The Hidden Data Problem That Stops Law Firms From Using AI
An AI audit for law firms finds that real matter data lives in Outlook, not Clio. Here is what the audit uncovers and what it actually takes to fix it.
I had a managing partner tell me that Clio was “basically our brain.” He meant it as a reassurance - proof that the firm had invested in modern infrastructure and was ready for AI. So we pulled up Clio together.
Time entries: populated. Invoice history: there. Contact records: present. Now I asked him to show me the last three substantive communications on the firm’s biggest open matter. He navigated to the matter in Clio. There were two time entries from the past month. That was it. The client emails, the opposing counsel correspondence, the memo his associate wrote summarizing the deposition - all of it was in Outlook. Some of it was attached to a SharePoint folder. One important document was in a PDF someone had emailed to him directly that he had never forwarded anywhere.
This is the hidden data problem that stops law firms from using AI. Clio is not their brain. Clio is their accounting software. The actual matter work - the substance of legal practice - lives somewhere else entirely.
Why “We Use Clio” Is Not an Answer
Clio has one of the better APIs in professional services software. Full REST access with webhooks. You can read and write matters, contacts, time entries, tasks, calendar events, and documents. From a Systems perspective, Clio is genuinely connectable - if the data is in Clio, you can build on it.
The problem is what firms actually use Clio for. In the audits I run on small and mid-size law firms, the pattern is consistent: Clio (or MyCase, or whatever practice management system the firm uses) is adopted primarily for time tracking and billing. Those features have hard business consequences - if you do not track time, you do not get paid - so they get used religiously.
Everything else is optional in practice, even if the firm has a policy otherwise. Client communications happen in Outlook because lawyers have used Outlook their whole careers and it is already where their email lives. Research memos are saved to SharePoint or a network drive. Draft documents live in Word. Deadline reminders go into Outlook calendar. Strategy notes might be in a physical notepad, a Teams channel, or a partner’s personal notes app.
The result: any given active matter has its real substance spread across email, SharePoint, and individual attorney devices - and a billing record in Clio that tells you how much time was spent but almost nothing about what was actually done.
The Legal Tech Landscape: What Is Wired and What Is Not
When I map the Systems layer for a law firm, the picture looks roughly like this.
Clio is the best-case scenario for practice management connectivity. The API is mature, documented, and reliable. If your firm is on Clio and has disciplined data entry habits, significant AI connectivity is possible - matter summarization, deadline tracking, conflict checking, document assembly triggered by matter status.
MyCase is more limited. The API has improved in recent versions but does not match Clio’s depth, particularly for programmatic document handling and webhook reliability.
Filevine occupies a different position - it was built with workflow and case management more centrally than billing, and it has API access. Mid-size litigation firms on Filevine often have better matter data inside the platform than Clio users, simply because Filevine’s workflow features make it more useful for substantive case tracking.
Document management systems - iManage and NetDocuments primarily - are where documents actually live at larger firms. Both have APIs. Both require authentication configuration and security policy alignment. But at least the documents are in one place, searchable, and programmatically accessible. Firms running iManage or NetDocuments are in a materially better position for document-layer AI than firms relying on SharePoint or network drives.
Outlook and Exchange are the real challenge. This is where half the matter data lives and where it is hardest to do something structured with it. Microsoft 365 has a Graph API that can read email and calendar data, but accessing attorney emails at scale raises conflict-of-interest, confidentiality, and client privilege questions that need legal review before you start connecting them to anything. The data is technically accessible via API. The question of whether you should access it programmatically - and how - requires careful scoping.
The Matter Intelligence Gap
Here is the specific problem this creates for AI. The two AI capabilities law firms most want right now are matter summarization (give me a quick brief on where this case stands) and status updates (draft a client email with current status). Both are genuinely useful. Both fail completely if the data for the matter is not in the system the AI reads.
If a client calls about their employment matter and the attorney uses an AI assistant to pull a current summary - and all that exists in Clio is four time entries and a billing invoice - the summary will be four billing line items. That is not useful. It is possibly harmful, because it gives the appearance of completeness when the substance is missing.
The audit finds this gap and documents it with specifics: on average, what percentage of matter activity is actually recorded in the practice management system versus existing only in email? In the firms we have assessed, the answer is typically 20-40% for active litigation matters and somewhat higher (60-70%) for transactional work where document production is the primary deliverable and documents do end up in a DMS.
Translating that gap into cost: if associates spend 45-60 minutes per week manually reconstructing matter status before client calls - pulling threads from Outlook, checking SharePoint for the latest draft, calling the partner who handled the last conference - and you have 12 fee earners doing this across 40+ matters each, the wasted hours add up quickly. At $150-250/hour associate billing rates, you are looking at $50K-$100K in write-offs or non-billable time annually that better data discipline would recover.
The Process Map: What It Reveals in Practice
The AI audit’s process map for a law firm traces the matter lifecycle from intake through close. For litigation practices, this typically surfaces the following disconnects.
Intake. New client intake forms exist - paper or digital. The information gets entered into Clio (contact, matter, basic details). But the intake conversation itself - what the client said, what the attorney’s initial assessment was, what documents the client brought - often goes into a handwritten note or a Word doc that never makes it into the matter record.
Active matter work. Communications are in Outlook. Research is in a folder. Drafts are in Word or Google Docs. Strategy is in people’s heads. The Clio matter record shows time entries with descriptions like “client conference” or “research and draft motion” - accurate but useless for any AI system trying to understand matter status.
