Dispatches
FIG. 19 AI enterprise search → FIG. 18 Athena Python SDK → FIG. 17 Athena Skills → FIG. 16 Web Monitor → FIG. 15 Athena for Operations → FIG. 14 Athena for Analytics → FIG. 13 AI search for your DMS → FIG. 12 Multiplayer AI agents → FIG. 11 Athena Anywhere → FIG. 10 Hey Athena → FIG. 09 Athena for Finance → FIG. 08 What code review becomes → FIG. 07 GTM without a CRM → FIG. 06 The company brain → FIG. 05 Two identical stacks → FIG. 04 Finance builds its own APIs → FIG. 03 Model routing in production → FIG. 02 Knowledge workers don't have GitHub → FIG. 01 Pricing, read from transcripts →
Dispatch 2026-09-30 Product
№13

AI Search for Your DMS: Find the Precedent You Know Exists

Make your legal team agent native with Athena. Ask for a precedent in plain language, and Athena finds it in your document management system, iManage, SharePoint and the rest, with a link to the page it came from.

Filed by Taylor Johnson — 13 min
Sepia engraved plate: a stone archive wall of arched niches filled with rolled papyrus scrolls, one scroll drawn out and unrolled across a stone table beside an olive branch, a bronze stylus and a lit clay oil lamp, with an open arch onto a hillside temple and a calm sea with one sail

Athena Intelligence is an agent platform that searches your document management system from a plain-language question. Ask for a precedent and Athena looks across iManage, SharePoint, OneDrive and Google Drive with your own access, then answers with a link to the file and page, so a lawyer can read the source before using it.

Most of what a law firm knows is already written down, in executed agreements, filed briefs and memos. It sits in a document management system (DMS) and in the SharePoint and drive folders around it, filed under a client and a matter number. The hard part is finding the one that answers the question in front of you, and that is the job AI search for your DMS takes on.

This dispatch covers what that search looks like when Athena does it: one precedent request followed from a partner's message to a cited answer, the wider set of searches legal teams run, and the controls that keep every result inside the access each person already has. Figure 1 shows the whole idea.

  • Athena searches iManage, SharePoint, OneDrive and Google Drive from a plain-language question and answers with a link to the file and page.
  • Results follow each person's access in the source system, so documents someone cannot open there do not appear for that person.
  • The lawyer reads the source and decides what applies, and nothing is sent or filed until a person approves it.
  • The same search can compare matches, start a draft from the precedent or be saved as an agent operating procedure (AOP) for the next request.
Contents
  1. 01The problem
  2. 02What Athena can find
  3. 03One request, start to finish
  4. 04How this differs from other tools
  5. 05Trusting the answer
  6. 06Security and deployment
  7. 07Getting started
  8. 08Questions teams ask
FIG. 01 / ASK ONCE, SEARCH EVERYWHEREA lawyer asksIn plain language, from chat, email,a text or a commentYour document systemsSearched with the asker's own accessiManageSharePointOneDriveGoogle DriveAthena01Search every driveWith the asker's own sign-in02Ask when unsureA question when two deals fit03Cite the sourceLinks the file and the pageA cited answerA short answer with a linkto the file and the pageThe source filesOpened in the threadto read in contextA filed memoDrafted from the signed copy,uploaded once approvedACitedEvery answer links to its sourceBYour permissionsAccess follows each personCReversibleEvery change can be undoneDHuman approvalNothing goes out unapproved FIG. 01 / ASK ONCE, SEARCH EVERYWHEREA lawyer asksIn plain language, from chat, email,a text or a commentYour document systemsSearched with the asker's own accessiManageSharePointOneDriveGoogle DriveAthena01Search every driveWith the asker's own sign-in02Ask when unsureA question when two deals fit03Cite the sourceLinks the file and the pageA cited answerA short answer with a link to the file and pageThe source filesOpened in the thread to read in contextA filed memoDrafted from the signed copy, uploaded once approvedACitedEvery answer links to its sourceBYour permissionsAccess follows each personCReversibleEvery change can be undoneDHuman approvalNothing goes out unapproved
FIG. 01Lawyers ask in plain language, and Athena searches the connected document systems, asks when unsure and returns a cited answer, the source files and a memo filed once a person approves. The four protections (A to D) apply to every search.

