Considering EvenUp? Five alternatives, and the part of the product that is people.

EvenUp Inc. sells drafting for the pre-litigation and early litigation stages of a personal injury matter. The demands page describes two tiers. Express Demands are drafted by the model in minutes and handed to your team to review, edit and finalize. Expert-Reviewed Demands are described as drafted by AI plus EvenUp's team of more than 150 legal professionals, and delivered in a finalized state in one to five days. The same page describes the review work as catching missing documents, gaps in treatment and other inconsistencies, with notes from a team of former paralegals, lawyers and adjusters.

Alongside demands the company sells MedChrons, which are medical chronologies built from the records in the file, and a product line called AI Drafts covering interrogatories, requests for admission and requests for production. There is a Microsoft Word add-in so the drafting happens inside the document your associates already work in rather than in a separate application they have to remember to open. That last detail is worth more than it sounds. Adoption failures in this category are usually failures of habit rather than failures of output quality, and a tool that lives where the work already happens survives contact with a busy practice better than one that requires a context switch.

The company's own website terms describe the offering, in a sentence written by its lawyers rather than its marketers, as an administrative demand drafting service. That phrasing is more accurate than the category name most people use for it, and it is the single most useful line on the entire site for a buyer trying to work out what they are purchasing.

Scale claims are published on the homepage and are dated only by the fact that we read them on August 12, 2026. The tiles read, verbatim, ten billion dollars or more in damages claimed and ten thousand cases processed weekly, alongside a plus sixty nine percent likelihood of hitting policy limit, review fifteen times more cases in the same amount of time, and cut nine or more hours of manual work per case. None of those figures carries a footnote, a sample definition, a baseline or a time period on the page that displays them. A customer testimonial on the same page states that settlement outcomes improved by three hundred percent. These are vendor claims and a single customer's self report respectively, and we repeat them here labelled rather than as findings, because a buyer will see them and should know what they are and are not.

The company is well funded and that is a real consideration rather than trivia. A press release dated October 7, 2025 on its own blog announces a one hundred and fifty million dollar Series E led by Bessemer Venture Partners, bringing total capital raised to three hundred and eighty five million dollars at a valuation above two billion. The same release states the company was founded in 2019 and that more than two thousand firms use the platform, including twenty percent of the top one hundred United States personal injury firms. For a firm choosing a vendor to hold years of medical records, the survival odds of the counterparty belong in the evaluation, and on that dimension EvenUp scores better than almost anything else in this series.

Read the three leading products in this category side by side and the same structure appears in all of them, stated openly, and largely ignored in the way firms talk about the purchase.

EvenUp's Expert-Reviewed tier is explicitly a hybrid. The model drafts, a team of more than one hundred and fifty legal professionals reviews, and the output arrives finished in one to five days. Supio's medical chronologies page says that unlike pure AI systems, human subject matter experts verify every chronology for accuracy. On its demand letter page Supio says the AI powered demand is then reviewed by human subject matter experts for an additional layer of quality assurance, and separately that for high value cases human experts verify medical bill summations and damage calculations. Note that the chronology promise is universal and the bill and damages verification is scoped to high value cases. Those are two different commitments and the difference will matter on a file that is not high value but is not simple either.

Eve is the contrast. Its demand letters page says its legal AI drafts using your style, formatting and case facts, and that drafts are generated instantly. We looked for a statement that a human reviews the output and did not find one on the demand letters page or the homepage. Absence of a claim is not proof of absence of review, and it should not be read that way. It does mean that if human verification is what you are buying, one of these three vendors does not appear to be selling it to you in writing.

Why this reframes the purchase. Software capacity is elastic and staffed capacity is not. If part of your turnaround depends on a review queue staffed by people, then your worst week is their worst week, and the one to five day window on a finished demand is a service level rather than a compute time. Volume spikes, holiday coverage and the vendor's own hiring pipeline become variables in your case cycle. None of that is a defect. Human review is very likely the reason the output is usable on a real file rather than a plausible looking draft that a paralegal has to rebuild. It simply means the vendor should be evaluated the way you would evaluate an outsourced medical record review vendor, with questions about turnaround under load, escalation, who the reviewers are and what happens when the queue is full, and not only the way you would evaluate a piece of software.