Deadline tracking. Many firms use Outlook calendar, not Clio’s task system, for actual deadline management. This is a risk issue as much as a data issue - if deadline reminders exist only in one attorney’s calendar, what happens when that attorney is on vacation or leaves the firm?
Close. Matters close in Clio for billing purposes. But the close-out materials - final agreements, settlement documents, close-out letters - often go into a client physical file or a disorganized SharePoint folder, not into a structured DMS with matter-tagged metadata.
Priority Matrix for Law Firms
The priority matrix from a legal AI audit typically looks like this.
High impact, lower effort:
- Mandate that matter notes and key communications get logged in Clio as document attachments or activity notes. This is a process change, not a technology change. It costs nothing and starts building the data layer immediately.
- Set up structured document naming conventions and a designated SharePoint library per matter. Not glamorous, but it makes the documents searchable and connectable when you are ready for the Intelligence layer.
High impact, medium effort:
- If the firm does not use a DMS, evaluate iManage or NetDocuments. The connectivity they enable for document-layer AI is significantly better than SharePoint, and for a firm billing $2M+ annually, the investment is justified by what it unlocks.
Intelligence layer (requires Data and Systems to be in order first):
- Matter summarization AI that reads actual matter history from a connected DMS and practice management system
- Automated client status update drafting based on recent activity logged in the matter
- Conflict checking automation using complete contact and matter data
- Deadline extraction from matter documents with automatic Clio task creation
The sequence is critical. Firms that spend money on AI document review or matter summarization tools before their documents are in a properly structured DMS end up with AI that cannot find the documents it needs to review. The tool exists. The input does not.
What Makes Legal AI Audits Different
Legal practices have two concerns I take seriously that do not come up in most other industry audits.
Client confidentiality and privilege. Any AI system that processes matter data is handling confidential attorney-client communications. The AI blueprint specifies clearly which data the AI reads, where it is processed, what the retention policy is, and how the system is isolated from other clients’ matters. These are not afterthoughts. They go in the deliverable explicitly.
Conflict checking. Before any AI system can be trained on or given access to matter data across the firm’s history, conflict checking protocols need to be confirmed. If the AI system can see matter A and matter B simultaneously, and those matters involve parties on opposite sides of a dispute, that is a problem - regardless of whether the AI “knows” what it is seeing.
These concerns do not prevent AI implementation in legal practices. They define the scope of the implementation. The AI blueprint in a legal audit is more constrained than in other industries, and deliberately so.
For a broader view of what the AI readiness framework looks like before engaging with any audit, the AI readiness audit guide is a useful starting point. For firms that also manage real estate transactions or property-related legal work, the AI audit for real estate agencies covers how MLS data constraints and agency data habits create similar matter-tracking gaps.
Frequently Asked Questions
Our firm uses SharePoint as our document store - does that count as having a DMS?
SharePoint is connectable via the Microsoft Graph API, and it can serve as a document repository. But it lacks the matter-centric metadata structure of a purpose-built DMS like iManage or NetDocuments. Documents in SharePoint are often inconsistently named, stored in arbitrary folder hierarchies, and lack the systematic tagging that makes AI document retrieval reliable. It is better than a network drive. It is not as good as a proper DMS. The audit assesses how your SharePoint is actually organized and whether it is structured enough to support AI connectivity.
Our firm has 4 attorneys - is an AI audit relevant at this size?
Yes, and in some ways it matters more at small firms. Four-attorney firms often have zero standardization around where matter data lives, because everyone just keeps things the way they personally prefer. The gap between “what is in the practice management system” and “what is in attorney inboxes and local folders” is often proportionally larger than at larger firms with IT departments enforcing structure. The AI audit at this size is a faster, lower-cost engagement that prevents spending money on AI tools that will not work with your actual data.
We have been using Clio for 5 years - should our data be in good shape?
Five years of Clio usage is a good foundation for billing and time tracking data. It is not automatically a good foundation for matter intelligence. What matters is whether your firm has used Clio’s full feature set - matter notes, document storage, calendar integration, task tracking - or just the billing module. We see plenty of firms with 7-10 years of Clio history whose matter records contain almost nothing except invoices and time entries. The audit assesses what is actually there, not how long the platform has been in use.
What AI tools are law firms actually using successfully right now?
The most successful current deployments I see are: automated conflict-check flagging using Clio’s contact data, document assembly for high-volume transactional work (NDA, engagement letter, standard contract generation), and AI-powered intake questionnaires that pre-populate Clio fields from client responses. These all work because they operate on clean, structured data - either from Clio’s well-maintained billing records or from structured form inputs. Matter summarization and status drafting at scale remain aspirational for most firms because the underlying matter data is not in a state that supports them yet.
How do we handle privilege and confidentiality if an AI system is reading matter correspondence?
The AI blueprint specifies this explicitly. For correspondence processing, options include: using a self-hosted or private-cloud AI deployment where data does not leave your environment, limiting AI access to document metadata rather than content, or restricting AI matter access to documents that have already been designated as non-privileged. Different firms make different choices based on their risk tolerance and practice area. What matters is that the choice is made deliberately and documented before any AI is connected to client data - not after.
Ready to Get Started?
Tell us what you're working on. We'll review every submission and respond within 24 hours.
Founder of Bosar Agency. Building AI systems that actually work for service businesses — voice agents, chatbots, and automations that capture leads 24/7.