Claude for Word is ridiculously useful

Claude decides to conform the 14th amendment to its current meaning

I just started using Claude for Word this week. It is spectacular. The idea is to work inside of Microsoft Word but to have the Claude Add-in work side by side as your drafting partner. It can provide comments about your entire draft: "tell me how I might shorten the second section of the article" or it can respond specifically to passages that you have selected in your document with the mouse: "This passage would benefit from an example or two; can you provide a couple?" Claude then comes up with responses in its side bar that are either just things you can read or actual suggestions on how to edit the document that you can choose to apply. The tool also reads comment threads and can respond to them, which supports iterative review workflows. That means Claude functions like another reviewer on the document rather than an outside box you shuttle text into and out of.

Claude isn't limited to the Word document before it in developing its responses. Just as in regular Claude, you can enrich its context by giving it access to particular documents you upload or to web search. And because Claude for Word appears to be able to access all the skills and connectors regular Claude can access, this means you can not only request conventional edits, but you can also subject your drafts to amazing new Claude for Legal features such as "Opposing Counsel Review" or "Persuasive Legal Writing" or "Red Team Verifier" or my very own "Belcher-proof" all without leaving the comforting Word environment. And when "Track Changes" is on in Word, Claude for Word creates a "proposal card" that you can choose to apply or not. The result is typical redlined text except that this time it is an AI doing the redlining.

I think the best way to prove my point is with a demonstration. I'm going to take a motion for summary judgment posted on the Internet and see how we can work with it in Claude for Word. After converting the PDF to Word, I load up the .docx file and open the Claude Add-In by pressing the Claude icon in the top ribbon. (To install Claude for Word, go to this website. You may also have to reboot your machine for reasons that are not clear to me. Also, you need a paid Claude account; it doesn't work with the free version.) If all goes well, your ribbon should look something like this, with the Claude icon on the far right and a sidebar on the right. (You won't see the red arrow or "The sidebar area." I just added those to the screen capture for what I hope is additional clarity.)

I then just want to get a feel for the document. In the Claude sidebar, I write "Summarize the argument made in this document in 150 words." Of course, I could have done the same thing inside of Claude itself by just uploading the document either in the web version or the Claude desktop app, but the advantages of making Word the primary app and Claude the helper should become apparent.

Here's what Claude tells us. (By the way, I asked Claude on a second pass to convert its answer to an HTML div so that it would appear more clearly here. I'm going to do similar things on some future prompts, but I won't bother telling you each time.)

Summary of Plaintiffs' Argument

Plaintiffs Williamsburg Climbing Gym and Fifth Concerto (guarantor) move for summary judgment discharging their obligations under a 2018 lease for a Brooklyn premises they intended to operate as a "BKBX" facility — a studio-based model built around instructor-led group fitness classes of 12–24 people. They argue two independent doctrines excuse performance.

Frustration of purpose (Restatement § 265; Krell, Benderson, Jack Kelly, LIALS, Arons): the landlord knew and approved the BKBX use, and Governor Cuomo's Executive Orders plus continuing NYC bans on indoor group fitness classes have destroyed the only purpose for which the premises were leased.

Impossibility of performance (Two Catherine, Kolodin, Leisure Time, Moyer): the pandemic and resulting governmental orders — unforeseeable in 2018 and unallocated by the lease — make operating the BKBX concept objectively impossible.

Plaintiffs lawfully terminated on May 1, 2020; Fifth Concerto's guaranty fell with the lease.

Fair enough. But now let's use some Claude superpowers. Before filing this draft, I might want to simulate a judge's first impression. So I can use the Judicial First Impression skill available from lawve (formerly lawvable), which can now be accessed using Claude and Claude for Word.

Use the lawve Judicial First Impression skill to tell me how a judge will understand this filing.

Judicial First Impression — Williamsburg Climbing Gym Co. v. Ronit Realty LLC

A reading of the plaintiffs' summary judgment papers only, before the defendant has been heard.

1. What I think this case is about

A commercial tenant on the Williamsburg waterfront, together with its corporate guarantor, took a ten-year lease in November 2018 to operate what they describe as a "BKBX" facility — a studio-based, instructor-led group fitness concept distinct from the traditional rock-climbing gyms operated by the parent group. The COVID-19 pandemic and the Governor's Executive Orders intervened during construction. The tenant purported to terminate the lease on 1 May 2020 and now seeks summary judgment that its obligations, and the guarantor's, are discharged on the alternative grounds of frustration of purpose and impossibility of performance. Roughly $1.25m in rent has been paid and some $2.26m expended on construction; the premises never opened for a single day.