It also explains the pricing shape. EvenUp describes its model as all in one, case based pricing. Per case pricing is the natural unit for a service where the marginal cost is somebody's hours. Per seat pricing is the natural unit for software. When a vendor in this category prices per case, that is a signal about the cost structure underneath, and it is the reason a firm cannot make the economics work simply by adding users the way it might with a research tool. Every case you send costs again.

The practical test for your own firm is a volume question, not a quality question. Take your last ninety days of pre-litigation files and count how many demands actually went out. Multiply by whatever the vendor quotes you per case, because they will only tell you in a call. Then compare it against what the same volume costs you today in paralegal and associate hours, and be honest that the internal number is a real cost that is currently invisible because it is buried in salary. Firms that run this arithmetic usually find the answer is not close in either direction, which is the useful kind of answer.

We checked five vendors on their own sites on August 12, 2026, looking for any dollar figure anywhere.

EvenUp has no pricing page at all. The URL where one would sit returns a 404. The demands page names the model, all in one case based pricing, and the call to action is Schedule a Call. Supio does have a pricing page, which is more than most, and it names two commercial shapes without pricing either of them. One is called Case Subscription and is described as a lower per case cost. The other is called Unlimited Firm Access and is described as the lowest per case cost. Both buttons lead to a sales conversation. Eve has no pricing page and the URL returns a 404, with no pricing link anywhere in the main navigation. Filevine's pricing page states that all packages are custom built for your team's needs and carries a Book a Demo call to action, with no figure on the page.

Five vendors, zero published prices. That is worth stating plainly rather than treating as normal, because it has a specific consequence for a mid sized firm. You cannot build a business case before you take the sales call, which means the vendor controls the sequence in which you learn things. You will see the demo before you see the number, and the demo is designed to make the number feel smaller. The counter is to fix your own walk away figure before the first call, write it down, and give it to whoever attends so it is not renegotiated in the room.

Third party sites do publish estimates for several of these products. We are deliberately not citing any of them. An aggregator figure that no vendor has confirmed is worse than no figure, because it anchors a negotiation on a number that the person across the table can correctly tell you is wrong, which costs you the credibility you needed for the rest of the conversation. If you want a comparison anchor, use the two facts that are verifiable: the pricing unit each vendor names, and whether the model is per case or per firm. Those tell you how your cost curve behaves as volume grows, which is the thing that actually decides whether a tool survives its second year.

One nuance on Supio's Unlimited Firm Access. An unlimited shape is the only structure in this comparison whose cost does not rise with case volume, which changes the calculus for a firm whose pre-litigation intake is growing. Whether it is priced within reach of a mid sized practice is exactly the thing that is not published, so treat it as a question to ask rather than an advantage to assume.

This is the section where the vendors differ most, and where the differences are documented rather than inferred.

Start with EvenUp's terms and conditions, last revised June 8, 2023 according to the page itself. It states that your use of the demand drafting service and products shall not be governed by that agreement, but rather by your company's agreement with the company covering those products. Read that again, because it is unusually candid and it is the most important sentence a buyer will find on the public site. The terms you can read do not apply to the thing you are buying. The terms that do apply are in a private contract you will receive after you have committed time to an evaluation. Nothing about that is improper and it is common in enterprise software. It does mean your data rights, your termination rights and your work product ownership are all unknowable from the outside, and it means the contract review has to happen before the firm builds a workflow around the product rather than after.

Supio's terms of service carry the same structure, stating that the agreement applies only to the site, and that a firm purchasing the products enters into another agreement covering their use. Neither company publishes a data processing agreement.

Eve is the exception and it deserves credit for it. Eve publishes a Master Service Agreement, last updated August 3, 2026, and section 5.1 states that as between Eve and the customer, the customer owns Customer Data, and that Eve acquires no right, title or interest in it. Critically, the agreement's own definition of Customer Data includes Output, which means the AI drafted work product is assigned to the firm in a public document you can read before you talk to anyone. That is a real differentiator. Of the vendors in this comparison, Eve is the only one where a firm can answer the question who owns the demand without signing anything first.

Now the training question, where the ranking reverses. EvenUp's privacy policy, last updated July 9, 2026, says that as an AI business it may use personal information to develop, improve, train and provide its proprietary AI technologies for all of its customers, while not permitting third party AI providers to use that information to train their own models for other customers. Its trust center is more specific and more reassuring on the mechanics: it says the models learn statistical patterns and are designed not to store or reproduce your documents, that third party AI providers including OpenAI and Anthropic are not permitted to train on case files, that retention is minimized with zero retention settings used where available, and that your case files stay with your matters and do not appear in another customer's work product. On deletion it states that requests are honored in accordance with applicable law and the customer agreement, and that backup copies may be retained for up to one year before deletion. That one year backup figure is the only specific retention period any vendor in this comparison publishes.

