AI for Drafting Client Letters in Law Firms (With Outlook + DMS Integration)

TL;DR: Most AI tools marketed to law firms either require uploading documents to a third-party vendor (privilege concern), don't integrate with Outlook and the firm's DMS (workflow concern), or train models on customer data (ethics concern). Aeralis sidesteps all three: it lives inside Gmail or Outlook as an add-on, reads only the message you're actively replying to, doesn't train on your content, and can reference your matter documents on demand via Knowledge uploads — without scraping the inbox or routing through anyone else's systems. This guide covers the actual setup, the ethics considerations, and where to start.

---

The legal market is full of AI tools claiming to draft client letters, summarize cases, and replace junior associates. Most of them have one of three problems that should disqualify them for a small or mid-sized firm:

1. They want full access to your inbox or your DMS — usually as a "service account" — which creates a privileged-information pathway that any malpractice carrier will flag.
2. They don't integrate with Outlook (which is what 80% of small and mid-sized firms use), forcing a separate app + copy-paste workflow that breaks during high-stress moments.
3. They train their models on customer data, sometimes contractually, sometimes opaquely. For a firm holding client confidences, this is a non-starter regardless of how well it's masked.

There's a workable AI workflow for client letters that avoids all three. This is what it looks like.

The constraints that matter for a law firm

Before you evaluate any AI tool, get clear on what's actually required.

Start with privilege and confidentiality: the tool can't retain or use client content for anything other than the immediate task. "Aggregated" or "anonymized" use of client data is still problematic for most malpractice carriers in 2026. The contract is the source of truth here, not the marketing page.

DMS coexistence matters next. The tool needs to coexist with iManage, NetDocuments, or whatever DMS your firm runs — not replace it, not route through it, not require a connector your IT team has to maintain. The simpler model: the AI works in the user's email client, the DMS holds the file.

Outlook-first is third. Most small and mid-sized law firms run on Microsoft 365. Tools that "support Outlook" but are really Gmail-first with an Outlook port are reliably worse — look for a native Outlook add-in.

An audit trail is fourth. Anything the AI helps draft should map to a specific attorney in the firm's records. If you're audited or face a malpractice claim, "the AI sent it" is not a defense.

And finally, no training on customer data. This should be contractual, not promised in a blog post. The privacy policy is the source of truth.

Where Aeralis fits these constraints

Aeralis runs as a Microsoft Outlook add-in (and a Gmail add-on, for firms still on Google Workspace). The architecture is intentionally narrow. Each lawyer installs the add-in on their own Microsoft 365 or Google Workspace account — there's no firm-wide service account, no tenant-wide read access. When an attorney composes a reply or drafts a new email, Aeralis sees that thread and nothing else; it can't search the inbox, list folders, or read other clients' matters. Email content is processed in real time and then discarded. Nothing is stored on Aeralis servers beyond anonymous generation metadata — no email bodies, no recipient names. There's no model training on client content; that's contractual, not just promised. Aeralis uses foundation models (Gemini Flash) via API with the no-training flag set. Knowledge uploads on the Business plan let the attorney reference firm-specific documents (form letters, internal style guides, jurisdiction-specific templates) without those documents leaving the firm's session.

The whole surface is deliberately narrow. Aeralis isn't trying to replace your DMS or your case management. It writes the email; your DMS holds the file.

The DMS workflow that actually works

The pattern most lawyers settle into:

1. Draft the response in Outlook, with the matter thread open. Aeralis is in the sidebar.
2. Optional: open the relevant matter folder in your DMS in a separate window for reference.
3. Tap "Generate" in Aeralis with a one-line prompt like "Draft a polite holding response acknowledging receipt and confirming we'll have substantive comments by Friday."
4. Edit the generated draft to add matter-specific details (case numbers, opposing counsel names, particular legal positions).
5. Save the final email to the matter in your DMS through your DMS's normal Outlook integration (iManage, NetDocuments, etc., all have one).

Critical: Aeralis doesn't touch the DMS. The DMS still gets the email saved via its own Outlook integration, exactly as it always has. The AI tool is upstream of the DMS, not parallel to it.

What it's good at vs. what to keep human

Good fit for AI assistance:

• Routine acknowledgments ("Got your email, will respond by [date]")
• Status updates on matters (no substantive legal content)
• Client communications about logistics (meeting scheduling, document requests)
• First drafts of demand letters in templated areas (debt collection, basic contract enforcement)
• Translating between formal and conversational registers
• Plain-English explanations of complex points for client letters

Keep human-written:

• Anything containing substantive legal analysis or strategy
• Demand letters where the language is the lever
• Court correspondence
• Anything involving privileged information you wouldn't share with a co-counsel without thought
• Engagement letters, fee agreements, settlement communications
• Anything where opposing counsel might later quote your exact words

The realistic time savings is on the first category — the templated 30–40% of an attorney's email volume. That's enough to pay for the tool many times over, without crossing the line into substantive legal work.

Practical setup for a small firm

For a 5–20 attorney firm running Microsoft 365:

1. One attorney pilots for two weeks on the free tier (15 emails/month). This is enough to test the workflow on real matters without committing the firm.
2. Set up one or two profiles in Aeralis — typically "Client correspondence" and "Opposing counsel correspondence" — with system instructions that match the firm's voice.
3. If it works, roll out Pro ($14/attorney/month) or Business ($29/attorney/month with Knowledge uploads) to the broader team.
4. Document the workflow in the firm's IT and ethics protocols. The tool's scope (per-attorney, no DMS access, no inbox read beyond active thread) is what your malpractice carrier will want to see.
5. Quarterly review of what's being drafted with AI assistance. Verify that the substantive legal work is still being written by attorneys, not AI.

Where Aeralis doesn't fit

A few honest limits. If your firm requires on-premise AI inference, Aeralis won't work — it runs on hosted infrastructure via the Gemini API, with no on-prem option. If you need full contract review or case analysis, Aeralis isn't a legal research tool. It's an email tool. Use it alongside Lexis, Westlaw, or Harvey, not instead of them. If your DMS doesn't support Outlook integration at all (rare in 2026, but it happens), the workflow above breaks and you'll need to save emails to matters manually. And if your firm has a strict no-cloud-AI policy, Aeralis won't fit — this is increasingly rare for non-government work, but it's not unheard of.

See also

• /solutions/law-firm-email-automation — the full law firm use case
• Are All AI Email Tools the Same? — what actually varies vendor to vendor
• /alternatives/transparency-matrix — pricing and privacy comparison
• AI Email Drafts: Best Practices & Tips — reliable output techniques

---

If you're at a small or mid-sized firm and want to test the AI email workflow without committing the firm, Aeralis's free tier is the right entry point. One attorney, fifteen drafts a month, two weeks to decide. No credit card.