Harbor×LegalFab
Confidential proposal · Draft for discussion

Clearing conflicts at volume —
automatically, and defensibly.

A proposed pilot for Clyde & Co: resolve the entities behind high-volume casualty conflict checks in place, collapse tens of thousands of false positives, and prove that low-risk volume business can be conflict-cleared straight-through — on the systems Clyde & Co already runs.

New-business conflict checks Entity resolution Auto-clear · human review on hits KYC screening — proof-of-technology only
Harbor × LegalFab · Prepared for Clyde & Co
Cover
Draft
Harbor×LegalFab
In partnership · Prepared for Clyde & Co
Executive summary

The conflict-check bottleneck — and how we clear it

Clyde & Co's high-volume casualty practice generates conflict-check demand at a scale manual clearance cannot meet. The parties that must be conflict-checked are the insured — overwhelmingly individuals, some organisations — and the incumbent system cannot resolve them as entities. It matches on names. The result is a flood of false conflict hits, reported in the tens of thousands, worked by a team of four at a pace that puts quality at risk.

The problem today
10,000s
false-positive hits
  • Name-matching, not entity resolution
  • A team of 4, clearing manually
  • Pace outstrips achievable quality
The pilot target
− 85%
false-positive reduction (pilot)
  • Auto-clear the low-risk majority
  • Humans review genuine hits only
  • Every decision explainable to source
Illustrative pilot target — to confirm on Clyde data.
The production trajectory
~99%
toward straight-through processing
  • Resolve-in-place — data never moves
  • Runs on Intapp, Elite 3E & MAR
  • Foundation for firm-wide onboarding
Production trajectory, not a pilot commitment.

What we are proposing

An 8-week, success-based pilot focused on conflict-of-interest checks, run entirely within Clyde's on-premise environment. LegalFab resolves the entities behind new-business casualty conflict checks in place — across Intapp and Elite 3E, integrated to the MAR publisher that already parses inbound instructions — and auto-clears the low-conflict-risk majority, routing only genuine conflict hits to a human with a full, explainable rationale. Harbor, a Clyde & Co risk services partner, runs the incumbent process in parallel to benchmark the result. KYC-style screening is included only as a proof of technology for a future engagement.

Why it matters now

The pilot solves an operational pain today, but it also proves the technology for a broader onboarding and KYC engagement across the Firm. That timing is deliberate: supervision of AML in the legal sector is transferring to the FCA1 — a more intrusive, data-driven regulator. The firms that will absorb that shift comfortably are the ones that can already evidence real-time, resolved, auditable client and matter risk. This pilot is the first step toward exactly that capability.

Harbor × LegalFab · Confidential
Executive summary
Draft
Harbor×LegalFab
In partnership · Prepared for Clyde & Co
Context

A new regulator, a new evidential bar

The supervision of legal-sector financial crime is moving to the FCA — and the same resolution capability underpins both AML evidence and conflicts

For a generation, anti-money-laundering supervision of law firms has sat with the profession's own regulators — the SRA and the Law Society among more than twenty sectoral bodies.1 That is now ending. In October 2025 the Government named the Financial Conduct Authority as the single professional-services supervisor for AML and counter-terrorism financing1,2, and the enabling clauses entered Parliament in the Financial Services and Markets Bill in May 20263, with HM Treasury’s implementation roadmap following in June.3,4

The change is not merely administrative. The FCA supervises in a different register from the SRA's guidance-led tradition: sharper scrutiny, broader powers, and a data-driven lens.1 The FCA has levied penalties into the tens of millions5, and firms deemed high-risk can expect particularly close attention.

The operative phrase in the Law Society’s guidance on the new regime is that success under it will depend on a firm’s ability to maintain a real-time understanding of client and matter risk3. That is an evidential standard, not a policy one — and for financial crime it cannot be met from a firm’s own records alone. It requires resolving each party against external data — sanctions and PEP lists, adverse media, corporate registries and beneficial ownership — and showing, with lineage, who a party is and how that judgement was reached.

The engine that produces this is the same one proven in the conflict-checking pilot: entity resolution in place, with a full audit trail. Conflicts is the immediate, high-volume proving ground; the identical capability, enriched with external data, is what a data-driven AML supervisor will expect to see. It is why the KYC screening in this pilot is included strictly as a proof of technology — ahead of a future onboarding engagement, not as today’s scope.

From

Guidance-led supervision

SRA / Law Society · sector-specific · principles and policies.

To

Data-led supervision

FCA single supervisor · intrusive, evidence-driven · multi-million-pound powers · systems and controls tested against reality.

What it rewards

Real-time, resolved, auditable risk

Firms that can evidence client and matter risk on demand — with provenance — absorb the shift comfortably.

References
  1. HM Treasury — Reform of the Anti-Money Laundering and Counter-Terrorism Financing Supervision Regime: Consultation Response, 21 October 2025 (GOV.UK).
  2. Financial Conduct Authority — Statement on the Government’s decision on reforming AML/CTF supervision, 2025.
  3. The Law Society — Changes to the UK Anti-Money Laundering Supervisory Regime, 2026.
  4. HM Treasury — AML/CTF Supervision Reform: Duties, Powers and Accountability — Consultation Response, June 2026 (GOV.UK).
  5. HKA — Changing of the Guard: What FCA Supervision Could Mean for AML in the Legal Sector, December 2025 (FCA enforcement figures).
Transition is subject to enabling legislation, funding and a detailed delivery plan; commencement timing depends on the availability of parliamentary time.
Harbor × LegalFab · Confidential
Context
Draft
Harbor×LegalFab
In partnership · Prepared for Clyde & Co
Slide 1 · The problem

The volume business is drowning the conflicts team

High-volume casualty · the insured are mostly individuals · the incumbent cannot resolve them

Volume of noise
10,000s
false-positive hits

Every inbound cover request throws a wall of name-matches. The signal is buried in the noise.