Supio makes a stronger sounding claim in a weaker place. Its marketing pages state that client data is never used to train outside models, that information stays private to the firm, and that the models powering Supio are trained separately and isolated from client files. Its actual privacy policy, effective March 28, 2025, contains no AI training clause at all. A promise on a marketing page is a representation and a promise in a privacy policy or a contract is an obligation, and the gap between those two is precisely what a diligence process exists to catch.

Eve and Filevine publish nothing on the subject. The word train appears zero times in Eve's public Master Service Agreement and zero times in its privacy policy, which carries an effective date of January 1, 2021 under the entity name Butler Labs. It appears zero times in Filevine's privacy policy, last updated July 22, 2026. Undocumented is not the same as unsafe, and it is not the same as safe either. It means the answer will be whatever the private contract says, so ask for it in writing.

Filevine is the only vendor here that links a Data Protection Agreement and a subprocessor list as named public documents from its own footer, alongside a subscription agreement and end user terms. For a firm whose diligence checklist includes subprocessors, that is a meaningful head start, and it is the kind of thing we look for when we run the security questions that belong in front of any AI build.

Every vendor in this category handles protected health information, because a demand package is built out of medical records. Every one of them says something about security. They do not say the same thing, and the wording differences are not decoration.

EvenUp's homepage says the company is SOC2-audited and HIPAA-attested. Its trust center expands that to independently assessed through a SOC 2 Type 2 examination and a HIPAA compliance assessment, and adds penetration testing at least annually. Supio's security page says Supio is SOC 2 Type II certified, and that it has implemented and maintained the security controls and practices required for HIPAA, PHIPA and GDPR, while elsewhere on the site the wording becomes simply HIPAA and GDPR compliant. Eve's security page says its information security program follows the criteria set forth by the SOC 2 Framework, and mentions independent third party assessments and annual penetration testing. We read that page in full and it contains no mention of HIPAA, PHIPA, HITECH or GDPR anywhere, on a product built for plaintiff personal injury work. Filevine's security page says its audits include HIPAA, CJIS and ISO 27001 security controls in addition to NIST 800-53 controls, that the company is pursuing ISO 27701, ISO 27017 and ISO 27018 certifications, and that it endeavors to comply with the HIPAA Security Rule and the HITECH Act. We searched that page for SOC and found only references to a security operations center, not to a SOC 2 report.

Ranked by how much they commit to, those four sentences run from strongest to softest roughly as follows: certified, audited and attested, includes those controls in its audits, and follows the criteria set forth by. Only the first two describe a completed third party examination in language a security reviewer would accept without a follow up question. Follows the criteria set forth by describes an internal program modelled on a framework, which may be excellent and may be nothing, and the way to find out is to ask for the report.

One technical note that cuts against the strongest sounding claim. SOC 2 is an attestation engagement resulting in a report, not a certification, so a vendor describing itself as SOC 2 certified is using the term loosely even when the underlying report is genuine. We are not treating that as evidence of anything, because the phrase is used loosely across the whole software industry. We flag it because the correct response to any of these four sentences is identical: ask for the report under an NDA, check the date, check the scope, and check whether the product you are buying is inside that scope. A current Type 2 report covering the right systems settles the question in a way no marketing sentence can.

The four vendors, on the four questions this page has been working through, each answered in their own published words and read on August 12, 2026.

Parrot is worth a line here, because a firm shopping this category will run into it. Its own site describes it as a platform for remote depositions, with products covering court reporting, deposition summaries, medical record summaries and transcripts. It does not sell demand drafting. Its banner announces that Parrot is now part of Filevine, and its footer legal entity is Filevine, Inc. So it is neither an independent company nor a demand vendor, and it belongs in a different evaluation than this one.

This series exists to be useful rather than to be negative, and on several dimensions EvenUp is ahead of the field.

It runs a trust center as a separate published surface, and the answers there are specific in a way marketing copy usually is not. Naming OpenAI and Anthropic explicitly as providers who are not permitted to train on case files is more concrete than any competitor's language on the same question. Committing to a backup retention window of up to one year gives a diligence reviewer an actual number to accept or reject rather than a phrase like as long as necessary, which is what the other privacy policies in this comparison offer.

