Modern Communications Surveillance in Financial Services: A Practical Guide to the Challenges and How to Get Ahead of Them

Financial firms have spent a decade building surveillance programs that satisfy regulators on paper. The 1LoD Surveillance Benchmarking Survey shows those same programs straining under data fragmentation, alert noise, expanding coverage mandates, and an AI transition almost nobody has resourced.

The answer is to consolidate capture, archiving, surveillance, and governance onto a single, AI-native platform, the category the industry calls Digital Communications Governance and Archiving (DCGA), with explainable AI Surveillance running on that same foundation as one system.

This guide explains the problems, who they affect, the solutions available, and why a unified DCGA and AI Surveillance platform is the right move.

The data behind this guide

This guide draws on the 1LoD Surveillance Benchmarking Survey & Report, among the most authoritative reads on the state of trade, e-comms, and voice surveillance across financial institutions. (Download the full report from 1LoD & Shield)

It is built on hundreds of conversations and survey results from surveillance leaders, including 1LoD’s Surveillance Leaders’ Network, and carries a foreword from the FCA’s Head of Secondary Market Oversight, published as the UK Market Abuse Regulation approached its tenth anniversary.

Shield was one of the lead sponsors of that research, because its findings describe, almost line by line, the problems Shield was built to solve.

This guide is a faithful summary of what surveillance professionals reported, followed by a practical playbook for acting on it.

The core tension: a function at an inflection point

Surveillance looks stable on the surface and is anything but underneath.

On one hand, 93% of firms are confident or very confident their surveillance function can detect future market abuse, and 78% say they have sufficient budget to run their business-as-usual (BAU) surveillance obligations.

On the other hand, almost a quarter of firms (22%) admit their current level of spend does not effectively manage market abuse risk. That figure is unchanged from 2024, and the report notes it would alarm regulators.

That contradiction is the story. Firms can keep the lights on. What they cannot do, at current resourcing and on current architecture, is make the leap from a reactive, compliance-driven control to a modern, preventive, intelligence-led risk function.

The technologies that would enable that leap (cloud data platforms, AI, large-scale behavioral analytics) now exist and are widely available. The capability is there. The open question is whether a firm’s data and operating model can evolve fast enough to use it.

1LoD calls this “the calm before the surveillance storm.”

The four hard truths from the 1LoD data

For each problem below: what the survey found, who it affects, and why it matters beyond the surveillance desk.

1. Alert noise is the industry’s most widely shared pain

What the data says: The high volume of false positives is the single most cited challenge in the survey. 52% of firms rate it a “high” challenge and a further 41% a “medium” one, meaning 93% of institutions consider false positives a meaningful operational problem. One global head of trade surveillance captured the absurdity of the status quo: teams are still fighting to move from a 99% false-positive rate to 95%, and questioning whether that effort is even worth it.

Who it affects: Level-1 analysts who burn out triaging noise; investigations teams whose real cases sit behind a wall of irrelevant alerts; the CCO who must explain review backlogs to regulators; and the firm, which carries the risk of a genuine signal lost in the noise.

Why it matters: Noise is not just an efficiency problem. When teams can only review a sample of alerts, real misconduct hides in the unreviewed remainder. In one anonymized real-world case, detailed later in this guide, a global institution could review only about 15% of randomly sampled alerts before modernizing. That’s a textbook compliance blind spot.

2. Data fragmentation is the Achilles heel

What the data says: 71% of firms say that fragmented data across silos, the lack of standardized formats and identifiers, and poor or inconsistent data quality are together the biggest hindrance to surveillance effectiveness. Asked to isolate the worst data problem, responses cluster tightly: 24% fragmentation, 24% lack of standardized formats or identifiers, 23% inconsistent or poor-quality data.

For 37% of firms, limited access to quality data is a “high” challenge in its own right, and legacy or outdated systems are a “high” challenge for another 37%.

Who it affects: Surveillance teams increasingly doing data-engineering work they were never meant to do: managing hundreds of connectors, reconciling feeds, chasing venue data.

Technology and data-governance teams pulled into surveillance remediation. And ultimately the analytics layer, because even the best AI model is only as good as the data it ingests.

Why it matters: This is the report’s central conclusion. The primary constraint on surveillance is the weakness of the underlying data foundation, not analytical capability or a shortage of clever tools. Poor data quality is largely a downstream symptom of fragmentation and missing standards.

Until fragmentation, standardization, and source-level quality are fixed, the more ambitious goals of holistic surveillance, AI-driven detection, and real-time monitoring remain out of reach in practice.

3. Scope is expanding faster than teams can scale

What the data says: Surveillance perimeters are widening on every axis. More than half of firms now trade on 50+ venues. 22% monitor more than 30 e-comms channels (up from 14% in 2024). The share of firms covering only five or fewer languages has almost halved to 26%. Coverage is spreading into custody (30%), wealth (44%), and retail (22%), business lines that historically sat outside the core market-surveillance perimeter.

Most strikingly, 89% of firms now use communications surveillance to proactively monitor culture and conduct, up from 59% in 2024, accelerated by the FCA’s PS25/23 (December 2025), which links bullying, harassment, and inappropriate tone to the regulatory Fit and Proper test.