01What problem does AI search for a DMS solve?

It closes the gap between knowing the work exists and being able to find it. A partner writes that the client wants a capped indemnity with a fraud carve-out and asks for the firm's precedent. The associate remembers that the firm did one last year but not which deal, so a DMS search for the obvious keywords returns nothing. The people who would know are out, so the associate opens file after file of drafts and clean copies, pastes a cap into a memo from the wrong version, and the partner asks which version was actually signed. Figure 2 sets that next to the same request handled by Athena.

FIG. 02 / THE SAME REQUEST, TWO WAYSTHE MANUAL WAYGuess keywordsZero resultsAsk aroundThe expert is outOpen filesDrafts and clean copiesPaste a memoCap from a draftWhich version?Did we sign v7?Time passesWITH ATHENAAskSigned version, plain wordsSearchEvery connected driveCheckAthena asks which dealFileA person approvesTime passesBars show relative effort on one scale for both lanes. The scenario comes from the launch video and is demo data. FIG. 02 / THE SAME REQUEST, TWO WAYSTHE MANUAL WAYGuess keywordsZero resultsAsk aroundThe expert is outOpen filesDrafts and clean copiesPaste a memoCap from a draftWhich version?Did we sign v7?Time passesWITH ATHENAAskSigned version, plain wordsSearchEvery connected driveCheckAthena asks which dealFileA person approvesTime passesBars show relative effort. Demo scenario.
FIG. 02The same request handled two ways. On the manual lane, the associate guesses keywords and gets zero results, asks around, opens file after file and pastes a cap from a draft. With Athena, the signed version is found, Athena asks which deal to use, and a person approves the filing. Demo scenario.

The DMS is where this work already lives. iManage reports that 79% of the Am Law 100 rely on its platform, so a precedent search has to work on the systems firms already run. Time lost to search was measured long before AI. In IDC's 2012 Information Worker Survey of 840 people, workers spent 2.3 hours a week searching for online documents they never found, and another 2.0 hours recreating documents because the right version could not be located. The survey is old and predates AI search, which makes it a baseline for what keyword search and folder browsing cost.

DMS vendors have started adding natural-language search of their own. iManage added cited answers across its platform to Ask iManage in January 2026, which helps a great deal inside one system. Legal work usually spans several: the executed copy in the DMS, the term sheet in SharePoint, the email that explains why a point was conceded. Athena searches across them and then works with the results.

02What can Athena find in a document management system?

Athena finds the documents, clauses and answers a team has already produced, in iManage, SharePoint, OneDrive and Google Drive, and works with the results in the same thread. Figure 3 groups twelve common searches by the team that runs them.

FIG. 03 / WHAT TEAMS SEARCH FORDEALSFind the clauseIndemnity caps, carve-outs,anti-sandbaggingCompare across dealsHow one clause wasnegotiatedStart from precedentPull the signed form into adraftLITIGATIONFind prior briefsMotions and declarations ona topicGet up to speedKey documents on a matteryou joinFind the expertDeclarations by author ortopicPRACTICE SUPPORTLocate formsThe firm's preferredlanguageCheck past workWhether a topic was coveredPull credentialsPrior pitches for anindustryIN-HOUSE AND RISKFind the signed copyExecuted agreements bycounterpartyCompare vendor termsLiability across vendorsAnswer an audit requestPolicies and approvals,citedEach one saves as an agent operating procedure, so the next request runs the same way. FIG. 03 / WHAT TEAMS SEARCH FORDEALSFind the clauseIndemnity caps, carve-outs, anti-sandbaggingCompare across dealsHow one clause was negotiatedStart from precedentPull the signed form into a draftLITIGATIONFind prior briefsMotions and declarations on a topicGet up to speedKey documents on a matter you joinFind the expertDeclarations by author or topicPRACTICE SUPPORTLocate formsThe firm's preferred languageCheck past workWhether a topic was coveredPull credentialsPrior pitches for an industryIN-HOUSE AND RISKFind the signed copyExecuted agreements by counterpartyCompare vendor termsLiability across vendorsAnswer an audit requestPolicies and approvals, citedEach one saves as an agent operating procedure.
FIG. 03Twelve common searches, grouped by the team that runs them. Each one can be saved as an agent operating procedure so the next request runs the same way.