Its integrations page names systems rather than showing logos. The page states that case files and data flow in from integrations including Litify, SmartAdvocate and CASEpeer, and describes automatic syncing from SharePoint, Google Drive and Dropbox. Supio's site, by contrast, says it connects with the case management systems your firm runs on and offers an API, without naming a single system on the pages we read. Naming three specific platforms is a checkable claim and a vague one is not, and the difference matters most to the firm whose stack happens not to be on the list.

It is also honest about the hybrid model rather than hiding it. Plenty of vendors would describe a staffed review team as proprietary quality assurance and leave the buyer to discover the headcount later. Putting a team of more than one hundred and fifty legal professionals on the product page is the opposite of that. Our disagreement is not with the model, which is probably why the output is usable. It is that the buying process treats a hybrid service like a software subscription, and the two behave differently under load.

Last, the funding position is real. A firm about to route three years of medical records through a vendor should care whether that vendor exists in three years, and a dated press release announcing a Series E at a two billion dollar valuation is a stronger answer to that question than most vendors in this series can give.

How to tell which one you are.

The choice is decided by three numbers that live inside your own firm, none of which a vendor can tell you.

The first is demands per month that actually go out. Below roughly the volume where a single paralegal is fully occupied assembling them, a per case fee is almost certainly cheaper than any build and quite possibly cheaper than the hire, and the honest recommendation is to buy. As volume rises the per case fee compounds while an owned system does not, and somewhere above that the arithmetic inverts. We are deliberately not naming the crossover number, because it depends on your quoted per case price, which nobody publishes, and on your own loaded cost per paralegal hour, which we do not know. Compute both and the crossover falls out of the two.

The second is where the hours actually go. Firms tend to assume the drafting is the expensive part because it is the visible part. Time it. If the majority of the elapsed time is records collection, chasing providers and reading treatment history, then a demand generator solves a small part of the problem and a chronology product solves a larger one. If the drafting genuinely is the constraint, the demand tier is the right target.

The third is how different your demands are from everyone else's. A firm with a genuinely distinctive demand format, its own damages theory and a house style that its senior attorneys defend is a poor fit for a service whose value proposition is a finished document in one to five days, because every returned draft gets rebuilt anyway. That firm is either an Express tier customer or a commission customer. A firm whose demands are competent and conventional gets more from the finished tier than it would from any build.

Two situations argue against doing anything at all this year. If your case management data is inconsistent enough that a human has to check every field before it can be trusted, then every option on this page inherits that problem and the first project is data readiness rather than drafting. And if nobody in the firm owns the outcome, meaning there is no named person whose job it is to decide in ninety days whether this stays or goes, then the purchase will drift into a subscription nobody can defend or cancel. The maturity assessment is a faster way to find out which of those applies than a quarter of trying.

What we would build, and what we have not built.

ColabContent commissions custom AI systems for mid market operators at a fixed fee, with the code owned by the client at handoff, and we say no when an off the shelf product is the better instrument. For a plaintiff firm the pattern we would build is an assembly layer rather than a writer: it reads the records already in the case management system, extracts the treatment timeline and the billing, reconciles what is present against what the file should contain, flags the gaps a reviewer would otherwise catch on a second pass, and drafts into your firm's own format for an attorney to approve. The differences from a subscription are that it runs on your own matter data, it does not charge again per case, and it belongs to the firm afterward.

We have not shipped a commission for a plaintiff personal injury firm, and we would rather say that here than let a comparison page imply otherwise. Our delivered work in law has been concentrated in intake and communications rather than demand assembly. That is a genuine limit on what we can claim, and it is the kind of thing worth asking any consultant to state plainly before you engage them. What we can offer on this specific decision is the evaluation itself, including the case where the answer is that you should buy the product and we should not be involved.

If you want the cost side before a conversation, what a law firm commission actually costs lays out the bands, and the generic version of the buy against build question covers the five tests we run on every diagnosis call.

What is EvenUp, and what does it actually produce?

