Law-firm intake becomes expensive when staff rekey an enquiry, chase missing party details, search several systems and prepare the same summary for multiple reviewers. AI can help organise the work, but it should not decide whether a conflict exists, whether the firm may act or whether a prospective client should receive legal advice.
This guide proposes a workflow for commercial law firms. It is not legal or professional-conduct advice, and it does not claim Binarify client results. Each firm must define the intake, conflict, confidentiality, privacy, supervision and matter-opening requirements that apply to its practices and jurisdictions. See our AI consultancy for law firms page for the wider implementation approach.
Define the intake decision before choosing AI
“Automate intake” can describe several different jobs:
- respond to a new enquiry;
- collect prospective-client and matter information;
- decide whether the enquiry belongs with a practice group;
- prepare names and relationships for a conflict search;
- review possible matches;
- approve or decline a matter;
- issue engagement terms; or
- open the matter in the practice-management system.
These jobs do not share the same owner or risk. A bounded first outcome could be:
Prepare a complete intake and conflict-search packet for authorised review, with every fact linked to the prospective client’s submission or an approved firm record.
The packet prepares a decision. It does not make one.
Use the practice platform first
Many firms already pay for useful intake and conflict features. Clio Grow intake forms can collect contact and matter information and route a completed form into a pipeline. Clio conflict checks support person, company and keyword searches, result review, status assignment and a final report.
Configure the firm’s existing forms, fields, roles and reports before building a parallel intake database. A custom layer is justified only when a measured gap remains, such as structuring information from mixed channels, preparing complex related-party searches or coordinating handoffs across several approved systems.
A controlled law-firm intake workflow
1. Create one prospective-client record
Capture the channel, time, contact details, preferred communication method, practice area, jurisdiction, urgency and responsible intake owner. Attach the original form, email, call note or referral rather than retaining only an AI summary.
Use a stable prospective-client ID before any conflict search. Deduplicate cautiously. Two people with similar names may be different, and two spelling variants may refer to the same person.
Do not ask for a full matter history before the firm knows what information it needs and how it may use it. Prospective-client information may itself be confidential under applicable professional rules.
2. Collect facts needed for routing and searches
Use an approved question set for the relevant practice area. AI may classify a narrative answer and propose missing questions, but the firm defines which fields are mandatory.
Relevant fields might include:
- prospective client and related entities;
- counterparties, affiliates, witnesses and other involved people;
- relationship between the parties;
- matter type, jurisdiction and material dates;
- referral source and existing firm relationships;
- urgency or limitation information for immediate human attention; and
- documents the firm has approved for pre-engagement collection.
If the message contains a possible deadline, injunction, arrest, transaction closing or other urgent circumstance, route it immediately under a deterministic firm rule. Do not let an AI confidence score delay review.
3. Normalise parties without replacing the source
Prepare searchable name variants, former names, trading names, corporate identifiers and relationships. Retain the exact submitted wording beside every normalised value.
The system should show why it proposed a variant. A company suffix, punctuation change or common abbreviation may be appropriate. An inferred parent company or personal relationship requires evidence or clarification.
Use separate statuses for confirmed, client-stated, inferred and unknown. Do not turn an uncertain relationship into a fact merely to make the search easier.
4. Prepare and run approved conflict searches
Send the reviewed terms to the firm’s authorised conflict-search process. Respect role restrictions, information barriers and search scope. Record the system, parameters, time, operator and result set.
AI may group possible matches and bring the most relevant evidence together. It should not hide a low-scoring result or label a match harmless. Names, matter descriptions and relationships can be ambiguous, and a conflict analysis may require information the intake workflow cannot access.
5. Assemble the reviewer packet
Give the authorised reviewer:
- the original enquiry and current intake answers;
- every party and relationship, including uncertainty;
- search terms, variants and systems searched;
- possible matches with source records;
- missing or conflicting facts;
- the proposed practice group and intake owner;
- any client or referral-source restrictions; and
- the next deadline.