Yet teams have not grown to match. In e-comms surveillance, 80% of firms have 20 or fewer full-time-equivalent reviewers while surveilling populations in the thousands across dozens of channels and languages.

Who it affects: Every line of the organization now generating regulated communications: wealth advisers, relationship managers, and client-facing staff, plus the surveillance leaders being asked to cover them without proportionate headcount.

Why it matters: The report’s blunt summary: surveillance is scoped like a broad perimeter function but staffed like a super-lean ops team. As one global head put it, the only realistic pitch for more resource is to “turn FTE into technology.”

Coverage cannot be solved by hiring. It can only be solved by scale that technology provides.

4. AI ambition is real; AI adoption is not

What the data says: There is a yawning gap between what firms want and what they run. AI-enhanced trade surveillance is present in just 11% of firms, but 89% say they want it. Generative-AI assistants for analysts exist in only 7% of firms, with 78% expressing demand. Almost no AI is fully operational: 0% of firms have AI fully embedded in trade or holistic surveillance, with only 8% in e-comms and 9% in voice. The majority are still in an exploration and research phase.

Trade lags furthest, with 44% still in research and 52% in proof-of-concept, because layering large language models onto structured, model-governed trade data raises validation and explainability problems that e-comms and voice do not.

Who it affects: Model-risk-management (MRM) teams now trying to retrofit probabilistic AI into governance frameworks built for deterministic, rules-based models. The survey’s top model-risk challenges include validating complex NLP/LLM models (37%) and insufficient explainability of AI outputs.

Why it matters: The firms that win the next phase will not be those with the flashiest AI claims. They will be those whose AI is explainable, validatable, and defensible to regulators.

Black-box AI that cannot show why it produced an alert is a governance liability in a surveillance context.

Why these problems matter beyond your desk

Surveillance failures do not stay contained. The survey, and the FCA foreword that opens it, make three escalations explicit.

• Regulators have raised the bar. The FCA now treats a granular market-abuse risk assessment, surveillance controls mapped to risks, venue inventories, and strong data governance as fundamentals, not optional extras, and signals it will consider regulatory action where they are absent. One UK surveillance head described a recent supervisory review with over a hundred information requests on surveillance alone.
• The firm carries the consequence. Enforcement actions in recent years have repeatedly cited gaps in channel capture, venue coverage, and the ability to see what clients are doing. A blind spot is no longer an internal inconvenience; it is a fine, a remediation program, and reputational damage.
• The market relies on it. Clean, fair markets depend on firms detecting abuse before it harms participants. As the FCA foreword frames it, today’s challenges become tomorrow’s fundamentals, and standing still means going backwards.

The practical playbook: how to get ahead of the storm

Here is how surveillance leaders can act on each finding.

1. Attack noise at the source, not the symptom. Reducing false positives by re-tuning legacy lexicons yields diminishing returns. The durable fix is contextual, behavior-aware detection that scores risk rather than matching keywords, paired with AI-driven triage that lets a small team run a large process. Demand precision metrics (alert rate, precision rate, escalation quality), not just volume reduction.

2. Fix the data foundation first. Before buying more analytics, map your data supply chain: every venue, every channel, every identifier. Standardize formats and resolve a single unique identifier per individual across voice and e-comms; the survey shows this remains unsolved at many firms and blocks everything downstream. Treat data completeness and reconciliation as a first-class control, because regulators now do.

3. Plan coverage as a permanent program, then scale it with technology. New channels, languages, venues, and business lines will keep arriving. Build an onboarding process for them, and use AI transcription, translation, and NLP to extend coverage without linear headcount growth. Use management information to detect unmonitored languages and channels from observed behavior rather than static assumptions.

4. Adopt AI where it is provably explainable. Start with the use cases where AI already delivers, language analysis in e-comms and voice, and insist on transparency: full documentation, validatable outputs, and the ability to explain any alert to compliance, MRM, and regulators. Treat explainability as a procurement requirement, not a feature.

5. Ringfence change capacity. The survey’s quiet warning is that almost no firm has dedicated transformation resource; change-the-bank work is done side-of-desk by overstretched teams. Protect a transformation budget and team, or partner with a vendor willing to carry part of the change burden, or the transition simply will not happen.

6. Consolidate the estate. Every problem above is made worse by fragmentation. The strategic answer, explored next, is to stop integrating point tools and move to a single, unified DCGA and AI Surveillance platform.

The strategic answer: a unified DCGA and AI Surveillance platform

Read the five findings together and a single conclusion emerges. Noise, data fragmentation, coverage strain, the AI gap, and the failure of holistic surveillance trace back to one root cause: surveillance built as a patchwork of disconnected systems sitting on a fractured data layer.

You fix it by unifying capture, archiving, surveillance, governance, investigations, and eDiscovery on one cloud-native platform with one data model, the approach the industry now refers to as Digital Communications Governance and Archiving (DCGA), with explainable AI Surveillance running on that same foundation. DCGA and AI Surveillance are two layers of one platform: DCGA is the governed data foundation, and AI Surveillance is the explainable intelligence that runs on it.

