Conflict Checks at Scale: The 1990s Workflow Most Florida Law Firms Are Still Running Against 2026 Liability Exposure

Conflict checks fail in three places, and only one of them is in the database. The first is at intake, where the client names a counterparty using their commercial name and your search runs against legal entity names. The second is in the matter history, where representation arcs are summarized in time-entry narratives that no name-search query indexes against.

The third failure point is the one firms rarely admit: the person running the check does not know what they are looking for. They run a name, get zero hits, mark it clean, and move on. Florida Bar Rule 4-1.7 does not care how fast you ran the search. It cares whether you actually identified the conflict.

I have come to believe that conflict checking is the single workflow in a mid-size law firm where the gap between the process operators think they have and the process they actually have is widest. The consequences are not theoretical. The ABA Standing Committee on Lawyers' Professional Liability publishes a quadrennial Profile of Legal Malpractice Claims, and conflict-of-interest failures appear consistently among the recurring claim categories by frequency. ALPS underwriters and Hanover's professional liability division have publicly discussed conflict-related claims as a persistent driver of mid-size firm exposure.

These are not edge cases from boutique operations. They are claims filed against firms with 10 to 50 attorneys running the same intake workflow they built in 2003.

Client-Reported Names at Intake

When a prospective client calls your intake coordinator and says, "I need you to represent me in a dispute with Coastal Development Partners," that name goes into your system as typed. Your conflict search runs against it. If you previously represented a party adverse to an entity legally registered as CDP Holdings, LLC (which does business as Coastal Development Partners), you get a clean result.

This is not a technology failure. This is a workflow design failure that technology can address.

The specific problem is that intake forms capture reported names, not legal entity names. Entity resolution, meaning the work of matching a reported commercial name against its registered legal entity, its parent, its subsidiaries, and its known aliases, requires a secondary lookup that most intake workflows skip entirely. Clio, one of the most widely deployed practice management platforms in Florida, runs conflict search across contacts, matters, notes, communications, and custom fields, but it does not perform external entity resolution against Sunbiz, PACER, or Dun and Bradstreet. If "CDP Holdings, LLC" is in your system and "Coastal Development Partners" is not, the search returns nothing.

The resolution logic has to be built into the intake process itself. The two options are a documented manual step (unlikely to survive staffing turnover) or an automated entity-enrichment layer that pulls from Secretary of State records, PACER, Westlaw entity data, or commercial databases like Dun and Bradstreet before the conflict query runs. The enrichment lookup adds 30 to 90 seconds to the intake workflow. That is acceptable. What is not acceptable is a 20-minute intake call where confidential information has already been shared before any check runs.

Matter Histories and the Representation Arc Problem

Most conflict search systems query against matter names, party names, and attorney-of-record fields. They do not query against time-entry narratives, correspondence logs, or file notes. This creates a structural gap that grows larger every year a firm has been in practice.

Consider a firm that represented a commercial landlord in a lease dispute in 2017. The landlord was one of three partners in a real estate holding company. The matter closed, the client record sits in the system, and eight years of time passing means nobody at the firm has active memory of the representation. In 2025, a new matter comes in: a tenant suing one of those original partners personally over a different lease. The conflict check runs against the 2017 matter name, the holding company entity, and the landlord's name. The personal name of the individual partner does not surface because he was never named as a party; he was documented in file notes as a principal of the client entity.

Florida Bar Rule 4-1.9 governs duties to former clients. It does not require that the former client was named as a party. It requires that the current representation is materially adverse to a former client in a substantially related matter. "Materially adverse" and "substantially related" are fact-specific determinations that Florida federal courts have repeatedly used as the basis for disqualification orders where the connecting tissue between matters lived in file notes and engagement context, not in formal party-name fields. The conflict check is supposed to surface the potential issue so a human can make that determination. If the matter history is not indexed in a way that surfaces the relationship, the human never gets the opportunity.

Agent-driven workflows parse time-entry narratives, file notes, and correspondence metadata to build a structured relationship graph that sits underneath the party-name index. The workflow does not determine whether a conflict exists; that is the attorney's call. It extracts entity mentions, relationship labels, and matter-role designations from unstructured text and populates a searchable graph. The search thus runs against relationships, not just names.