The people
4
clearing it by hand

A four-person team works the queue manually, at a pace set by inbound volume, not by risk.

The consequence
Quality
at risk
speed vs. defensibility

When pace outstrips capacity, the honest position is that clearance quality cannot be assured.

One inbound party · the false-positive flood
INBOUND PARTY 1 insured NAME MATCH → TEAM OF 4 TENS OF THOUSANDS OF FALSE HITS · CLEARED BY HAND

The business context: Clyde's primary clients here are the insurers. The parties that must be conflict-checked are the insured — overwhelmingly individuals, occasionally organisations — arriving at high frequency through routine casualty cover. The conflict risk on any single party is low. The problem is not difficulty; it is volume without resolution.

Harbor × LegalFab · Confidential
1 / 5
Draft
Harbor×LegalFab
In partnership · Prepared for Clyde & Co
Slide 2 · Why it happens

The incumbent matches names. It cannot resolve entities.

One root cause explains the tens of thousands of false positives

Name-matching — today

"J. Smith" matches 900 records

  • Every string that looks alike becomes a hit
  • No notion of which J. Smith this is
  • No sense of whether a match is even adverse
  • Human effort scales with noise, not risk
The system throws the decision back to the person.
Entity resolution — the fix

One resolved party, scored once

  • Identifiers, context and relationships resolve the party to a single entity
  • Cleared against actual clients & matters, not strings
  • The low-risk majority clears itself
  • Human effort scales with genuine risk only
The system makes the decision, and shows its working.
900 look-alikes vs. one resolved entity
NAME MATCH Smith 900 look-alike records → all hits ENTITY RESOLUTION 1 one resolved party → scored once

The insight: the false-positive flood is not a tuning problem to be chipped away at — it is a structural artefact of matching on names. Resolve the entity first, and the overwhelming majority of hits simply never occur.

Harbor × LegalFab · Confidential
2 / 5
Draft
Harbor×LegalFab
In partnership · Prepared for Clyde & Co
Slide 3 · The approach

Resolve in place. Auto-clear the majority. Explain every hit.

On Clyde's own systems — nothing is migrated, nothing leaves the boundary

The pipeline · nothing leaves Clyde's boundary
01 · INGEST MAR 02 · ENTITY RESOLVE 03 · AUTO CLEAR 04 · HITS ONLY ROUTE auto-clear human review
Resolve-in-place

Data stays in Intapp, Elite 3E and MAR — on-prem, in Clyde's boundary. No warehouse, no copy, no migration.

Auto-clear · human-in-the-loop

The low-conflict-risk majority clears straight through. People see genuine hits only — and their time goes to judgement, not triage.

Explainable by design

Every hit carries its reasoning and its lineage back to the source record — audit-ready for internal risk and, increasingly, for the regulator.

Harbor × LegalFab · Confidential
3 / 5
Draft
Harbor×LegalFab
In partnership · Prepared for Clyde & Co
Slide 4 · The pilot

Eight weeks. Benchmarked by Harbor. Measured on the number that matters.

Success-based · in Clyde's on-prem environment · new-business casualty conflict checks

ParameterProposed
ScopeNew-business conflict-of-interest checks, high-volume casualty.
KYC screening — proof-of-technology only
Duration8 weeks · to confirm
EnvironmentClyde on-prem — Intapp, Elite 3E, MAR publisher
ModelSuccess-based / at-risk · to confirm
BenchmarkHarbor runs the incumbent process in parallel
Human roleReview on genuine hits only, with explainability
Primary success metric
− 85%
reduction in false positives (pilot)
~99%
production trajectory
STP
the end state: straight-through
Figures illustrative — pilot target to be confirmed against Clyde data. Production trajectory is a direction of travel, not a pilot commitment.
False positives collapse — pilot to production
INCUMBENT 100% PILOT TARGET 15% PRODUCTION ~1%
Also measured
Lap-time / turnaroundThroughput clearedAuto-clear rateExplainability of hits
Harbor × LegalFab · Confidential
4 / 5
Draft
Harbor×LegalFab
In partnership · Prepared for Clyde & Co
Slide 5 · The path

Clear the volume first. Then onboard the whole firm.

The pilot is a wedge, not an endpoint

Land the wedge · compound the estate
STEP 01 Casualty STEP 02 Onboarding STEP 03 Firm-wide ONE RESOLVED GRAPH · REUSED ACROSS EVERY DOWNSTREAM USE CASE
Why start here
  • Highest-volume, lowest-conflict-risk work — the cleanest place to prove straight-through clearance
  • An acute, measured pain with a team already at capacity
  • Runs on systems Clyde & Co already owns — no migration to prove the point
Where it goes
  • One resolved graph of clients, matters and parties — reused across every downstream use case
  • KYC proven in-pilot as the bridge to firm-wide onboarding
  • Real-time, auditable client & matter risk — ahead of the FCA transition, not behind it
The volume business is where we prove it. The firm is where we scale it.
Harbor×LegalFab
Harbor × LegalFab · Confidential
5 / 5
Draft