A 6-Attorney Firm Cut Staffing Costs 27% With AI — Without Replacing a Single Person
A 6-Attorney Firm Cut Staffing Costs 27% With AI — Without Replacing a Single Person
When a senior associate left Ad Astra Law Group, the San Francisco litigation firm faced a decision every small firm owner knows well: do we hire a replacement, or do we figure out how to do without?
They chose a third option.
Managing partner Katy Young deployed Legion, an AI litigation drafting platform, to absorb the departed associate's workload. The result: staffing costs dropped 27%. And the work didn't fall behind — it got faster.
Get the full picture. Go premium.
Weekly intelligence briefings, deeper analysis, and direct access to the full archive.
This is not a story about AI replacing staff. It's a story about what happens when a small firm uses a natural transition point to ask a different question.
The Decision Most Firms Don't Think to Make
When someone leaves a 6-person firm, the instinct is immediate: post the job, start interviewing, fill the role. The disruption risk feels highest during the gap.
But that gap is also when the existing team's actual workload becomes most visible. Before Ad Astra posted a job listing, Young assessed where the departing associate's time was actually going — not in general terms, but task by task.
The largest single time consumer: first-draft document generation. Litigation is documentation-intensive. Complaints, motions, briefs, discovery responses — all require a first pass before any attorney can do real legal work on them. That first pass is expensive when it takes two full days of an experienced lawyer's time.
AI handles first drafts differently than it handles most other legal tasks. The structure is known. The language patterns are learnable. The facts still require attorney input, but the scaffolding — the organizational logic, the boilerplate language, the format — can be generated and refined rather than built from scratch.
After implementing Legion, a 45-page complaint that previously required two full days came back in 2.5 hours.
The math doesn't require a spreadsheet: that's roughly 13 hours recovered per complaint. For a litigation practice running 15-20 significant matters per year, that's 200+ hours of senior attorney time redirected toward work that actually requires a senior attorney.
Why This Worked — and Why It Often Doesn't
Ad Astra's implementation succeeded for three reasons that most "we tried AI" stories miss.
They targeted the biggest bottleneck, not the easiest wins. First drafts are time-intensive and uncomfortable to delegate — which is exactly why they stayed manual for so long. Low-value tasks are the ones that get automated first in most firms. High-value, high-time tasks like first drafts are where the actual money is.
They used a purpose-built tool, not a general one. Legion is litigation-specific. It knows how complaints are structured, what language works in different jurisdictions, and where attorney input is required versus where the platform can generate confidently. General AI tools (ChatGPT, Claude in basic mode) can do this work, but require significantly more prompt engineering and review time to hit the same output quality.
They used a transition moment to force the decision. There is no good time to restructure how your firm works. Staff departures create urgency that breaks the inertia. Young could have kept the old process and hired a replacement; instead, she used the disruption as cover for a change that would have been harder to make during a stable staffing period.
The Principle Behind the Tactic
The Ad Astra approach reflects a pattern that appears consistently in small firm AI implementations that actually produce measurable results.
Find the task that costs the most time per output, and ask whether AI can replace the first pass.
In litigation: first-draft complaints, motions, and briefs. In accounting: first-draft tax return preparation, audit memos, and client communications from financial data. In consulting: first-draft market analysis and project status reports. In staffing: first-draft candidate summaries and client intake reports. In marketing agencies: first-draft client briefs, deliverable summaries, and campaign reports.
The principle is identical across firm types: a professional's time is expensive, and most professionals spend a significant fraction of it on structured first drafts that don't require judgment — they require format knowledge, language fluency, and information assembly. That's what AI does well.
The review still requires judgment. In Ad Astra's case, attorney review of an AI first-draft still takes time. What dropped was the time to have something worth reviewing — from two days to 2.5 hours.
How to Evaluate Whether This Works for Your Firm
Three questions determine whether this approach is worth pursuing:
1. What is your highest-cost structured output? Not the document that's hardest to write — the one that requires the most attorney (or accountant, or consultant) hours to produce, not because it's complex, but because it's long and format-dependent. That's your target.
2. Does a purpose-built tool exist for that document type in your practice area? General AI can be prompted to draft most documents, but purpose-built tools calibrated for your document type (litigation, tax, contract) will produce better first drafts with less prompt engineering. Legion for litigation, Black Ore for tax returns, ContractPodAi or Spellbook for contracts. If no purpose-built tool exists, general tools can still work — budget more review time.
