Start with the non-billable hours
A firm selling time should automate the time it cannot bill. Administration, chasing, file setup, and reconstructing what happened last Thursday.
That converts overhead into capacity without touching the professional judgement clients are actually paying for.
Five that reliably pay
- Time capture from calendar, email and document activity, as draft entries
- Client onboarding — extracting data from forms and identity documents
- Document assembly from your templates and matter data
- File and correspondence summaries when a matter changes hands
- Deadline and obligation extraction from incoming documents
Time capture alone typically recovers three to six per cent of billable hours that were being worked and never recorded.
What not to automate
- Advice to a client, in any form that reaches them unreviewed
- Anything where your professional indemnity position depends on judgement
- Conflict checks, unless as a first pass a person confirms
- Anything a regulator would expect a named individual to have done
Confidentiality is the first conversation
Client material has obligations attached that go beyond data protection. Check your engagement terms and any professional rules on subprocessors before sending anything to a model provider.
Redaction, business-tier terms and regional processing usually resolve it. Doing that after the fact does not.
The realistic economics
| Integration | Typical build | Typical return |
|---|---|---|
| Time capture | £5,000–£9,000 | 3–6% billable recovery |
| Onboarding extraction | £6,000–£12,000 | 40–60% less admin per client |
| Document assembly | £8,000–£15,000 | Hours per matter |
| File summaries | £4,000–£8,000 | Faster handovers, fewer errors |