The reviewer records the firm’s permitted disposition, notes and required follow-up. The workflow must support “needs more information” and “refer to risk or conflicts counsel” rather than forcing a yes-or-no result.
6. Separate conflicts, fit and engagement
Conflict review is not the same as deciding commercial fit, capacity, expertise, credit, client due diligence or engagement terms. Keep these approvals separate even if one team coordinates them.
Do not send advice, promise representation or imply that the firm has accepted the matter before the authorised decision and required engagement steps are complete. Use approved holding language for prospective clients.
7. Open the matter only after approval
When the firm accepts the engagement, write approved data into the practice-management and document-management systems. Preserve the prospective-client record, conflict report, engagement approval, responsible lawyer and matter-opening checklist.
Use idempotent writes so a retry does not create a second client or matter. Failed writes need an owned exception queue. Confirm that permissions and information barriers are correct before documents move into the matter workspace.
Binarify intake decision packet
This proposed artifact keeps the evidence and authority for each step visible.
| Packet section | Source | Automated preparation | Authorised disposition |
|---|---|---|---|
| Enquiry | Original form, email, call note or referral | Classification and missing-field proposal | Intake owner confirms route and urgency |
| Parties | Prospective-client statements and firm records | Name normalisation and relationship map | Staff confirm terms for search |
| Conflict search | Approved firm systems | Result grouping and evidence assembly | Authorised reviewer records status |
| Matter fit | Practice criteria and capacity | Checklist and missing information | Practice owner decides fit |
| Engagement | Approved terms and client requirements | Document assembly and field population | Lawyer approves issue and acceptance |
| Matter opening | Approved client and matter record | System writes and validation | Matter owner confirms successful opening |
Store each version and reviewer change. A corrected party name should not erase the search terms and evidence used for an earlier decision.
Confidentiality, supervision and prospective clients
The Solicitors Regulation Authority’s 2026 warning notice says AI does not diminish or transfer professional responsibility and highlights confidentiality, accuracy, supervision and human oversight. ABA Formal Opinion 512 discusses duties involving current, former and prospective client information in the US model-rules context.
The firm should approve which tools may receive prospective-client data, what supplier terms apply, where data is processed, whether inputs are used for training, who can see the record, how long it is retained and how information barriers are enforced. Client or court instructions may impose additional restrictions.
These controls require jurisdiction-specific professional advice from the firm’s responsible people. A vendor’s enterprise label or general security certification does not by itself decide confidentiality, privilege or professional compliance.
Measure intake quality and conversion separately
Baseline by practice area and enquiry channel. Track:
- time from enquiry to review-ready packet;
- staff minutes spent rekeying, chasing and assembling searches;
- enquiries missing a required party or relationship at first review;
- possible matches omitted or incorrectly grouped;
- matters opened with incorrect client, owner, permissions or metadata;
- time awaiting prospective-client information and internal review;
- accepted, declined, referred and abandoned enquiries using agreed definitions; and
- complaints, confidentiality events and advice given before engagement.
A higher consultation-booking rate is not automatically better. The workflow should help suitable enquiries reach the right reviewer while urgent, conflicted, unsuitable or incomplete cases receive the firm’s approved next step.
Pilot one practice area
Choose a practice area with a stable intake form, sufficient volume, named conflict reviewers and a known matter-opening process. Run in shadow mode first: prepare packets without changing the current decision or system of record.
Sample accepted, declined, urgent and ambiguous enquiries. Compare completeness, search coverage, preparation time, reviewer corrections and downstream matter-opening errors. Define stop conditions for missing parties, inappropriate communication, permission failures and incorrect system writes.
Continue with source-linked AI legal document review after a matter is opened. Use the law-firm AI ROI guide to determine whether any released capacity becomes measurable firm value, or book a 30-minute conversation about one intake queue.