2. Immediate points of confusion

The lease clause set out in the brief, Section 62(b), defines the permitted use as "primarily as an indoor climbing facility," with fitness studios listed as one of several incidental uses. The plaintiffs say their purpose was studio-based group fitness. I cannot, on this reading, reconcile the asserted purpose with the words on the page of the lease the parties signed.
The brief leans heavily on pre-execution emails and post-execution design correspondence to establish that the landlord "knew and approved" of the BKBX model. It is not yet clear to me whether the plaintiffs say the lease was modified, whether they invoke parol evidence, or whether they accept the written use clause and contend that the parties' subjective purpose nonetheless controls.
Gyms in New York City were permitted to reopen, subject to restrictions, on 2 September 2020. The plaintiffs purported to terminate on 1 May 2020. I should like to understand on what footing termination was justified four months before the regulatory picture had clarified.
The relief sought is framed as "rescission" in places and "discharge" in others. These are not the same remedy and have different consequences for the $1.25m already paid and the $2.26m spent on the build-out.

3. What feels strong

The factual narrative of the pandemic and the cited Executive Orders — particularly No. 202.3 closing gyms and No. 202.13 halting non-essential construction — is uncontroversial and plainly supplies an external, supervening event.
The continuing prohibition on indoor group fitness classes, as distinct from the partial reopening of conventional gyms, is a serviceable line on which to rest a "this tenant in particular" argument, if the underlying premise about the lease's purpose can be sustained.
The frustration cases cited — Krell, Jack Kelly Partners, LIALS, Benderson — are the orthodox authorities a court would expect to see, and they are accurately summarised so far as I can tell from the brief itself.
The unforeseeability point is, in the abstract, well taken: a 2018 commercial lease cannot fairly be read to have allocated the risk of a global pandemic.

4. What feels weak or unconvincing

The gap between the written use clause ("primarily as an indoor climbing facility") and the asserted "principal purpose" (studio-based group fitness) is the central difficulty of the motion, and the brief does not, to my eye, confront it squarely. Frustration of purpose under the Restatement requires the frustrated purpose to have been a basic assumption of both parties. The plaintiffs' own contractual language points the other way.
The reliance on a course of emails and design approvals to recast the purpose of a fully integrated commercial lease is asserted rather than argued. No discussion of the parol evidence rule, of any integration clause, or of why the written instrument should yield to the correspondence appears in what I have read.
The cited authorities are largely tenant-friendly cases in which the written purpose matched the frustrated use — restaurant leases for restaurants, office leases for offices. None of them, as summarised, involves a tenant whose claimed purpose differed from the use clause. That distinction will not escape the defendant.
The impossibility cases — Two Catherine, Kolodin, Leisure Time Travel, Moyer — are each tied to their own peculiar facts (building department inaction, a no-contact stipulation, a hotel fire, a 666% landfill price increase). The brief gathers them without explaining why the doctrine, as a New York court would apply it to a ten-year commercial lease, fits the present facts. New York law on impossibility in commercial leases tends to be considerably stricter than the brief's tone suggests.
The May 2020 termination is asserted to have been "lawful" but the brief does not appear to grapple with whether a temporary closure — however severe — suffices, or whether the doctrine requires permanence or near-permanence of the frustrating condition.

5. What I suspect (but cannot yet see proved)

I suspect the lease contains an integration clause, a force majeure clause, and a rent abatement or casualty provision, none of which is discussed in the brief. If so, the omission is itself telling.
I suspect the defendant will say that the landlord's "approvals" of design drawings were approvals of alterations under §43.05(b), not amendments to the use clause, and that the distinction matters.
I suspect there is a body of New York commercial-lease COVID jurisprudence by now that bears directly on these doctrines. The brief cites no such authority. Either the plaintiffs were unable to find favourable cases, or they have chosen not to engage with adverse ones. Neither inference flatters the motion.
I suspect the guarantor's position depends entirely on the tenant's; nothing in the brief suggests an independent ground for discharging the guaranty, and a well-drafted guaranty often survives the tenant's defences. The brief treats this as automatic, which I should not.

6. My provisional level of confidence

Low. The brief is competently written and the cited cases are real, but the motion's central premise — that the principal purpose of the lease was a use the lease itself classifies as incidental — is asserted rather than established. Until I see the lease in full and hear the defendant on the use clause, the integration question, and the force majeure provisions, I am not in a position to grant summary judgment to a tenant that walked away from a ten-year obligation four months into a public-health emergency that the parties' own contract may have addressed.

7. What I would expect to see next

The full text of the lease, particularly any force majeure, casualty, rent abatement, integration, and "no oral modification" clauses.
The defendant's opposition, which I expect will press the §62(b) use clause hard.
Authority — from either side — on how New York courts have treated frustration and impossibility claims by commercial tenants arising from the COVID-19 closures. There must be such authority by now.
A clearer account from the plaintiffs of the remedy actually sought: discharge of future obligations only, or rescission with restitution of the $1.25m in rent and $2.26m in construction expenditure.
Evidence, if it exists, that the BKBX use was contemplated by both parties as the lease's principal purpose at the moment of execution, and not merely tolerated as an incidental one.

