Billables AI typically achieves 80–90% accuracy on client/matter matching (auto-assignment of clients and matters to your time entries). Actual accuracy varies by firm, workflow, and the complexity of your client/matter structure (for example, firms with many similarly-named matters or infrequent correspondents tend to see more variance). The good news: matching is a learned behavior, not a fixed setting, so accuracy climbs steadily as you use the product. Below are the practices that move the needle most.

Best Practices for Client Matter Matching

1. Edit and approve time entries consistently

Billables learns from every entry you touch. The system also weights recently used clients and matters more heavily, so the matters you're actively working tend to get matched faster and more reliably than ones you haven't touched in a while.

Two nuances matter here:

• Edit before you approve, not after. Approval is treated as the "source of truth" signal — once an entry is approved, Billables learns from that final version. If you approve first and fix mistakes later, the system doesn't see the correction as a learning signal.

• Make it a daily habit. If entries sit untouched until the end of the week, Billables gets no learning signal for that entire stretch, and any matching drift compounds across the week instead of correcting itself day by day. A quick daily pass keeps the feedback loop tight.

2. Favorite the clients and matters you use most

Star your most frequently used clients and matters on the Billing Codes page. Favorited codes are weighted more heavily in matching, which is usually the fastest way to boost accuracy on your core caseload with almost no ongoing effort.

3. Fall back on billing guidelines when implicit learning isn't enough

The first two practices rely on implicit learning — Billables inferring matches from your edits, approvals, and favorites over time. For most matters, that's sufficient. But if a particular client or matter keeps mismatching despite consistent edits and favoriting (for example, a correspondent who spans multiple matters, or a client whose activity looks similar to another client's), use billing guidelines to give Billables an explicit rule instead — for example, "emails with [correspondent] should be matched to [client/matter]." Treat this as a targeted fix for the matters implicit learning struggles with, not a first-line habit for everything.

4. Give the system a head start with historical data

If you're still early in onboarding, consider uploading a representative sample of previously approved time entries across different clients, matters, and staff. Pre-training on real historical entries helps Billables pick up your firm's narrative style and matter-specific conventions faster than starting from zero. Note that this may be available only on premium tiers — reach out to your Billables AI account rep to learn more.