3. Do you have a natural transition point coming? A staff departure, a shift from hourly to retainer, a new practice area launch, a partner bringing in significant new work — any of these creates a window to implement without fighting against an established workflow. If no transition is imminent, you can still implement. It's just harder to get people to change how they work without an external forcing function.
What 27% Actually Means
Staffing costs are typically the largest line item in a professional services firm's P&L — often 50-70% of revenue. A 27% reduction in that line item, at a 6-attorney firm generating $2-3M annually, is a material number. Not "we saved some time" material. "This changes what we can invest in and what we need to charge" material.
Ad Astra achieved it by doing one thing: not hiring a replacement and deploying the right AI tool instead.
They didn't restructure the firm. They didn't retrain staff. They didn't run a four-month implementation project. They identified their most expensive structured output, found a tool purpose-built to draft it, and used an associate departure as the forcing function.
Most small firms have an equivalent transition point available to them within any given 12-month window: a staff change, a service line that's stagnating, a role they've been thinking about eliminating or restructuring. The question is whether they use it.
The Action Item
Identify your firm's equivalent of Ad Astra's complaint drafting — the one output that takes the most time per deliverable, not because it's complex, but because it's long and format-dependent.
Then find out whether a purpose-built AI tool exists for it. Not ChatGPT. A tool built specifically for that document type in your practice area.
You don't need a staff departure to test it. Run a trial on three documents. Measure the before-and-after time honestly. If it passes the test — if the first draft is good enough that your review time stays the same or drops — you have your answer.
The question isn't whether AI can replace that work. For most structured legal, financial, and consulting documents, it already can. The question is whether you're using that capability yet.
The Crossing Report is published weekly for professional services firm owners navigating AI adoption. Subscribe free →
Frequently Asked Questions
What is the Ad Astra Law Group AI case study?
Ad Astra Law Group is a 6-attorney litigation firm in San Francisco that cut staffing costs 27% by deploying Legion, an AI litigation drafting platform, after a senior associate departed. Rather than hiring a replacement, managing partner Katy Young used the transition as an opportunity to absorb the associate's workload through AI-assisted first-draft document generation. The most cited result: drafting a 45-page complaint dropped from two full days to 2.5 hours.
What AI tool did Ad Astra use to cut costs?
Ad Astra deployed Legion, a litigation-focused AI drafting platform designed for first-draft document generation in litigation contexts. Legion is purpose-built for law firms rather than a general AI tool. The firm targeted their biggest workflow bottleneck — time-intensive first drafts — rather than attempting to automate everything at once. This targeted approach is a key reason the implementation succeeded.
Can a small firm cut staffing costs with AI without layoffs?
The Ad Astra case demonstrates a specific path: using a planned or natural staff transition (a departure, a retirement, a reduction in contract work) to absorb workload through AI rather than backfilling the role. This approach avoids morale and ethical complications of AI-driven layoffs while capturing the same cost efficiency. The firm reduced costs by not hiring, not by eliminating existing staff. For most small professional services firms, this is the least disruptive path to meaningful AI cost reduction.
What tasks can AI actually replace at a small law firm?
Based on the Ad Astra implementation and broader small firm adoption patterns in 2026, the tasks with the highest AI absorption rates are: (1) First-draft pleadings, complaints, motions, and contract drafts — AI handles the structure and boilerplate, attorneys revise. (2) Research memos on routine legal questions — AI generates the first pass, senior review catches judgment calls. (3) Client intake questionnaire generation — AI creates tailored intake forms from matter-type descriptions. (4) Billing entry summaries — AI drafts time narratives from call notes and document activity. Tasks that resist replacement: relationship management, strategic client counseling, and judgment calls on novel fact patterns.
Is 27% cost reduction realistic for most small law firms?
A 27% staffing cost reduction is achievable, but depends on the context. Ad Astra's result was driven partly by not backfilling a departed associate — a high-cost role. Firms that achieve the largest savings typically start with a specific high-cost workflow where AI absorption is clean (first drafts, research memos) and measure the result before expanding. Firms that try to automate broadly without a specific target rarely achieve the same numbers. The 27% is a real result, but it shouldn't be treated as a baseline expectation without matching the implementation conditions: a targeted workflow, a purpose-built tool, and a clear measurement of before-and-after.
Get the weekly briefing
AI adoption intelligence for accounting, law, and consulting firms. Free to start.
Free weekly digest. No spam. Unsubscribe anytime.
Related Reading
This is the kind of intelligence premium subscribers get every week.
Deep analysis, cross-sector patterns, and the frameworks that help professional services firms make the crossing.