
Good AI won't fix a bad data problem: Why compliance needs to get its records straight first
Opening the prospect of much faster reviews, automated monitoring, better insight into risk, and the ability to interrogate vast volumes of records in seconds, Artificial Intelligence promises to transform communications compliance.
But there is an absolute first stage to get there that any serious enterprise IT project, compliance-related or not, cannot afford to overlook: the data must be fit for purpose first.
Yes, AI can process information at extraordinary speed, but simply by being given access to it, it cannot turn inconsistent, incomplete, or poorly governed records into reliable evidence. The UK Government's January 2026 guidance on making government datasets ready for AI, for example, says data can is only AI-ready data if it’s accurate, complete, consistent, secure, offers useful metadata, alongside governance and clear data stewardship.
In fact, the opposite can occur: by producing an apparently coherent answer from inconsistent underlying information, AI can make an existing data-quality problem harder to spot.
For compliance teams, that makes data preparation a critical part of the control environment. Let’s see how.
Not just a problem for AI…
It’s worth noting that data quality isn’t, of course, a problem that only emerges when AI enters the picture. In fact, this is a problem that has confronted compliance for many years.
When an audit takes place, the required information and records need to be found, as rapidly as possible. Any such enquiry, therefore, depends on an organised compliance repository: the need for good records is not new.
But, the application of AI accentuates this problem, because AI depends on data availability – and, if we wish to apply AI to deliver new insights and benefits, we need also to get the data right. So…
AI amplifies the quality of the data we serve it
The fact is that data analytics, Business Intelligence, and all serious pre-AI enterprise IT applications and automated decision-making all depended on good underlying data.
As McKinsey says, not having that is the primary obstacle to enabling AI for more than two-thirds of high-performing companies, with organisations needing data that is reliable, traceable and reusable.
AI, then, simply increases both the potential value of good data, but also the potential negative consequences of reliance on poor data. Analysts agree. Gartner predicts that through 2026, organizations will abandon 60% of AI projects that aren’t supported by AI-ready data.
AI can genuinely speed up responses to audits and regulatory enquiries, but only if the data foundations are in place. Good data management should already exist, but in practice it may not provide the foundation AI needs.
The compliance record problem
Before asking what AI can do with compliance records, a key question is: can you confidently explain what those records contain? An organisation in a regulated sector like financial services or which has a contact centre may have years of recorded customer conversations, but that does not necessarily mean that it has a usable compliance dataset for possible AI extension.
Why: recordings may be stored in different systems or formats, your metadata may be incomplete, customer identifiers may be inconsistent and retention rules may have changed over time. Moreover, different teams may use different terminology for the same issue.
Through context and judgement, a human investigator can often compensate for these inconsistencies. Can we really ask a new AI system working across thousands or millions of records to do as good a job?
From ‘just keeping records’ to a fully integrated compliance process
This is why normalisation matters. The objective is not to make every record identical. It is to establish a consistent structure and vocabulary around the information: what happened, when it happened, which customer or interaction it relates to, what category it belongs to, what version of a policy applied at the time, and where the underlying evidence came from.
That creates something much more valuable than a pile of recordings: a consistent, searchable and governable compliance framework.
The mission, therefore, is bigger than deploying an AI tool. You must be sure you can give it what it needs in the right context. A properly structured set of compliance records can become a foundational asset, as once information is consistently classified and its provenance understood, the same foundation can support multiple applications: QA, regulatory monitoring, complaint analysis, thematic reviews, training and management reporting, and much more.
It’s important to remember that regulation such as the EU’s AI Act and other national legislation keeps raising the bar on data governance and traceability.
Bottom line: before trying to make them intelligent, you need to make your records trustworthy.
What trustworthy records look like in practice
This is where the recording layer matters. Touch captures communications from more than 60 systems, covering voice, mobile, SMS and messaging platforms. Each interaction is recorded with consistent structure and metadata, regardless of the channel.
An AI working on top of that recording layer reads one coherent record set, not 60 inconsistent ones. The same consistency makes a regulator’s request, or a GDPR subject access request, far quicker to answer.
Consistency is half the problem. The other half is control over who and what can see the data. Touch’s role-based access controls apply to AI exactly as they do to people.
Whether it’s our own AI features or a customer’s own tools connected through governed access: an AI sees only the records its user is permitted to see. And every access is written to an audit log, whether by a person or an AI, so compliance teams can see who looked at what, when, and why. They can answer “how is AI being used on our records?” with evidence and confidence.
That’s the point of building AI on the recording layer rather than beside it: the same standards that make the records defensible make the AI’s use of them defensible too.
How Touch is helping customers like you start to unlock the value of AI
As AI is rapidly becoming a strategic capability for businesses, Touch is investing to ensure our customers can capture its benefits, with the addition of secure, trusted AI capabilities to provide value from the compliance records we protect.
We’re working closely with a select group of strategic partners and early adopters, who are already using our new AI-enabled capabilities on their recorded interactions: intelligent transcription, search, and conversational queries that let authorised users interrogate large volumes of recordings quickly and effectively.
Their real-world experience is shaping how these features develop, and we expect to move them to general availability in the coming months.
These innovations will turn recordings from a passive record into a richer source of insight, supporting faster decision-making, better customer outcomes and more effective compliance. We will also ensure that our technology can work alongside customers’ own AI and analytics environments, providing secure, governed access to the data they need as their strategies evolve.
Innovation with trust as DNA
The Touch AI approach is underpinned by a simple principle: AI can only ever be deployed within the same rigorous governance framework as the data on which it operates. AI-derived information will therefore remain subject to appropriate access controls, retention and deletion policies, audit trails and customer-specific compliance requirements.
To be clear, that’s the same principle that has underpinned our efforts to deliver consistent, coherent, and accurate compliance recording frameworks.
We are also working to make sure that customers using Touch’s AI capabilities will be able to configure AI-enabled monitoring and surveillance around their own Codes of Conduct, regulatory obligations and industry best practice, with human oversight maintained where it matters.
By bringing AI into the established quality and compliance environment, Touch can help customers extract substantially more value from their recorded interactions while maintaining the security, privacy and governance on which their businesses depend.
But that journey must start with good data. If building a trustworthy foundation for AI is on your agenda, get in touch to see how Touch can help.

