First 90 days and the business case
Systems before headcount: measure where the volume comes from, fix the causes in one practice area, prove it, then scale to all five, and hire for what is left.
A 30/60/90 plan with a checkpoint at each stage, and a simple model of what the fixes are worth: messages avoided × handling time → hours saved → hires deferred. Every assumption is labelled and editable.
Timeline, week by week
Thirteen weeks. Bold week numbers are checkpoints (day 30, 60 and 90), where the work stops if the numbers do not clear.
Days 1 to 30
Baseline, and the quick wins that need no software
- Sit in every queue: client success, case processing, intake, billing. Map how a message moves from inbox to answer, and where it waits.
- Hand-label 500 recent messages by practice area, intent and root cause. This is the test set for any classifier, rules or model.
- Baseline by practice area: first-response time, CSAT, Google rating and volume, complaint and escalation rates.
- Inventory the case management fields and matter stages, the inboxes and the client portal events a status update could be triggered from. Agree data handling with the COO and the lawyers: privileged content stays inside the firm's systems.
- Quick wins: one owner per inbox, a visible SLA queue, a complaint log with a root-cause field, a weekly operations review.
- People: structured screening and scorecards for the open roles, an interviewer load limit, draft compensation bands.
Days 31 to 60
Build and pilot in one practice area
- Pilot the triage queue in one practice area, where the fixes have most to remove. In this synthetic data that is Family sponsorship, with the highest share of silence-after-payment messages.
- Silent-file alert and proactive status updates triggered by matter milestones, plus a 14-day "nothing has changed" note.
- Consult no-show alert to the team lead within 15 minutes; no file closed without a lawyer-approved closure notice; a handoff note whenever a file changes owner.
- Draft replies from lawyer-approved templates only, with a lawyer signing off anything with a deadline, an escalation or a closed file.
- One document checklist per matter type in the pilot area.
- Onboarding automation live for the hiring wave; the other four areas continue as they are and act as the comparison.
Days 61 to 90
Scale to all five practice areas
- Roll out area by area with a short playbook for each: its milestones, its document checklists, its escalation contacts.
- Weekly dashboard to the COO and practice leads: CSAT, reviews, complaint and escalation rates, SLA, root-cause mix.
- Team leads own the routing rules and templates, with a change log, so the system improves without a ticket to IT.
- Re-run the business case with measured, not assumed, avoided volume, and reset the customer success hiring plan to match.
Scorecard
One headline number, escalations per 100 open files, because it captures what clients feel and cannot be improved by growth alone. Baselines are the last four weeks of synthetic data; targets are proposals to agree with the COO at day 30.
| KPI | Definition | Baseline (synthetic) | Day 60 target | Day 90 target |
|---|---|---|---|---|
| Escalations per 100 open files (headline) | Escalations to a manager, lawyer or outside body in 4 weeks, per 100 paid open files. Normalised so growth alone does not move it. | 1.77 | −25% in the pilot area vs. the other four | −35% across all five (≤ 1.15) |
| Silent files | Paid open files with no client update in 14 or more days. | 1,617 (29%) | Under 10% in the pilot area | Under 5% everywhere; none over 30 days |
| Speed to first contact | Median hours from a new lead or consult request to a person contacting them. | 13.9 h | Under 4 business hours | Under 2 business hours |
| Missed consultations | Booked consultations the firm did not call or join. | 9.1% | Under 3%, every miss called back the same day | Under 1% |
| CSAT | Share of survey responses scoring 4 or 5 out of 5. | 82% | Pilot area not below its baseline | +5 points across all five |
| Google rating, new reviews | Average star rating of reviews posted in the period. | 4.44 | Hold at or above baseline | Back above 4.6 |
Business case: systems before headcount
Defaults use the last four weeks of synthetic volume and the last eight weeks of root-cause shares. The avoidable rates and costs are assumptions; the 30-day baseline replaces them.
Inputs
Edit any number. FROM DATA comes from the synthetic history on the CSAT page; ASSUMPTION is a planning assumption to replace with the firm's own numbers in the first 30 days.
System fixes
Result
4 customer success hires deferred
at next year's volume, from 7,249 hours a year that no longer need doing.
- Messages avoided per week
- 467 (43%)
- Hours saved per week, messages avoided
- 70.1
- Hours saved per week, AI drafts on the rest
- 22.8
- Hours saved per year, today's volume
- 4,833
- Full-time equivalent, next year's volume
- 4.8
- Gross saving per year (whole hires only)
- $288,000
- Net of run cost
- $258,000
- Payback on the build
- 5.6 months
If every fix works half as well as assumed: 2 hires deferred, payback 12.6 months. That downside is why the pilot measures avoided messages in one practice area before anything is scaled.
Formula: messages avoided = weekly messages × Σ(share × avoidable). Hours = messages avoided × handling time, plus remaining messages × handling time × draft saving. FTE = annual hours × growth multiple ÷ productive hours. Hires deferred rounds down, because a fraction of a hire is not a saving.
What the model leaves out, on purpose
- Revenue effects: better reviews and fewer complaints should help conversion and referrals, but they are not counted. The case stands on cost alone.
- Lawyer and consultant time: fewer status calls frees licensed time too; only client success time is counted.
- Partial hires: a fraction of a person is not a saving until volume grows into it, so hires deferred rounds down.
- Real fees, salaries and volumes: none are known from outside the firm. Every cost figure here is a placeholder.