Every legal billing software claims to save time. But how much time, exactly? And does AI-generated billing actually capture more than a lawyer typing entries from memory? We decided to find out with a controlled comparison.
We took a single 45-minute client conference — a real commercial litigation matter involving settlement negotiations, costs estimates, and next steps — and put it through two parallel billing workflows. One solicitor billed the meeting manually using their practice management system. Another used LexUnits, uploading the meeting recording and reviewing the AI-generated entries. Both started immediately after the meeting ended.
The results were illuminating, and not just in the ways you might expect.
The Test Setup
To keep the comparison fair, we established clear ground rules. Both solicitors attended the same meeting and took their own notes. The manual biller used Actionstep, which is representative of how most mid-tier Australian firms handle time recording. The AI biller used LexUnits with the meeting's audio recording.
We measured four things: the total time spent on the billing task from start to finish, the number of discrete entries generated, the total billable time recorded (in six-minute units), and the quality of the billing descriptions as assessed by a supervising partner who was also in the meeting.
Here is what happened.
Round 1: Speed
The manual process began with the solicitor opening Actionstep, navigating to the correct matter, and starting to type. She worked from a combination of her handwritten notes and memory. The first entry took about two minutes — she had to recall the precise wording, estimate the time, and select the correct activity code. Each subsequent entry took roughly 90 seconds. In total, she produced five entries in 12 minutes and 40 seconds.
The AI process began with uploading the 45-minute MP3 file to LexUnits. The transcription and AI processing took 2 minutes and 15 seconds. The solicitor then spent about 3 minutes reviewing the eight generated entries — tweaking one description, merging two entries that covered related topics, and adjusting a time estimate. Total elapsed time: 5 minutes and 30 seconds.
| Metric | Manual | LexUnits AI |
|---|---|---|
| Total time to complete billing | 12 min 40 sec | 5 min 30 sec |
| Time spent writing descriptions | ~10 min | ~2 min (editing only) |
| Time spent on admin (navigation, codes) | ~2 min 40 sec | ~1 min 15 sec |
The AI workflow was 57% faster. But speed alone is not the interesting part.
Round 2: Completeness
This is where the comparison became genuinely revealing. The manual solicitor produced five entries covering the main topics she remembered: the settlement discussion, costs estimate, evidence review, a brief mention of the upcoming mediation, and a catch-all entry for general file review and instructions.
LexUnits produced eight entries from the same meeting. The five core topics were all there, but the AI also identified three additional billable activities that the manual biller overlooked:
- Limitation period analysis — A five-minute discussion about whether a particular claim was time-barred. The solicitor had forgotten this entirely because it was a brief tangent in the middle of the settlement discussion.
- Costs disclosure obligations — The partner had spent three minutes explaining costs disclosure requirements under the Legal Profession Uniform Law. This was captured as a separate entry by the AI but lumped into the catch-all entry by the manual biller.
- Follow-up instructions — The client gave specific instructions about documents to request from the other side. The AI captured this as a distinct billable activity; the manual biller did not record it at all.
| Metric | Manual | LexUnits AI |
|---|---|---|
| Number of entries | 5 | 8 (after merging 2) |
| Total billable time recorded | 36 min (6 units) | 42 min (7 units) |
| Topics missed or underrecorded | 3 | 0 |
The revenue impact: At $500/hour, the one extra unit captured by the AI is worth $50 per meeting. For a solicitor who has four meetings a day, that is $200/day or roughly $48,000/year in recovered revenue — from a single feature.
Round 3: Description Quality
We asked the supervising partner to review both sets of entries without knowing which was AI-generated and which was manual. The partner assessed each entry on three criteria: accuracy (does it correctly describe what happened), specificity (is the description detailed enough to withstand client scrutiny), and professional tone (does it read like proper legal billing).