Deals and transactional teams

Deal teams start from precedent on almost every agreement. Athena can find a clause across prior purchase agreements, such as an indemnity cap with a fraud carve-out or an anti-sandbagging provision, compare how the same clause was negotiated across deals, and start a draft from the signed form.

Litigation teams

Litigators reuse arguments. Athena can find prior briefs and expert declarations on a topic, pull the key documents and correspondence for a lawyer who joins a matter in progress, and find declarations by author or subject.

Practice support and knowledge teams

Knowledge lawyers field the same questions again and again. Athena can locate the firm's preferred language and forms, check whether a topic was covered in earlier work, and pull prior pitch materials and credentials for an industry.

In-house legal and risk

In-house teams hold years of signed agreements. Athena can find the executed copy by counterparty, compare liability and indemnity terms across vendors, and gather policies and approvals, each with a citation, when an audit request arrives.

These searches sit on the same platform as everything else Athena does, so a result can feed a table, a memo or a deck. The legal solutions page and the enterprise knowledge search use case cover more, and the connectors page lists the systems Athena reaches.

03How does it work? One precedent request, start to finish

A lawyer asks in plain language, Athena searches the connected systems with that lawyer's access, and the answer comes back with a link to the file and the page. The launch video below follows one request from the first question to an approved filing, and Figure 4 breaks it into four steps.

VIDEOThe launch video (43 seconds, vertical): the current way of finding a precedent, then the same request run by Athena from the first question to an approved filing. The video has music and on-screen text and no narration. People, deals and figures in the video are demo data.

In the demo, Maya Chen, a partner, messages at 4:12 PM that the client wants a capped indemnity with a fraud carve-out and asks for the firm's precedent. Leo Park, an associate, answers eight minutes later that they did one last year and can't remember which deal. The first half of the video shows the current way: a DMS search returns zero results, nobody on the deal chat remembers, five files get opened one by one, and a memo gets pasted from version seven, which turns out not to be the version that was signed. The second half runs the same request through Athena. Names, deals and figures are demo data.

FIG. 04 / ONE REQUEST, STEP BY STEP01AskIn plain language,asking for the signedversion, not the drafts02SearchEvery connected drive,searched with theasker's own sign-in03CheckIf two deals fit,Athena asks which oneto use as precedent04FileA memo is drafted fromthe signed copy, and aperson approves the uploadPEOPLE DECIDE HEREAthena searches and reads in the first two steps. A person decides in the last two.Nothing is filed until a person approves it. FIG. 04 / ONE REQUEST, STEP BY STEP01AskIn plain language,asking for the signedversion, not the drafts02SearchEvery connected drive,searched with theasker's own sign-in03CheckIf two deals fit,Athena asks which oneto use as precedentPEOPLE DECIDE HERE04FileA memo is drafted fromthe signed copy, and aperson approves the uploadPEOPLE DECIDE HERENothing is filed until a person approves it.
FIG. 04One request in four steps, matching the chapters in the video. Check and File are where people decide: Athena asks which deal to use as precedent, and nothing is filed until a person approves.
  1. A lawyer asks in plain language and says to use the signed version, not the drafts, with no filename, folder or keyword to guess.
  2. Athena searches every connected drive with the asker's own sign-in. In the demo it finds SharePoint files for "indemnity" and "carve-out" and Google Drive files for "SPA", including the draft, the executed agreement and a second signed deal.
  3. Two signed deals have a fraud carve-out, so Athena asks which one to use as precedent and waits. The lawyer picks the Halvorsen SPA, and Athena opens the executed agreement to read it.
  4. Athena drafts a memo from the signed copy, and a person approves before it is uploaded to the deal folder in SharePoint. Nothing is filed until that approval.

The answer carries a citation to page 14 of the executed agreement. The video closes on four promises: every answer cited, every permission kept, every upload approved and every version recoverable.

