Know which data your AI should use, and which it should not
ETT helps you locate relevant data, identify sensitive content and establish ownership and handling rules before it enters an AI workflow.
Who this service is for
CISOs and security leaders
Understand where sensitive data lives and how it could be exposed through AI systems.
Data leaders
Build clean, structured and reliable data foundations for AI, analytics and automation.
IT leaders
Manage fragmented systems, legacy platforms, cloud apps and data estates that have grown over time.
Compliance teams
Prepare for AI governance, audits, privacy reviews or regulated data use.
AI and machine learning teams
Get reliable training, grounding or retrieval sources before AI systems can perform effectively.
What you would be buying
A picture of what data exists across an agreed part of your estate, how sensitive it is, who owns it, and which sources are safe for an AI workflow to use.
What you receive
- Scoped data inventory across the repositories agreed at the start
- Sensitivity labels applied against your classification policy
- Ownership mapping for the data found
- Exposure findings: over-permissioned, duplicated or misplaced content
- Prioritized remediation plan
- Approved source shortlist for AI use
What you provide
- The repositories in scope, agreed in writing
- Read access, or a supervised scan window
- Any existing classification policy
- Responsible data owners for the areas covered
How success is measured
- Coverage: how much of the agreed estate was actually reached
- Classification quality, scored on a human-reviewed sample
- Exposure identified
- Remediation progress against the prioritized plan
Scanning, classification, document extraction and answer retrieval are four different things. This service does the first two. It does not build the extraction or retrieval layer, and discovery tooling is not a document extraction engine.
What can hold it up
- Scope agreed before work starts; an unscoped discovery has no completion point
- Access to repositories, including any that require a change request
- An owner for the remediation plan — findings without an owner do not get fixed
After it goes live
Discovery is a point-in-time picture. Estates drift, so agree a re-scan cadence if the inventory needs to stay current.
Deliverables and measures describe the standard shape of this engagement. Exact scope, duration and commercial terms are agreed and confirmed in writing before work starts.
How we do the work
Data discovery
We scan and map data across cloud apps, data lakes, legacy systems, shared repositories and operational platforms, creating a clearer view of where data lives, how it is used and what may be hidden.
Structured and unstructured data mapping
We identify and organize both structured data, such as databases and system records, and unstructured data, such as documents, emails, transcripts, call logs and knowledge files.
Data classification
We classify data by sensitivity, compliance category, business function and potential AI use case, helping organizations understand what can be used safely and what needs additional controls.
Sensitive data identification
We help identify data such as PII, PHI, customer information, financial records and sensitive business data before it enters AI workflows.
Data structuring and metadata enrichment
We support the deduplication, normalisation and enrichment of data with the context and metadata needed for AI systems to use it more effectively.
AI readiness preparation
We help prepare data for use in AI workflows, including grounding, retrieval-augmented generation, knowledge repositories, vector databases and semantic relationships where appropriate.
Why this matters
Most enterprises want to use AI more effectively, but many are starting from a difficult position.
Their data is spread across cloud platforms, legacy systems, data lakes, shared drives, spreadsheets, emails, documents, transcripts and custom tools. Some of it is structured. Much of it is not. Some of it is sensitive. Some of it is duplicated, outdated or poorly labelled.
Without a clear view of what data exists and how it should be handled, AI becomes harder to trust.
ETT helps organizations create that visibility. We discover where data lives, classify it by sensitivity and business function, and prepare it so it can be used more safely and effectively inside AI systems.
Why discovery and classification come before AI deployment
AI systems need reliable data, but reliability starts with knowing what data the business actually holds.
If sensitive information is hidden in unstructured files, customer records are inconsistent, internal knowledge is scattered or data ownership is unclear, AI systems can introduce risk quickly. They may surface the wrong information, expose data that should be protected, or operate from sources that are incomplete or poorly understood.
Discovery and classification create the control layer that AI needs. They help organizations identify what data can be used, what needs protecting, what requires cleaning and what should not enter an AI workflow at all.
Hidden sensitive data, unclear ownership, inconsistent records, unstructured sources, higher AI risk.
Known data estate, clear labels, sensitivity mapping, AI-ready sources, stronger governance.
How we prepare enterprise data for AI
Orient
We identify where data lives, how it is currently used, what systems are involved and where hidden or sensitive data may create risk.
Prove
We prove the approach on a defined data set, validating that discovery and classification surface the right risks and AI-ready sources.
Govern
We define the classification approach, data-handling rules, metadata structure and governance needed for safe AI use.
Scale
We apply discovery, classification and structuring across the agreed data sources, creating clearer visibility and more usable AI foundations.
Compound
We maintain data visibility over time, review classifications, update governance rules and improve readiness as AI use cases evolve.
Often delivered alongside
Timing, cost, access and ownership
How long does this take?
It is set by the scope agreed at the start and by how long access approvals take. We will not run an unscoped discovery: without a boundary there is no point at which it is finished.
What drives the cost?
The size and variety of the estate in scope, how many repository types are involved, and how much of the classification needs human review to be trustworthy.
What access do you need?
Read access to the repositories in scope, or a supervised scan window where read access is not permitted, plus any existing classification policy and the responsible data owners.
Do you take our data off-site?
No. Discovery and classification run against your estate under your controls. What leaves is the findings, not the content.
What happens next?
A conversation about the specific process or decision you have in mind. If there is a workable opportunity we will describe a scoped first engagement; if there is not, we will say so and explain why. Requesting a session does not commit you to anything.
About this service
What is data discovery for AI?
The process of locating and mapping the data an organization holds, including structured and unstructured sources, so it can be assessed for AI readiness, governance and safe use.
What is AI data classification?
Labelling data by sensitivity, compliance category, business function and potential AI use case. This helps organizations decide what data can be used, protected, restricted or excluded from AI workflows.
Why is data classification important before using AI?
AI systems may access, retrieve or act on business data. Classification helps prevent sensitive, inaccurate or inappropriate data from being used in ways that create risk.
What types of data should be discovered and classified?
Databases, documents, emails, transcripts, call logs, spreadsheets, customer records, internal knowledge bases, cloud app data and legacy system data.
How does data discovery support RAG and AI grounding?
Discovery helps identify which sources can be used for retrieval-augmented generation and grounding. Classification and structuring then make those sources more reliable, traceable and suitable for AI use.
Do you know what data your AI will rely on?
Request an AI session to explore whether your data estate is ready for AI, where hidden risks may exist, and what needs to be discovered, classified or structured before deployment.
Around four minutes. Indicative guidance based on your answers.