iManage · Antidote

Legal DMS

Matter work product, email, and document management inside the firm. The corpus is the firm's. Microsoft 365 is the bundling and AI-invasion surface.

As of 19 August 2026. This memo covers matter work-product filing and retrieval inside a law firm. No live-product test has been run.

Work product is the firm's. Microsoft is the invasion surface.

iManage is a work-product system: custody, versioning, and email filing. The documents are typically firm- or client-owned. There is no exclusive corpus. Direct AI substitution of a DMS fails because the job is not generation. The invasion path is a platform attaching retrieval and generation to the estate the firm already runs.

Matter taxonomy and Outlook filing can create switching costs. Those remain unobserved until export completeness, metadata fidelity, and migration time exist. Workflow capability and operating economics stay open without an independent product test. Generation add-ons are a different treatment.

Exposed vs scarce. Work product is the firm's. Generation add-ons are reconstructable. Custody, versioning, and Outlook filing are the product — not a scarce corpus.
Still untested. Retrieve-and-file completeness, version error, and migration fidelity versus SharePoint.
The comparison that would change the call. A declared SharePoint or generic DMS versus iManage on the same matter set and taxonomy budget.

Decision questions

On a frozen matter set, does iManage clear a prespecified retrieve-and-file gate relative to the current process?
What isolated increment comes from taxonomy, email filing, or integrations, each changed one at a time?
What is full cost per accepted retrieve, including admin FTEs, not seat list price?
Can work product leave the system without material metadata loss?

Filing errors, missed versions, broken email capture, and failed search are different classes. They stay separate. A correct document with the wrong matter is a different error than a missed document with a good query.

Published benchmarks attach to the steps that have an honest public floor. ParseBench bounds whether filed PDFs parse. CUAD and LegalBench bound add-on search and generation over work product. They do not measure retrieve-and-file, versioning, or Outlook capture. Those remain the scored job.

Workflow breakdown

Each step of the job is mapped to a published benchmark or another reputable quantitative measure only where the mapping is honest. A public figure is a floor or a labor-market context on that task type. It is never a score of this product.

Task	Published measure	Pack fit	What it does not measure
Parse filed PDFs and attachments	ParseBench	Pack fit strong	Custody still requires a faithful parse. Not miss rate or version error.
Find a clause inside a filed contract	CUAD	Public floor only	Add-on search over work product. Not email filing or matter taxonomy.
Matter retrieve-and-file, versioning, and Outlook capture	None	No honest public floor	The scored job. No public pack measures custody, version error, or email filing.

Findings at a glance

Capability, cost, and defensibility are assessed separately. They are not averaged into a score. Where evidence is missing, the finding stays open.

Status	Area	Finding
Open	Workflow capability	A DMS is custody and retrieval, not legal reasoning. No independent retrieve-and-file test has been run.
Open	Operating economics	Admin load, storage, and full cost per accepted matter retrieve are unmeasured. A seat is not that quantity.
Judgment	Defensibility	Work product is typically the firm's. Metadata CSV export is documented, but full-fidelity document, version, permission, audit-history, taxonomy, and migration behavior remains unverified. Microsoft 365 is the bundling and AI-invasion surface.
Vendor-stated	Vendor claims and open questions	Public materials describe work product, email, and document management for legal and professional services. Those are vendor declarations.

Workflow capability

Search coverage is not retrieval performance

Work product, email, and matter folders describe an input surface. They do not describe miss rate, version error, or time-to-retrieve on a frozen set. No independent iManage test has been run.

A DMS can be the system of record and still fail the cases that matter: a prior version produced as current, an email that never filed, or a search that misses a known work product. Those classes stay apart.

Published benchmarks

ParseBench bounds whether filed PDFs parse. CUAD and LegalBench bound add-on search and generation over work product. None of them measures retrieve-and-file, versioning, or Outlook capture.

ParseBench

Filed PDFs still have to parse. Tables, charts, faithfulness, formatting, and grounding stay separate. Parse quality is not retrieve-and-file.

Third-party leaderboard · not an Antidote run · not a product score

Current frontier VLMs on ParseBench

Official ParseBench leaderboard, observed 13 August 2026. Five dimensions kept separate. Gemini 3 Flash is Thinking High. GPT-5.6 Sol and Fable 5 are Reasoning None / default API settings in the source.

