademark Audit Reports

Every mark has a story scattered across the register — filings, objections, hearings, assignments, half-recorded changes. The Audit Report reads it in one pass: give Legwork the mark's Online Status and whatever else the file holds, and it returns the firm's audit format — particulars verified, history reconstructed, issues numbered, actions proposed with the forms to file. Facts come from the documents alone; what isn't there stays blank, and inference is marked as inference.

• The full particulars, verified. Mark, application number, class, applicant as reflecting on the online records, use claim, the specification verbatim, present status and renewal position — one table at the top.
• A documents inventory. Every paper in the file on its own numbered, dated line — “Copy of the examination report dated 04 July 2024” — and “undated” when the paper is.
• History as a narrative. The prosecution story in formal legal prose — examination grounds by section number, responses, hearings, journal advertisements, oppositions, assignments — with inferences italicised as inferences.
• Issues met by actions. Numbered problems — address mismatches, running deadlines, incomplete assignment chains, conditions restricting enforceability — each answered in Proposed Action with the form to file: TM-M, TM-P, TM-16, TM-44, and Option 1 / Option 2 where paths diverge.
• The client's file, woven in. The mail chain supplies the client's own reference, their comments and instructions — and every question they asked is answered in the report. Client papers inform; they never override the register.
• One click from the register. Finish a Talk to TMO fetch and Create Audit Report drafts the audit from the fetched documents in the background — filed to My Work before you've read the status.

1. Start from the status The mark's Online Status printout from the TMO website is the one document the audit cannot be built without. Everything else is optional depth.
2. Add what the file holds TMO records — the application as filed, examination reports, journal extracts, certificates, POAs, assignment deeds. The client mail chain. Prior-agent papers. Or skip the uploads entirely: a finished Talk to TMO fetch feeds the audit with one click.
3. The register is read Particulars are extracted only from the documents — the applicant as the online record shows it, the goods and services verbatim, every paper inventoried with its date. A fact the documents don't state stays blank.
4. The analysis is applied Status-specific playbooks: a registered mark gets its renewal position checked; an objected one its response deadline and strategy; an opposed one its counter-statement date; an abandoned one its revival prospects; a Madrid designation its provisional-refusal check. Where absolute grounds appear, a Section 9 scoring framework is applied against the actual goods and services.
5. The report lands in the firm's format Particulars table, History, Issues, Proposed Action — and a Note only when genuinely warranted. The client's instructions are summarised and answered. Out comes an editable Word file, filed to the matter.

An audit is how a firm takes custody of a mark — the new client, the prior-agent handover, the portfolio acquired mid-prosecution, the pre-enforcement check. The question is always the same: what is this file, really? The Audit Report answers it in the format the firm already signs, from the documents the register already holds — and with Talk to TMO on the same platform, the whole loop closes: fetched from the Registry in the morning, audited before lunch.

One report, one mark. An audit is built for a single mark and taken to the bottom of its file. A portfolio is audited mark by mark — there is no one-report-per-client shortcut, and the depth is the reason.

Patent Prosecution and Drafting

Patent work carries a different burden of precision. Legwork supports specification drafting and prosecution with the claim set, the prior art and the examiner's own language held together, so an amendment in one place does not quietly break something in another.

• Patent drafting assistance. Specification, claims and abstract drafted from disclosure material and inventor input.
• Prosecution workflows. First and subsequent examination reports tracked to their statutory dates.
• Office action analysis. Each objection separated, characterised and matched to a response strategy.
• Claim review. Dependency, antecedent basis and support checked across every amendment.
• Prior-art context. Cited art read against the claims, with the distinguishing features stated.
• Structured documentation. Forms, working statements and filing sets assembled from the matter record.

Patent Personas

Different patent tasks want different instincts. A drafter reaches for breadth; a searcher reaches for the closest art that defeats it. Legwork runs specialised personas tuned to each part of the lifecycle, so the assistant answering you is the one suited to the question.

• Drafting Patent drafting persona Works from disclosure to specification, holding claim scope and support in view as the language settles.
• Prosecution Patent prosecution persona Reads the examination report on its own terms and builds the response around the statutory grounds.
• Prior art Prior-art research persona Searches to defeat the claim, not to confirm it, and reports what it found against each feature.
• Claims Claim analysis persona Tests dependency, antecedent basis, breadth and infringement reach across a claim set.
• Portfolio Patent portfolio persona Reads a family across jurisdictions: coverage, renewals, gaps and where the next filing belongs.

Litigation Drafting

IP litigation drafting, held to the record. Legwork produces pleadings and applications grounded in the matter's own facts and documents, and keeps every issue consistent from the plaint to the last rejoinder.

• Pleadings. Plaints and civil suit drafting built from the matter file and the documents on record.
• Written statements. Paragraph-by-paragraph responses, with nothing left unanswered by accident.
• Replies and rejoinders. Replications drafted against what was actually pleaded, not against the last draft.
• Applications. Interlocutory applications and supporting affidavits assembled as a set.
• Legal research support. Authorities pulled and applied to the issue they were cited for.
• Consistency and issue tracking. Every issue tracked across the file, so positions do not drift between drafts.