EvenUp Inc. is a drafting product for plaintiff personal injury firms. Its main outputs are demand packages, medical chronologies which the company calls MedChrons, and drafts of discovery responses including interrogatories, requests for admission and requests for production. It also ships a Microsoft Word add-in so drafting happens inside Word rather than in a separate application. Demands come in two tiers on its own demands page. Express Demands are drafted by AI in minutes and finalized by your own team. Expert-Reviewed Demands are described as drafted by AI plus EvenUp's team of more than 150 legal professionals and delivered in a finalized state in one to five days. The company's own website terms describe the offering as an administrative demand drafting service, which is the most precise description of it published anywhere on the site.

How much does EvenUp cost?

EvenUp publishes no price. There is no pricing page on the site and the URL where one would sit returns a 404. Its demands page names the commercial model as all in one, case based pricing without attaching a figure to it, and the call to action is to schedule a call. Third party sites publish estimates, and we are not citing them, because a number no vendor has confirmed will be corrected by the salesperson in the room and will cost you the credibility you needed for the rest of the negotiation. The useful preparation is to decide your own walk away figure before the first call, based on what the same volume of demands costs your firm today in paralegal and associate hours, and to give that figure to whoever attends the call.

Does a human review the demand, or is it purely AI?

It depends on the tier and the vendor, and this is the most consequential difference in the category. EvenUp's Expert-Reviewed tier is explicitly hybrid, described on its demands page as drafted by AI plus a team of more than 150 legal professionals, with review notes from former paralegals, lawyers and adjusters. Its Express tier is AI drafted and left for your own staff to finalize. Supio states on its medical chronologies page that human subject matter experts verify every chronology, and on its demand letter page that the AI powered demand is then reviewed by human subject matter experts, with bill and damages verification specifically scoped to high value cases. On Eve's demand letters page and homepage we found no statement that a human reviews the output, which is not proof that none happens, only that it is not a published commitment.

Why does it matter whether part of the product is staffed by people?

Because software capacity is elastic and staffed capacity is not. If the finished demand depends on a human review queue, then the one to five day turnaround is a service level rather than a compute time, and your worst week coincides with the vendor's worst week. Volume spikes, holiday coverage and the vendor's own hiring become variables in your case cycle. It also explains why pricing in this category is per case rather than per seat, since the marginal cost of another case is somebody's hours. None of this is a criticism of the model, which is very likely why the output is usable on a real file. It means the vendor should be evaluated with the questions you would put to an outsourced record review provider, covering turnaround under load, escalation paths and what happens when the queue is full, and not only the questions you would put to a software vendor.

Who owns the demand that the AI drafts?

For most of these vendors you cannot tell from the public site, and one of them tells you clearly. Eve publishes a Master Service Agreement last updated August 3, 2026 whose section 5.1 states that as between Eve and the customer, the customer owns Customer Data and Eve acquires no right, title or interest in it, and the agreement's own definitions include Output within Customer Data, which covers AI drafted work product. EvenUp's public terms and conditions state explicitly that they do not govern the products and that a separate private agreement does, so ownership is not determinable from outside. Supio's terms of service carry the same structure, applying only to the website. Filevine links a Data Protection Agreement and a subprocessor list publicly from its footer, which is more than the others publish, though the subscription agreement is what would settle ownership.

Will my clients' medical records train these vendors' AI models?

Only two of the four vendors address it, and they address it in different places. EvenUp's privacy policy, last updated July 9, 2026, states that as an AI business it may use personal information to develop, improve, train and provide its own proprietary AI technologies for all of its customers, while not permitting third party AI providers to train their models on it. Its trust center adds that the models learn statistical patterns and are designed not to store or reproduce your documents, names OpenAI and Anthropic as providers not permitted to train on case files, and states that zero retention settings are used where available. Supio states on marketing pages that client data is never used to train outside models and that its own models are trained separately and isolated from client files, but its privacy policy effective March 28, 2025 contains no AI training clause at all. The word train appears zero times in Eve's public Master Service Agreement and in its privacy policy, and zero times in Filevine's privacy policy last updated July 22, 2026. Undocumented means the answer sits in a private contract, so ask for it in writing.

How long is my data kept if we cancel?

EvenUp is the only vendor in this comparison that publishes a specific number. Its trust center states that deletion requests are honored in accordance with applicable law and the customer agreement, and that backup copies may be retained for up to one year before deletion. Supio's privacy policy uses generic language about retaining information as long as necessary to fulfil a legitimate business need, with no period stated. Filevine's privacy policy states that personal data is retained while services are provided and that even after the account is closed it keeps personal data in order to comply with legal and reporting obligations, again with no period. We could not locate a specific retention or deletion on cancellation clause in Eve's public documents. For a firm holding protected health information on behalf of clients, a stated window is materially better than an undefined one, and it belongs in the contract rather than only on a trust page.