The verdict was surprisingly close. The manual entries scored well on professional tone — the solicitor was experienced and knew the conventions. But they lacked specificity in places. The entry "Conference with client regarding settlement" is accurate but tells the client very little about what was actually discussed or advised.
The AI entries were more granular. Instead of a single settlement entry, LexUnits produced "Conference with client regarding proposed settlement terms, including consideration of quantum for general damages and assessment of costs exposure if proceedings continue." This level of detail makes the entry far more defensible if the client queries the bill.
The partner did flag one AI entry as slightly awkward in phrasing and noted that one time estimate seemed generous. Both issues were resolved during the review step in under 30 seconds. Overall, the partner rated the AI entries as equal or better than the manual entries on all three criteria.
Round 4: Consistency Across the Day
Here is something the single-meeting comparison does not fully capture: the degradation of manual billing quality over the course of a day. The first meeting of the morning gets billed relatively well because the details are fresh. By the fourth meeting at 5 pm, when the solicitor finally sits down to record time, memory has faded significantly.
With AI-assisted billing, there is no degradation. The fifth recording of the day is processed with the same accuracy as the first. Every meeting gets the same thorough treatment regardless of when the lawyer gets around to uploading it. This consistency compounds over time — the revenue recovery is not just about individual meetings but about eliminating the systematic erosion that occurs with end-of-day or (worse) end-of-week billing.
What About Cost?
A fair comparison must account for the cost of the tool. LexUnits pricing starts at $19 AUD per month for the Starter plan with 50 credits. Processing a single audio file uses one credit. A busy solicitor averaging four recorded meetings per day, twenty per week, would need approximately 80 credits per month, fitting comfortably within the Pro plan at $29/month or the Premium plan at $49/month.
Compare that to the revenue recovery. Even conservatively, if the tool helps capture just one additional six-minute unit per meeting at a $500 hourly rate, the return is $50 per meeting. Over 80 meetings per month, that is $4,000 in recovered revenue against a $29–$49 monthly cost. The return on investment is not marginal; it is roughly 80 to 1.
Where Manual Entry Still Has a Place
To be fair, there are scenarios where manual entry remains appropriate. Quick file reviews, brief phone calls under five minutes, and administrative tasks that do not involve substantive discussions are often faster to record manually than to record and process through AI. LexUnits is not designed to replace manual billing entirely — it is designed to handle the high-value, high-complexity entries where the most revenue leakage occurs.
The ideal workflow for most solicitors combines both approaches: use AI billing for meetings, conferences, and any substantive work where a recording or transcript exists, and use manual entry for quick tasks where the overhead of uploading a file would not be justified.
The Verdict
Our side-by-side comparison produced a clear winner across every metric that matters. AI-assisted billing through LexUnits was faster, captured more billable time, produced more detailed descriptions, and maintained consistent quality regardless of when in the day the billing was done. The cost is negligible relative to the revenue recovered.
The most telling moment came after the test, when the manual biller reviewed the AI-generated entries and said: "I genuinely did not remember we discussed the limitation period issue." That single forgotten entry was worth $50. Across a year, those forgotten entries add up to tens of thousands of dollars.
The question is not whether AI billing software works. It is how much revenue you are comfortable leaving on the table by not using it.
FAQ: Does LexUnits replace my practice management software?
No. LexUnits is designed as a "last mile" tool that complements your existing system. It generates time entries from audio, documents, and emails, then exports them in the correct format for Actionstep, LEAP, Clio, or Smokeball. Your practice management software remains your system of record.
FAQ: How accurate are the AI-generated time estimates?
The AI estimates time based on the proportion of the recording spent on each topic. For a 45-minute meeting, the estimates are typically accurate within one six-minute unit. You can adjust any estimate during the review step before exporting.
FAQ: Can I use LexUnits for phone calls as well as meetings?
Yes. Any audio recording can be processed, whether it is a face-to-face meeting, a phone call, a Zoom conference, or a voice memo dictated after a hearing. The workflow is the same: upload the audio, review the entries, and export.