04How is this different from other AI search tools for legal?

Athena is one platform that searches, reads and then works with the results, on the systems and permissions a firm already has, with no documents to move and no new search box to learn. Table 1 compares the manual way with Athena, task by task.

TaskThe manual wayWith Athena
Start the searchGuess a filename or a search termAsk the question in plain language
Find the right dealOpen folders organized by client and matterSearch iManage, SharePoint and more at once
Check the answerOpen each file and scan for the clauseOpen the cited page and read it in context
Compare across mattersCopy language into a document by handAsk for the matches side by side in a sheet
Reuse the precedentSave a copy and edit from memory of the dealStart a draft from the signed form
Repeat next monthAsk around againSave the search as an AOP
TABLE 1The same six tasks done by hand and with Athena. The right-hand column describes what a lawyer asks Athena to do in each case.

A dedicated search product finds documents and stops there. In Athena the same thread can compare the results in a sheet, draft from the precedent in a document or summarize them in a deck, and the same platform runs the other work a legal team hands to agents.

Athena connects to iManage, SharePoint, OneDrive and Google Drive and uses each person's access in them, so there is no second copy of the library and no second set of permissions to maintain. Every answer links to its source, every change can be undone, and nothing goes out until a person on the team approves it.

Athena deploys in managed cloud, your VPC, on-premises or air-gapped. A search that works can be saved as an agent operating procedure, and the people who know the work can write the next one themselves. Read more about agent operating procedures.

05Can lawyers trust the answers?

Lawyers can check every answer, because each one links to the file and page it came from and nothing is used until a person has read the source. Figure 5 shows the citation and the access rules side by side.

FIG. 05 / CITED, AND INSIDE EACH PERSON'S ACCESSIndemnity precedentFound SharePoint files for “indemnity”The signed SPA caps indemnity at 10% of the purchaseprice. Fraud is carved out.SPA_Halvorsen_EXECUTED.pdfp.14SPA_HALVORSEN_EXECUTED.PDF / PAGE 149.4 Limitations. The aggregate liability of Seller shall notexceed ten percent (10%) of the Purchase Price; provided,that the limitation shall not apply to any claim arising from Fraud.SAME QUESTION, DIFFERENT ACCESSA partner with matter accessSees SPA_Halvorsen_EXECUTED.pdfand the page it came fromA teammate without accessThe file is not shown. The source systemsays no, so Athena does tooAccess follows the source system. Revoke it in iManage or SharePoint and it takes effect in Athena right away. FIG. 05 / CITED, AND INSIDE ACCESSIndemnity precedentFound SharePoint files for “indemnity”The signed SPA caps indemnity at 10%of the purchase price. Fraud is carved out.SPA_Halvorsen_EXECUTED.pdfp.14PAGE 14 OF THE SIGNED SPA9.4 Limitations. The aggregate liability ofSeller shall not exceed ten percent (10%) ofthe Purchase Price; provided, that the limitshall not apply to a claim arising from Fraud.SAME QUESTION, DIFFERENT ACCESSA partner with matter accessSees SPA_Halvorsen_EXECUTED.pdfand the page it came fromA teammate without accessThe file is not shown. The sourcesystem says no, so Athena does tooAccess follows the source system.Revoke it there and it takes effect right away.
FIG. 05An answer with its citation, and the same question asked by two people with different access. Results follow the access each person has in the source system. Names, files and figures are demo data.

In the demo, the answer carries a chip for SPA_Halvorsen_EXECUTED.pdf, page 14. Clicking it opens an excerpt of the page with the cap and the fraud exception highlighted, so the lawyer reads the clause in the agreement itself.

Access follows the source system. Documents a person cannot open in iManage or SharePoint do not appear in that person's results, and when access is revoked in the source system the change takes effect in Athena right away. Athena does not store data from connected sources. Human approval applies to any tool, so a filed document or an email waits for a person to approve it, as the memo upload does in the demo.

06Is AI search for a DMS secure enough for a law firm?

Athena is built for regulated work: SOC 2 Type II and HIPAA, zero data retention agreements with model providers, and deployment inside a firm's own environment. Figure 6 shows where each part runs.