The legitimate counterargument here is confidentiality. Running a language model over privileged client communications, time-entry narratives, and file notes raises Rule 4-1.6 obligations and squarely implicates Florida Bar Ethics Opinion 24-1 on generative AI use. The answer is architectural: on-premise inference or private-cloud deployment, contractual no-training terms with the model provider, and an audit log that confirms no privileged content left the firm's perimeter. If a vendor cannot describe their deployment model in those terms, you are trading a conflict problem for a confidentiality problem.

Subsidiary and Affiliate Chains

This is the failure mode that produces the largest claims. A firm represents a major commercial bank on a routine matter. The bank's mortgage subsidiary, a separately registered LLC operating under a different trade name, becomes an adverse party in a foreclosure defense case the same firm picks up six months later. Name search: clean. Entity resolution against the parent's subsidiary chain: disqualifying conflict.

The Florida Secretary of State's Sunbiz database is publicly queryable and contains registered agent data, officers, and directors for Florida-registered entities. What it does not contain is parent-subsidiary disclosure; Florida filings do not require it. What Sunbiz surfaces is shared officer and registered-agent overlap that may indicate affiliation between entities, which is useful but not dispositive. No standard practice management system queries Sunbiz as part of a conflict check. The lookup is manual, optional, and functionally absent in most intake workflows.

PACER contains party-name data from federal filings that can surface relationships not visible in state corporate records. Thomson Reuters Company Investigator provides entity relationship mapping with documented parent-subsidiary linkages where they are public. None of these are integrated into a standard Clio or MyCase conflict workflow. They require either a manual step performed by someone who knows to perform it, or an automated enrichment pipeline that queries these sources when a new party name enters the system.

The agent-based approach connects intake data to external entity registries, runs the subsidiary chain lookup, and returns a structured result that includes flagged relationships before the conflict search runs against internal matter history. The workflow is the same; the input to the search is richer.

Rate Exposure When Checks Fail

I want to be specific here because operators respond to financial data faster than they respond to process arguments.

Conflict-related malpractice claims at smaller and mid-size firms tend to settle in the low- to mid-six-figure range for defense and indemnity costs, with outliers reaching seven figures when the underlying case has significant damages. Carriers track these patterns at the portfolio level. Hanover's professional liability application, along with the supplemental risk questionnaires from ALPS and most other PL carriers, explicitly asks whether the firm has documented conflict-check procedures, system logs, and supervisor sign-off. Firms that can produce that documentation receive better underwriting outcomes than firms that cannot.

The documentation question is separate from the accuracy question. A firm can have a poor conflict check process that is well-documented, or an undocumented process that happens to catch most conflicts. Neither is adequate. From an insurance standpoint, the documented process is the one that demonstrates diligence and supports a coverage defense if a claim is filed.

An agent-driven workflow produces a timestamped, logged output for every conflict search. The search parameters, the sources queried, the entity enrichment steps, and the result are all captured. That log is the documentation. It exists as a byproduct of the workflow, not as a separate compliance task someone has to remember to complete.

Four Questions for the Vendor Demo

If you are evaluating a platform or an agency offering conflict-check automation for your firm, these four questions will separate vendors who have built the workflow from vendors who have built a demo:

1. When a new party name enters the system at intake, which external entity registries does your workflow query before running the internal conflict search, and can you show me the API call logs from a live matter?

2. How does your system handle the subsidiary chain for a Florida-registered entity where the parent is incorporated in Delaware and the adverse party is a second-tier subsidiary? Walk me through the lookup.

3. Can your conflict search return results from unstructured text in time-entry narratives, not just structured party-name fields? If yes, what is the extraction method, what is the false-positive rate on your test set, and how do you keep the privileged content inside our perimeter under Florida Bar Ethics Opinion 24-1?

4. What does the audit log look like, and does it capture enough information to demonstrate reasonable diligence to an E&O carrier if a conflict claim is filed two years after the matter closes?

If the vendor cannot answer questions two and three with specifics, you have a name-search tool wearing automation language. That is the 1990s workflow with a new interface. Florida Bar Rule 4-1.7 does not grade on interface design.