Are these vendors SOC 2 and HIPAA compliant?

They each say something different, and the wording is the finding. EvenUp's homepage says SOC2-audited and HIPAA-attested, and its trust center says independently assessed through a SOC 2 Type 2 examination and a HIPAA compliance assessment. Supio's security page says it is SOC 2 Type II certified and has implemented the security controls and practices required for HIPAA, PHIPA and GDPR, while elsewhere its own site simplifies that to HIPAA and GDPR compliant. Eve's security page says its information security program follows the criteria set forth by the SOC 2 Framework, and that page contains no mention of HIPAA, PHIPA, HITECH or GDPR anywhere. Filevine's security page says its audits include HIPAA, CJIS and ISO 27001 controls in addition to NIST 800-53, that further ISO certifications are being pursued, and that it endeavors to comply with the HIPAA Security Rule, and we found no SOC 2 claim on it. Treat all four sentences the same way: request the report under NDA and check its date, its scope, and whether the product you are buying sits inside that scope.

Which case management systems do they integrate with?

EvenUp names systems and the others largely do not. Its integrations page states that case files and data flow in from integrations including Litify, SmartAdvocate and CASEpeer, and describes automatic syncing from SharePoint, Google Drive and Dropbox. Supio's site states that it connects with the case management systems your firm runs on and offers an API for custom workflows, without naming a specific system on the pages we read. We did not find a published integration list for Eve. A named platform is a checkable claim and a general statement is not, so if your firm runs something outside the named set, make the integration a written condition of the order rather than a topic in the demo. The same discipline applies whichever product you choose, and it is the single most common reason these purchases underdeliver in year one.

What happened to Filevine's DemandsAI, and is Parrot a competitor?

Both have moved, and a firm working from an older comparison article will get this wrong. The Filevine URL that served a demand specific product now resolves to a page titled for LOIS for Word, Filevine's general AI drafting and contract redlining product, and the product path under its products directory returns a 404. Filevine's current AI is presented under the LOIS umbrella covering in-Word drafting, question answering over case data, and an AI drafting module whose published feature list includes human verified output. Parrot describes itself on its own site as a platform for remote depositions, with products covering court reporting, deposition summaries, medical record summaries and transcripts, and its site banner announces that Parrot is now part of Filevine, with Filevine, Inc. as the footer legal entity. So Parrot is neither independent nor a demand drafting product, and it belongs in a different evaluation.

Should a firm doing forty demands a month buy this or build something?

Buy first, and let the number decide the second year. At that volume a per case fee is almost always cheaper than a build and usually cheaper than the additional hire, and you learn what good output looks like on your own files without committing capital. The thing to protect is your ability to leave. Get the per case price in writing, get the ownership and deletion terms in the contract rather than off a trust page, and set a named person and a ninety day date for the decision to renew or stop. If volume grows and the per case line becomes one of the larger numbers on the firm's software spend, that is the moment to price an owned assembly layer against it, because a build's cost stops rising as case count rises and a per case fee does not.

How should we run the evaluation so it produces a decision?

Measure your own baseline before a demo is booked. Pull the last ninety days of pre-litigation files and record three numbers: elapsed days from records complete to demand sent, demands sent per paralegal per month, and how many drafts came back needing substantive rewriting rather than editing. Write down what each would have to reach for the purchase to be worth renewing, and write it before anyone sees a demo. Then run one vendor on a real block of live files rather than a curated sample, with the same attorney reviewing throughout so the quality judgment is consistent. Name the person who decides and the date they decide. A pilot designed this way ends in a decision somebody signs, which is the only output that makes the exercise worth running.

Related reading.

More on AI in plaintiff and mid market legal practice: what we would build for a law firm, the twelve tools mid market firms actually shortlist, and AI for mid market law firms beyond the hype. The neighbouring comparisons cover different layers of the same stack: Filevine and case management AI, Harvey against a custom build, the Harvey alternatives field, and Clio's AI against a commission. The integration path for Salesforce firms is documented in the Litify playbook. To size your own numbers before a call, the law firm diagnostic takes a few minutes. If the vocabulary is new, start at what custom AI means and what AI commissioning means, then what a mid market engagement costs and how we work. The rest of this series sits on the comparisons hub.