If I were a practicing lawyer, I would find this feedback extremely useful — even where some first impressions aren't quite right. Here's the follow up I asked Claude inside Word.

Some of this stuff we can not fix due to the fact that this is an academic teaching exercise and not real litigation. For example I don’t have a copy of the actual lease. But what textual changes can you recommend that might improve our first impression on the judge.

Claude came up with nine suggestions, which I summarize here.

Open with the strongest fact: indoor group fitness classes remained prohibited in New York City when the memorandum was filed. Make that the first sentence and frame it as the practical holding sought.
Replace overstatements. Drop “once in a century” and “undisputed foreseeability.” Say instead that the parties did not, and could not reasonably have been expected to, allocate the risk of a state-ordered ban on the leased premises’ specific intended use.
Do not present Krell, Benderson, Jack Kelly, and Long Island Airports Limousine as an undifferentiated citation block. Concede that some involved risks existing at execution, then argue that their principle applies even more strongly when the loss of lawful use results from a later sovereign act.
Address the adverse COVID-frustration cases directly, including BKNY1 and the Gap/Victoria’s Secret line. Distinguish them on the ground that this case involves an ongoing prohibition of the premises’ specific use — boutique group fitness — not a temporary interruption of general retail, restaurant, or gym operations.
Compress the BKBX-specificity point into one forceful paragraph: this was not simply a “gym” lease, but a lease for a boutique group-fitness studio whose model depended on 12-to-24-person indoor classes, the precise activity New York prohibited for over a year.
Remove “as a matter of law” from the Preliminary Statement. The summary-judgment posture already conveys that.
Quote the lease’s use clause early in the facts. The § 265 argument depends on showing the lease’s principal purpose.
Treat the May 1, 2020 termination as requiring doctrinal support. Add a citation or sentence explaining why frustration permits discharge and surrender, rather than merely excusing future performance.
Replace heated pandemic rhetoric such as “raging on” with neutral formulations like “remain in effect.”

Claude then helpfully wrote: "If you want, I can stage these as tracked-change proposals against specific paragraphs — say the word and I'll pull paragraph indices and propose diffs. Given Track Changes is on and these are substantive rewrites, they'll flow through review cards rather than straight edits." I told it to do so but to use yet more superpowers such as a connector to CourtListener to make sure that the revisions were grounded in reality. I also forbade it from using cases after April 16, 2021, because that is when the real brief was filed.

Here is what I would like you to do. Make the edits you originally suggested but before the detailed discussion of case law, do legal research, use your connector to CourtListener to do the legal research that will prevent our analysis from having mistakes. Do not use cases after 4/16/2021. Do not use midpage as I want to show what can be done without a paid midpage account.

Claude asks about the form of its response. Should it make the changes directly or make proposals in "response cards." I choose the latter. Here is an example of what it proposes.

After review, I ask Claude to apply them all. Because I turned "Track Changes" on in Word, we can see the actual redlining.

At this point, I am concerned that the original document and the edits may contain grammar errors. So I can use a built-in skill to clean up the document, all within Word. It's a simple slash command called /copy-edit. (There's also its cousin, /check-doc, which does a structural integrity check, looking for things like misnumbered sections, cross-references, open comment blocks – the kind of stuff you would want to get rid of before filing.)

The /copy-edit slash command finds a number of issues and makes the corrections automatically. So long, Grammarly?

So far we've been using skills and algorithms to edit. But we can also have Claude collaborate in a more immediate way with the author. I give Claude this prompt:

Here’s the current opening of the brief. %%% [[[Opening paragraph inserted here]]] %%% It’s technically fine but it’s dull. Remember what New York City has just been through at the start of the COVID pandemic. It was calamitous. try and capture the idea that this was not ordinary minor frustration. It was a cataclysm that forced the government into unprecedented action that no one could anticipate let alone parties to a commercial lease. Try and capture that spirit right in the opening before going into the more technical parts.

Claude for Word gives me two choices both of which are far better than the dull original. Claude reasons: "The current opening reads as bureaucratic motion language. Recasting it leads with the visceral fact of what March 2020 was in New York — empty streets, refrigerated morgue trucks, a Governor closing every gym in the state — and frames the doctrines of frustration and impossibility as the law's response to a true cataclysm, not a routine business disappointment. The technical Rule 56 recitation is preserved verbatim in a second paragraph so nothing is lost." Here's the first "City stopped" option:

And here's the "Foundation destroyed" option.

I prefer the Foundation-destroyed option as it reinforces the idea of frustration of purpose. I apply it and the draft changes. There's no copy and paste between apps with the attendant formatting issues. Everything happens inside good old Microsoft Word.