Dimension	Gemini 3 Flash	GPT-5.6 Sol	Claude Fable 5	Claude Opus 4.8
Tables	91.5%	89.34%	89.79%	89.65%
Charts	64.79%	59.57%	52.21%	49.75%
Faithfulness	90.87%	86.17%	90.02%	89.02%
Formatting	68.31%	50.42%	72.62%	71.38%
Grounding	59.77%	25.09%	49.24%	18.69%

Third-party leaderboard · not an Antidote run · not a product score

Specialist parsers vs a frontier VLM

Same ParseBench leaderboard. LlamaParse Agentic, Pulse Ultra 2, and Reducto Agentic against Gemini 3 Flash Thinking High. The source Overall column is omitted.

Dimension	LlamaParse Agentic	Gemini 3 Flash	Pulse Ultra 2	Reducto Agentic
Tables	90.74%	91.5%	75.45%	80.42%
Charts	78.11%	64.79%	90.82%	73.4%
Faithfulness	89.68%	90.87%	79.49%	86.37%
Formatting	85.24%	68.31%	73.05%	57.6%
Grounding	80.62%	59.77%	66.56%	67.07%

CUAD

Matter files often include contracts. CUAD bounds extractive clause location on public contracts. It does not measure custody, versioning, or email filing.

Paper table · Hendrycks et al., 2021 · not an Antidote run · not a product score

CUAD precision-recall, selected models

Hendrycks et al., 2021, Table 2. AUPR and precision at 80% recall are different estimands and are not averaged.

System	AUPR	Precision @ 80% recall	Precision @ 90% recall
DeBERTa-xlarge	47.8%	44%	17.8%
RoBERTa-large	48.2%	38.1%	0%
RoBERTa-base	42.6%	31.1%	0%
ALBERT-xxlarge	38.4%	31%	0%
BERT-base	32.4%	8.2%	0%

How to read the published benchmarks

ParseBench is the floor on filed PDFs. CUAD and LegalBench are floors on add-on search and generation. None of them is a retrieve-and-file score, and none of them is iManage. Add-on generation, if present, is a later treatment. If it cannot be turned off in a comparison, say so and leave attribution unproven.

Current systems, cost, and harness

Official list prices observed 8 August 2026 bound the model bill only. They are not a product subscription and not full cost per accepted output. ParseBench cents per page is a parser-layer cost. Full cost still has to include failed attempts and reviewer minutes.

ParseBench tables vs cents per page

ParseBench official leaderboard, observed 13 August 2026. Horizontal axis is list cents per page. Vertical axis is the Tables metric (GTRM), not an overall score.

SharePoint is the named competitor, not a straw man

The failure mode that matters for iManage is treating a CUAD or LegalBench cell as retrieve-and-file performance. Custody, versioning, and Outlook capture are the scored job.

The interesting paired tests are shared drive versus a declared generic DMS, and generic DMS versus iManage, each with the same matter set and taxonomy budget. A product that only ships the last comparison cannot claim the increment. Generation add-ons are a later treatment. If they cannot be turned off in the comparison, say so and leave attribution unproven.

Operating economics

A seat is not admin load

List seat price is a commercial term. Full cost still has to include admin FTEs, storage, professional services, and lawyer minutes spent filing and searching, with failed retrieves in the denominator.

No iManage all-attempt cost is in this report. Copilot-style search can cut find-time and still leave the DMS bill in place. Any leverage case remains modeled and assumption-bound.

Defensibility

Public facts, not measurements

Named surface and packaging, from iManage's published pages. iManage documents AI processing inside its governed repository, permission-bound Microsoft Copilot and third-party integrations, and a default that neither it nor model providers train on customer cloud data; limited-insight programs require opt-in. Implementation partners are a named part of adoption. Metadata CSV export is documented; full-fidelity portability is not. A 2023 DLA Piper CIO interview cited Microsoft alignment as a selection factor.

Competitive set, as a hypothesis: Microsoft 365 / SharePoint / Copilot, and other matter DMS products. That is not a bake-off.

Embed can bind. The corpus is not the moat. Microsoft is the exposure.

Work product is usually the firm's. iManage is a custodian. Direct AI substitution of a DMS fails because the job is custody, versioning, and email filing, not generation. The invasion path is a platform attaching retrieval and generation to the estate the firm already runs.

Matter taxonomy, Outlook filing, and practice integrations can raise switching costs. Vendor documentation describes metadata CSV export; that does not establish full-fidelity document, version, permission, audit-history, or taxonomy portability. Migration time, services hours, and support attach remain unmeasured.

Defensibility

iManage's load-bearing questions are embed, export, and Microsoft bundling. Work product is usually the firm's. That is not a data moat.