FIG. 06 / WHERE IT RUNSINSIDE YOUR ENVIRONMENTManaged cloud, your VPC, on-premises or air-gappedDocument systemsiManage, SharePoint,OneDrive, Google DriveAthenaSearch, read and citeunder your permissionsModelsFrom every frontierprovider, your choicereadreasonAthena does not store data from connected sources.SOC 2 Type IIAudited controlsHIPAAFor regulated dataZero retentionWith model providersNo trainingOn customer data FIG. 06 / WHERE IT RUNSINSIDE YOUR ENVIRONMENTManaged cloud, your VPC, on-premisesor air-gappedDocument systemsiManage, SharePoint,OneDrive, Google DriveAthenaSearch, read and citeunder your permissionsModelsFrom every frontierprovider, your choiceAthena does not store data from connected sources.SOC 2 Type IIAudited controlsHIPAAFor regulated dataZero retentionWith model providersNo trainingOn customer data
FIG. 06Athena, the document systems and the models all sit inside the customer's environment, whether that is managed cloud, a VPC, on-premises or air-gapped. The band lists the compliance and data-handling commitments.

Athena runs in managed cloud, your VPC, on-premises or air-gapped, and you own the data plane. Athena has zero data retention agreements with Anthropic, OpenAI and Gemini, does not train on customer data, and works with more than 60 models from every frontier provider. A Big Four audit firm has run Athena in production for over a year, and Athena passed more than a dozen enterprise security and AI vendor reviews this year. The deployment options and permissions and identity pages go into detail.

07How do we get started?

Start with a question the team asks every month, such as a clause precedent. Connect one system, ask the question, and read the sources together. A search that returns the right agreement is a good first agent operating procedure to save.

Curious how it would work on your own documents? We're happy to show you.

08Questions teams ask

What is AI search for a document management system?

AI search for a document management system lets a person describe what they need in plain language instead of guessing a filename or keyword. The AI plans the searches, reads the matches and answers with links to the source files. In Athena, the search runs across iManage, SharePoint, OneDrive and Google Drive using the asker's own access.

Does Athena search iManage and SharePoint?

Yes. Athena connects to iManage, SharePoint, OneDrive and Google Drive, among more than 150 data connectors. A lawyer asks once, and Athena searches the connected systems together, then answers with a link to the file and page. In the demo, Athena searches SharePoint and Google Drive together and cites the signed agreement.

Will Athena show documents a person is not allowed to see?

No. Athena uses each person's own access in the source system, so documents someone cannot open in iManage or SharePoint do not appear in that person's results. If access is revoked in the source system, the change takes effect in Athena right away. Athena also does not store data from connected sources.

Does Athena copy our documents out of the DMS?

No. Athena does not store data from connected sources. Documents stay in iManage, SharePoint or wherever they live, and Athena reaches them through the connection, with the asker's access, when a search runs. Athena can run in your own cloud environment, including your VPC, on-premises or air-gapped, so the data plane stays yours.

How do I know an answer is right?

Every answer links to the file and page it came from, so the lawyer reads the language in the source before relying on it. When two documents fit, Athena asks which one to use. A person decides what applies to the matter, and anything Athena would send or file waits for a person to approve it.

Can Athena do more than find documents?

Yes. Athena can compare the matches side by side in a sheet, draft a memo from the signed copy and, once a person approves, upload it to the deal folder in SharePoint. A search that works can be saved as an agent operating procedure (AOP), so the next request, such as a clause precedent for a new deal, runs the same way.

Where does my data go, and can it run in our VPC?

Athena runs in your own cloud environment: managed cloud, your VPC, on-premises or air-gapped. Athena has zero data retention agreements with Anthropic, OpenAI and Gemini, does not train on customer data and does not store data from connected sources. SOC 2 Type II and HIPAA apply, and you own the data plane.

Athena Library

Run this on a schedule

Scheduled briefs for legal and audit teams, delivered as files.

№12 Multiplayer AI Agents Teammates and agents edit and comment in the same sheet, document or deck, including the copy on a laptop. next dispatch

All dispatches → Work on this →