Loud and clear – Why accurate voice transcription has multiple compliance benefits

• The accuracy of voice transcription is improving, and the technology is unlocking multiple compliance benefits
• From identifying market abuse to non-financial misconduct, transcription is becoming a vital tool for monitoring communications
• With regulators’ expectations around recordkeeping increasing, transcription gives firms a more complete and more clear data picture

Advances in the accuracy and dependability of voice transcription as a risk management tool have come at a critical time for those working in regulated industries. With hybrid work resulting in trading and business taking place over an ever-increasing range of voice channels, transcription is becoming an increasingly vital tool in ensuring that firms are able to enact effective communications monitoring and surveillance – and to keep records that are regulator-ready should the investigators in jurisdictions where voice monitoring is a requirement to come knocking.

As a channel, voice has yet to be specifically highlighted by regulators as a key focus in the same way as channels like WhatsApp (although the Australian securities regulator recently listed it alongside other channels firms must supervise and capture). In a recent LSEG webinar, Rob Mason, Director, Regulatory Intelligence at Global Relay, identified that current attitudes suggest that transcription records may not need to be retained unless a compliance alert has been flagged, but that “voice orders are still orders, and are required to be captured”. 

Whether it is to meet regional recordkeeping requirements or monitor for misconduct, there are a number of compliance benefits that capturing accurate voice transcription offers firms. 

One of the clearest applications of high-quality transcription is using these records to flag potential instances of market abuse. The fast-paced nature of financial services decision making necessitates the use of voice over other communication channels.

There is the belief that voice channels can be harder to surveil for signs of market abuse. We’ve seen examples where traders have used channels like Xbox 360 voice chat to discuss abusive practices because those channels “can’t be traced”. Previously, there was perhaps some truth to this – voice channels produce hundreds of hours of recordings that would have had to be listened to and transcribed manually. Undertaking this can be a cost prohibitive exercise when few records yielded any signs of wrongdoing.

It’s no wonder those looking to disguise misconduct may have turned to voice as a means of covering their tracks, as it left no physical paper trail. However, the advent of more accurate transcription and AI tools means that conversations can be reviewed automatically, in context, and at speed and scale. Policies can be set that flag any potential signs of market abuse – meaning those looking to manipulate markets will need to think again. 

Market abuse and insider trading aren’t the only kinds of misconduct that perpetrators might seek to hide via voice channels. We have recently seen a sea change, with regulators increasingly prioritizing non-financial misconduct and expecting firms to be actively monitoring for bullying, harassment, and discrimination. 

According to Rob Mason, firms are reporting more internal escalations and disciplinary measures stemming from non-financial misconduct identified via voice channels than instances of market manipulation. Accurate transcription plays a key role in this, producing a precise record of any instances that can be reviewed, and text-format data that allows for compliance functions to set up policies that flag keywords (such as racist, sexist, or other derogatory language) automatically.

In the same LSEG webinar, Carroll Barry-Walsh, Lawyer, Speaker, and Founder of Barry-Walsh Associates, highlights that it can be challenging to discern if behaviour constitutes non-financial misconduct without context. Accurate transcription gives us statements and their context, which makes investigation much easier and enables reviewers to establish if a comment was an isolated, offhand remark that was misinterpreted – or part of a more concerning behaviour pattern – and escalate accordingly. 

Transcription can also assist in identifying whether reviewers might need to look elsewhere for signs of misconduct, such as off-channel communications. By setting up a policy alert or AI model to flag phrases like ‘let’s discuss elsewhere’ or ‘message me on WhatsApp,’ firms can identify where speakers might be trying to hide something – and follow them across channels accordingly.

During the webinar, Rob Mason stated that “recordkeeping has never been more important than it is now.” We see a consistent drumbeat of enforcement cases against firms and individuals that have failed to meet regulatory recordkeeping requirements. We’ve also seen regulators urge firms to proactively cooperate with investigations and self-report to secure better investigation outcomes. 

Cooperation with investigations requires complete, coherent, easily accessible data – something accurate transcription plays a huge part in. Producing complete, clear records that can be sorted and searched in a compliant archive, advanced transcription can assist firms in quickly and accurately identifying necessary information. Knowing who said what, when, to who – and being able to evidence it – goes a long way to helping firms in relevant jurisdictions to respond to regulatory requests more quickly. This starts relations with regulators off on the right foot, collaboratively, confidently, and cooperatively. 

Firms that are unable to provide accurate transcription records have gaps in their compliance systems and can be much slower to provide information to regulators – risking seeming obstructive, rather than cooperative. Regulators take a dim view of firms that are unable to provide records when asked, as it means they cannot ‘do their job’ of investigating potential non-compliance.

There are many other clear benefits to accurate transcription, such as its ability to free up employees from lengthy manual transcription efforts to focus on higher-value compliance tasks, the related cost and efficiency savings, and providing firms with another stream of high-value data to unlock data insights and train AI models. Solutions are becoming increasingly advanced, able to accurately transcribe in multiple languages and overcome challenges like background noise, poor audio quality, colloquialisms, and slang. Firms operating in geographies where monitoring voice is required are beginning to catch on to the fact that, when it comes to voice transcription for compliance, the benefits are loud and clear.