Dimension	Rating	Judgment	Missing evidence
Data rights	Exposed	Matter work product, email, and documents are typically firm- or client-owned. iManage is a custodian. Custodianship without reuse rights is not an exclusive asset.	Custody versus ownership terms and client-file reuse rules.
Exclusivity	Exposed	Document-management concepts, matter taxonomy, and Outlook filing are category features. There is no exclusive corpus. The differentiator, if any, is configured taxonomy and integrations, which are customer-specific.	What, if anything, cannot be reproduced in SharePoint or another DMS.
Rebuild difficulty	Conditional	A generic DMS can host files. Reconstructing matter taxonomy, email-filing rules, and historical versions is a migration project. Hard is not impossible. AI generation add-ons are a different treatment and are not this report.	Export completeness, metadata fidelity, and timed migration.
Freshness	Open	Version currency and email-capture completeness are controlled within each customer's repository. iManage's published default bars shared-model training on customer cloud data, so customer count is not a cross-firm content-learning loop.	Version-error rate, email-capture completeness, target opt-ins, and any permitted telemetry fields.
Switching cost	Conditional	Outlook filing, matter workspaces, and practice integrations can bind lawyers' daily path. That is the plausible switching-cost story. It stays contingent until export and migration are observed.	Switching interviews, migration time, and metadata loss on export.
System of record	Conditional	For many firms iManage is the work-product system of record. Time, finance, and DMS-of-record conflicts are unobserved. Generation add-ons must not be allowed to impersonate record authority.	Write-authority map across DMS, finance, and matter systems.
Integrations	Conditional	Email, identity, and practice tools create gravity. Microsoft 365 is both the integration surface and the bundling threat.	Integration inventory and a SharePoint paired comparison.
Network effects	Exposed	A DMS does not get better because another firm uses it. Network effects, if any, are local to a firm's own matter graph.	Any claimed cross-firm learning would need rights and a measured increment.
Trust and compliance	Conditional	Legal buyers pay for confidentiality, audit, and ethical-wall machinery. Those claims need an audit, not a brochure.	Access-control tests, ethical-wall evidence, and incident history.
Distribution	Conditional	Law-firm distribution is a specialized channel. It does not block Microsoft, NetDocuments, or a practice-management bundle from attaching DMS features to existing seats.	Channel concentration and bundling exposure, separately from retrieval quality.
Services dependence	Open	iManage maintains a broad implementation and migration partner ecosystem. Specialist services are structurally part of deployment and switching, while their customer-level intensity remains unmeasured.	Admin FTE, professional-services hours, and escalation rate.
Rate of improvement	Open	No measured improvement curve on retrieve-and-file error rates is in this report.	Dated filing and retrieval error series.

How AI can compete

Replacement, a new AI product, a platform bundle, cheaper labor, and a customer building it themselves are different risks. They are not averaged into a score.

Mechanism	Assessment
Direct replacement	SharePoint or another DMS can host files. Substitution of configured taxonomy and email filing is a migration, not a file copy.
New AI product	Copilot-on-the-corpus attacks search, not custody. The invasion path is Microsoft attaching retrieval and generation to the existing estate, not a legal-model product replacing the DMS.
Platform bundle	The material exposure. Microsoft 365 already sits under email, identity, and files. A bundled DMS-plus-generation offer can price against a standalone legal DMS.
Labor compression	AI search can cut lawyer minutes spent finding a document. That is value-creation on top of the DMS. It does not retire the need for a system of record.
Customer builds it	A firm's own IT can host the corpus on the SharePoint estate it already pays for. That is the bundling exposure executed by the customer. The binding constraint is migration: export completeness, metadata fidelity, and taxonomy rebuild.

Data flywheel

A flywheel needs every step below. On this evidence it is unproven. Having data is not the same as being allowed to reuse it, and reuse is not the same as a measured quality gain.

Link	Status	Note
Unique data	Vendor-stated	Matter exhaust is the firm's. iManage's exhaust, if any, is telemetry and configuration.
Permission to reuse	Open	iManage's published default bars customer cloud data, including AI inputs and outputs, from model training. Optional limited-insight programs require explicit opt-in; executed terms remain controlling.
Reliable feedback	Open	No audited filing-error labels.
Proven quality gain	Open	No paired DMS-only versus add-on result.
Better outcomes	Open	No frozen retrieve-and-file miss rate.
More usage or data	Open	Retention from taxonomy embed is unobserved.

What public sources resolve

Public filings and vendor documentation can settle some structural questions. They cannot replace a test on the live product.

