A boutique firm was drowning in intake forms, follow-up emails, and document prep. Two AI tools and 45 days later, the partners got 12 hours a week back.
Hargrove & Associates is a boutique firm with eight attorneys specializing in business litigation, contracts, and employment law. Good reputation. Strong client retention. A billing rate that should have made the practice extremely profitable. The problem was that too much of what the attorneys were doing wasn't billable.
The breakdown, tracked over a two-week audit period: each attorney was spending an average of 3.1 hours per day on administrative work. Not complex legal strategy. Not client consultation. Not depositions or filings. Admin: chasing intake forms, sending follow-up emails, formatting documents, coordinating scheduling, and reviewing client files to answer basic status questions.
At an average billing rate of $385/hour, 3.1 daily hours of admin across 8 attorneys equates to roughly $38,000/week in revenue that can't be captured. The firm's managing partner, Diana, had known this was a problem for years. The solution had always been "hire another paralegal." They were on their third paralegal hire in four years. The admin work kept expanding to fill whatever capacity existed.
"The work wasn't getting simpler," Diana said. "It was getting more complex and there was more of it. We were treating the symptom, not the cause."
“We weren’t spending 40% of our time on admin because we lacked staff. We were spending it because we hadn’t designed our processes to not require it.”
Diana H., Managing PartnerHargrove's IT consultant recommended a two-tool stack: XBert AI for financial and operational anomaly detection across client accounts, and Make.com to build the automation workflows connecting their intake, document, and CRM systems.
The decision to combine them rather than use a single platform was intentional. XBert surfaces insights from financial and operational data that would otherwise require an attorney to manually review. Make.com handles the workflow automation, moving data, triggering communications, and routing documents without a human in the loop.
| Process | Before | After | Time Saved |
|---|---|---|---|
| Client intake | Manual form collection, paralegal data entry, attorney review | Automated intake form → CRM entry → attorney notification with pre-summarized profile | 2.5 hrs/new client |
| Status update emails | Paralegal drafts, attorney reviews, attorney sends | Triggered automatically at case milestones, attorney reviews in under 60 seconds | 45 min/client/week |
| Document prep | Template pulled manually, attorney edits, admin formats | Matter data auto-populates templates; attorney reviews final draft only | 1.2 hrs/document |
| Billing anomaly review | Monthly manual review of client accounts for write-offs and overages | XBert flags anomalies daily; attorney reviews flagged items only | 3 hrs/week firm-wide |
| Follow-up scheduling | Email chains to coordinate attorney + client availability | Automated scheduling links triggered at case stage completion | 30 min/appointment |
Week one was process mapping. The IT consultant and Diana spent two days documenting every administrative touchpoint in the firm's client lifecycle, from first inquiry to case close. This step is often skipped. It shouldn't be. You can't automate what you haven't mapped.
Weeks two and three were Make.com workflow builds. The intake automation was first: new inquiry comes in via web form, data populates the CRM, a client profile is generated, and the relevant attorney receives a structured briefing note rather than a raw form. That alone saved the intake paralegal approximately 90 minutes per new client.
Week four integrated XBert into the financial review process. Each client matter was connected to the firm's billing system. XBert monitors for write-off patterns, billing gaps, and invoice anomalies, surfacing them daily rather than waiting for the monthly review session that attorneys dreaded.
Weeks five through six were testing, refinement, and attorney training. The training took less time than expected: attorneys adapted quickly once they saw that the AI was reducing their workload rather than adding a new system to manage.
The firm tracked admin time per attorney before and after deployment. At 90 days, average daily admin time had dropped from 3.1 hours to 1.85 hours, a reduction of 40.3%. Across 8 attorneys, that's 10 hours per day reclaimed firm-wide. At a blended billing rate of $385/hour, the theoretical recovered revenue ceiling is $3,850/day, though the real figure is lower, as not all recovered time is immediately billed. The firm estimated approximately $22,000/month in additional billable capacity created in the first quarter.
The more immediate, measurable number: paralegal overtime dropped to zero. The third paralegal hire, brought on 14 months earlier specifically to handle intake and follow-up volume, was reassigned to complex document review work that actually requires a trained professional. No one was let go. The capacity got redeployed to higher-value work.
The two-tool stack cost the firm approximately $340/month. The admin time recovered in the first 90 days represented an estimated $66,000 in additional billable capacity. The paralegal whose time was previously consumed by automatable work is now doing work that genuinely requires her expertise. The firm didn't shrink its team to save money. It automated the work that was burying its team, then redeployed that talent upward. That's the right way to do this.
The document automation required more attorney involvement in the template-building phase than anticipated. Attorneys had to review and approve each template before it could be automated. This created a two-week slowdown in week three. Lesson: involve the attorneys in template design from day one, not after the automation is built.
XBert's billing anomaly detection required calibration for the firm's specific billing practices. The initial alert volume was too high, flagging things that were intentional (like pro bono write-offs) as anomalies. Two weeks of tuning brought the false positive rate down to acceptable levels. This is normal for any AI monitoring tool and shouldn't be a deterrent, but plan for it.
The best AI tools, real case studies, and actionable guides, delivered every Thursday. No noise. Just signal.