A unified DCGA and AI Surveillance platform directly answers each survey pain point:

• One data model removes the fragmentation, inconsistent identifiers, and missing standards the survey names as the biggest hindrance to effectiveness.
• Native cross-channel surveillance across e-comms, voice, and trade makes the holistic view that 48% of firms cannot achieve today architecturally possible rather than an aspiration.
• AI-native, explainable detection cuts noise and surfaces real risk while remaining defensible to MRM teams and regulators, closing the gap between the 11% who have AI trade surveillance and the 89% who want it.
• Cloud-native scale lets coverage expand across channels, languages, and venues without scaling headcount in lockstep.
• A single archive and surveillance layer ends the duplication, migration pain, and case-management strain that come from stitching acquired components together.

The market alternatives, legacy suites assembled through acquisition or black-box AI that cannot explain itself, recreate the very fragmentation and governance risk the survey identifies.

The right move is consolidation onto a unified DCGA and AI Surveillance platform designed from the ground up as one system.

Why Shield

Shield is a unified DCGA and AI Surveillance platform: communications surveillance, archiving, governance, and investigations on one system, powered by explainable AI.

It was built from the ground up as a single, modern, cloud-native system rather than assembled from acquisitions, which is precisely the architecture the survey’s findings call for.

This is not only Shield’s own assessment. In the 2025 Gartner® Magic Quadrant™ and Critical Capabilities reports for Digital Communications Governance and Archiving (DCGA) solutions, Shield was named a Visionary, was the fastest-rising vendor in the Magic Quadrant, ranked among the top 5 of 14 vendors for Completeness of Vision, and scored as a Top 3 vendor across all Critical Capabilities use cases: Connectors, Archive & Retention, Regulatory Compliance, Investigations, Internal Analytics & Insights, and User Governance.

Notably, Shield’s highest scores came in Investigations and Connectors, outperforming legacy providers in capabilities that were historically their strongholds.

What that delivers in practice:

• Noise reduction with precision. Shield reports 3x less noise than legacy systems and an average alert rate of 0.054%, alongside 3x more actionable escalations, directly addressing the industry’s most cited pain.
• Explainable, agentic AI. Shield’s AmplifAI multiagent suite improves detection accuracy and transparency. Its GenAI identifies nuanced risk while reducing false positives and accelerating triage, and its GenAI investigative assistant enables natural-language search across investigations. The architecture is model-agnostic, so firms can integrate their own proprietary models. This gives MRM teams and regulators the transparency they now require, where black-box opacity tends to stall AI adoption. Gartner recognized Shield’s generative and agentic AI for two consecutive years.
• Native cross-channel coverage. Shield brings e-comms and voice into one investigative view, supported by a proven low Word Error Rate on voice transcription, so holistic review becomes something teams can run day to day.
• Unified, cloud-native architecture. One platform for capture, archive, and surveillance, with strong data residency and federated search, so coverage scales without bolting tools together. Flexible deployment across SaaS (AWS), with a modular design that lets firms start where they need and scale as demands evolve, including a deployment delivered in seven weeks for a major APAC sovereign wealth fund (Case Study available upon request.)
• Transparent commercial model and data portability. Shield uses per-employee pricing for predictable total cost of ownership, with no-cost data exports and no exit fees, removing the financial and operational barriers that often trap firms in legacy archives. Gartner specifically cited this commercial model and data accessibility as a strength.

A representative outcome from Shield’s own client base: a leading global financial institution with 15,000+ employees across 38 countries had spent four years failing to implement voice surveillance and could review only about 15% of sampled alerts.

After consolidating onto Shield’s unified surveillance and voice platform, it achieved integrated risk detection across 1,500 monitored employees, 16 data sources, 9 languages, and 32 risk behaviors, with broader coverage and significantly reduced noise.

At another institution, Shield delivered a 0.22% alert rate across roughly a million communications a month and 3x more escalations than the previous vendor.

The pattern is consistent: when firms replace fragmented, legacy, or black-box estates with a unified, explainable DCGA and AI Surveillance platform, the survey’s hardest problems become tractable.

About this report

This analysis summarizes and interprets the 2026 Surveillance Benchmarking Survey & Report, a 1LoD survey of surveillance leaders across financial institutions, for which Shield was one of the lead sponsors. All survey statistics are drawn from that report. Shield performance figures and the customer outcome described are Shield’s own. Analyst recognition refers to the 2025 Gartner® Magic Quadrant™ and Critical Capabilities reports for Digital Communications Governance and Archiving (DCGA) Solutions.

Gartner® is a registered trademark and service mark, and MAGIC QUADRANT is a registered trademark, of Gartner, Inc. and/or its affiliates in the U.S. and internationally and are used herein with permission. All rights reserved. Gartner does not endorse any vendor, product or service depicted in its research publications and does not advise technology users to select only those vendors with the highest ratings or other designation. Gartner research publications consist of the opinions of Gartner’s research organization and should not be construed as statements of fact. Gartner disclaims all warranties, expressed or implied, with respect to this research, including any warranties of merchantability or fitness for a particular purpose.