Question	Conclusion	Still open
Is customer content used to train shared AI models?	iManage says neither it nor third-party model providers train on customer cloud data by default. Limited-insight programs require explicit opt-in; broader public improvement language makes the executed cloud agreement controlling.iManage AI; Terms of Service	Executed terms and any target opt-ins, including what limited insights contain and how long they persist.
Can work product be exported without material loss?	An absolute no-export claim is contradicted by documented metadata CSV export. Full-fidelity portability remains unproven.Exporting metadata of items in the Grid view	A migration test covering document bytes, versions, metadata, permissions, audit history, links, taxonomy, elapsed time, and external-services hours.
Is iManage services dependence low?	A specialist implementation and migration ecosystem is structurally part of adoption and switching. Public evidence does not support a low-services assumption.Implementation Partners; DLA Piper's CIO talks through the iManage Cloud selection	Target quotes, migration and training hours, support attach, partner margins, escalation rates, and five-year total cost.

Required comparisons

A decision-grade test would run these side by side. None has been run here.

Comparison	Note
Current workflow	Requires imported filing and retrieval process. Not assumed from Outlook folders.
Shared-drive floor	File shares and mailbox search without a matter DMS.
Named competitor	Declared SharePoint or generic DMS with the same taxonomy budget.
This product	Needs authorized iManage access and matter-level traces.
Isolated changes	Taxonomy, email filing, and integrations must be separable or labeled unproven.

Open claims

Each claim stands on its own. They are not averaged into a moat score. If only the product surface is observable, retrieval outcome and price can still be reported. Add-on stories stay unproven. The corpus is typically the firm's. Microsoft search is not a DMS substitute until a paired comparison exists.

Claim	Status	What is still missing
Matter retrieval is faster and more complete than a generic DMS.	Open	Frozen retrieve tasks, miss rate, and a SharePoint paired comparison.
Published documentation shows metadata CSV export; full-fidelity portability is unresolved.	Open	Full-fidelity testing across document bytes, versions, permissions, audit history, taxonomy, and migration time.
Work product is typically the firm's, not iManage's exclusive corpus.	Judgment	Custody versus ownership terms and client-file reuse rules.
Microsoft bundling risk is low.	Open	A DMS-native generation paired comparison and channel exposure.
Implementation and migration partners may make specialist services part of iManage adoption and switching.	Open	Admin FTEs, partner and professional-services hours, implementation margin, support attach, escalation rate, and source evidence.

What this report does not conclude

Replacement of the current filing process
Retrieval completeness versus a generic DMS
Full cost per accepted retrieve
A data moat; the corpus is typically the firm's
A switching-cost, Microsoft-bundling, or services-dependence figure

What would settle these questions

Authorized iManage access with exportable traces.
A sealed matter set with independently adjudicated retrieve keys.
Imported observations of the current filing process.
A declared SharePoint or generic DMS cell, same matters, same budget.
Export and migration completeness, including metadata.
A DMS-native generation paired comparison versus iManage-only retrieve.
Admin-FTE and lawyer-minute measurement, including failed retrieves.

Sources

Terms of ServiceiManage · Official contract · 16 July 2025
iManage's public terms distinguish customer content from usage information and say AI inputs and outputs are processed to provide services, while also granting a general product-improvement license.
iManage states that customer accounts are governed by a nonpublic Cloud Services Agreement rather than the website terms. The executed CSA, DPA, AI addendum, opt-ins, and provider terms are controlling.
The iManage Platform: SecurityiManage · Official policy
iManage lists independent ISO, SOC, and CSA STAR assurance, annual audit cadence, granular access controls, encryption, monitoring, and regional controls.
Certification breadth does not establish secure target configuration, report exceptions, availability, incident absence, or retrieval quality. Current reports remain gated.
iManage AIiManage · Official product documentation
iManage says AI processing remains in its governed repository, supports Microsoft Copilot and third-party tools through permission-bound integrations, and does not train models on customer cloud data by default. Optional limited-insight programs require opt-in.
Vendor description and default policy. It does not establish retrieval accuracy, implementation effort, or the executed target terms.
Exporting metadata of items in the Grid viewiManage Work Help · Official product documentation
iManage documents CSV export of configured metadata columns for selected items or a grid.
This disproves an absolute no-export proposition but does not establish complete document, version, audit-history, permission, or taxonomy fidelity in a migration.
Implementation PartnersiManage · Official product documentation
iManage maintains a broad specialist implementation ecosystem for migration, deployment, integration, training, and support. The services layer is structurally part of adoption and switching.
Partner count and capability do not reveal customer-level hours, attach rate, implementation margin, migration cost, or continuing services dependence.
DLA Piper's CIO talks through the iManage Cloud selectionPaul Gilford interview, Legal IT Insider · Named interview · 18 September 2023
DLA Piper's CIO described a long diligence process, long-standing iManage use, roadmap and Microsoft alignment, and the significant change program a different decision would have required.
One named customer selection, not a representative switching-rate estimate. The account concerns movement to iManage Cloud, not an observed exit from iManage.