We could certainly continue with human-Claude line edits and in real life I probably would. But before I went too much farther I would like AI's help in identifying the most serious flaws in the argumentation. Some of those flaws may be incurable – it may be our motion is not particularly strong – but it also might be that we can anticipate certain lines of attack and shore up our position. To do this, I deploy the Opposing Counsel Review skill through the already-beloved lawve connector. After some bizarre complaining from Claude that it can't find the skill, and some not-so-gentle coaxing on my part, Claude produces a devastating report. I show the beginning of the opposing counsel review and an additional excerpt. This skill is great at demolishing any inflated sense of legal competency.

Some of these complaints are the consequence of my not having the file underlying this brief. But perhaps we can enlist Claude to help us be more persuasive on some of the issues. Again, I enlist a skill made available through the lawve connector.

OK. Claude, you got me. unfortunately, because this is somewhat of a pedagogic exercise in how to use Claude for Word, i don;’t actually have the underlying lease. And I don’t want you to fake up a lease, although that might be an option. Instead, could you use your lawve (or lawvable) persuasive writing skill to figure out some passages we can develop to help weaken some of the most devastating attacks developed by opposing counsel. Why don’t you come up with three edits of the brief for me to review.

Here's an example of the kind of edit that Claude recommends. It is, in my opinion, pretty darned good.

I thus press "Apply All" on all three changes Claude recommends. Again, I could do a lot more editing but I also want to make sure that my brief doesn't have any "hallucinations." I use CourtListener not because it is better than midpage ai, another service for checking against hallucinations, but because I want to show how even people who don't have the money for a paid Midpage account can still at least begin to check citation accuracy – and, again, without leaving Word.

Now use the CourtListener connector to review every citation in the brief and make sure that (1) the case actually exists and is cited correctly and (2) all quotations from cases are accurate; (3) the case actually stands for the proposition we cite it for; and (4) the case is still good law on the point cited.

It's a good thing we checked! In the course of improving the writing in response to my earlier efforts, Claude went above and beyond the call of duty. It fabricated. Claude for Word is no better than regular Claude at avoiding hallucinations. Here's what CourtListener tells us.

The one urgent problem — staged for your review

407 East 61st Garage
— fabricated quote. I authored this citation last turn. The case exists at 23 N.Y.2d 275, but the quoted language "the principal object the parties had in view, not on the particular words by which they described it" is not in the opinion at page 281 or anywhere else. Worse, the opinion's actual frustration discussion is adverse — it holds frustration is "inapposite" where purpose was frustrated by a party's own business decision and rejects unilateral abrogation based on financial disadvantage. Both of those propositions sit close to Landlord's reply theory.

I've staged cite-407e61st-fix above. It removes the bad quote and the 407 East 61st Garage citation, and substitutes a verified quote from Jack Kelly Partners — which the brief already relies on and which articulates the same "so completely the basis of the contract" standard. The structural move (anchoring shared assumption over the lease's outer permissions) is preserved.

Smaller flags worth knowing about, not urgent

LIALS (96 A.D.2d 998 / 62 N.Y.2d 846) — the brief frames the 96 A.D.2d opinion as a frustration-of-purpose holding. The Third Department's actual rationale is failure of consideration (the opinion cites Restatement § 265 in passing but doesn't apply it as the holding), and the Court of Appeals affirmed on contract interpretation grounds without reaching frustration. Quote text checks out. Not a misquotation, but the doctrinal label is loose. Easy to clean up if you want — change "frustrated" to "frustrated/failed of consideration" or note the affirmance grounds.
Jack Kelly Partners carries a "Caution" treatment flag in the citator (6 citing opinions). Worth a closer look before relying on it as load-bearing authority post-fix; I can run the citing-cases pull if you want.
Two Catherine St. quote is verbatim, but the statement comes from a motion to vacate a default judgment, where the court held only that impossibility was a "meritorious defense" — not that it was established. The brief reads as if the court ruled on the merits. Probably fine to leave (every casebook cites it the same way), but worth knowing if Landlord pushes back.
Moyer — minor: brief says "future performance by plaintiff [was] excused"; opinion reads "future performance by plaintiff must be excused." Bracketed substitution is defensible Bluebook practice, but a vigilant opponent will note it.
Arons v. Charpentier reporter cite is shared. 36 A.D.3d 636 also reports Almonte v. Tobias. The brief's case name disambiguates correctly — not a defect, but parallel-cite suggestion: add 36 A.D.3d at 637 on first cite to lock it.

I then tell Claude to implement the needed changes. Here's a sample of the redlining that results.

There is a ton more you can do from within Claude for Word. But this post is intended to be more inspiration than encyclopedic. Here's my final example. I couldn't figure out how to end this blog post. I wrote something cruddy. Then I stuck it in Word and asked Claude to edit it. So here's how the blog post ends.

Much better!