Connecting the Invisible Dots: Data Before AI

Ask most organisations where their data lives and you will get a confident answer.

Ask them to prove it, and the picture changes quickly.

A CRM that has not been properly maintained in two years. Financial records split across three versions of the same spreadsheet. Operational data that exists in a system nobody quite understands anymore. Customer records that appear under slightly different names depending on which platform you are looking at.

This is not unusual. It is the normal state of data inside a growing organisation. And it is the single most reliable predictor of whether an AI initiative will deliver value or quietly stall.

Why AI Exposes What Was Already There

AI does not create data problems. It reveals them.

When we start an AI Opportunity Discovery with an organisation, one of the first things we assess is data readiness. Not because it is a box to tick, but because it determines almost everything else. Which use cases are viable. What will need to be addressed before anything can be built. How long it will realistically take to get from exploration to something that works in production.

What we consistently find is that the data challenge is not about volume. Most organisations have plenty of data. The challenge is that it is fragmented, inconsistently structured, and often without clear ownership.

An AI system drawing on five different sources with conflicting records will not resolve those conflicts. It will reflect them at speed. The outputs become unreliable. The initiative loses credibility. And the organisation concludes that AI did not work, when the real issue was that the data was not ready for it.

The Questions That Need Honest Answers

Before any AI initiative is scoped, there are a small number of data questions worth sitting with properly.

Do we actually know where our data lives? Not where it is supposed to live. Where it genuinely is, including the spreadsheets, the shared drives, the tools teams have adopted without IT's knowledge.

Is it good enough to be useful? Data quality problems come in many forms. Inaccuracies that crept in during manual entry. Fields that mean different things in different parts of the business. Nuance that was never captured because the system was not designed for it. Gaps where information simply was not recorded. AI will not identify and correct these issues. It will work with what it is given and produce outputs that reflect the quality of what went in.

Can it be accessed and integrated? Data that exists but cannot be extracted or queried is not useful for AI purposes. Infrastructure and access matter as much as the data itself.

Who owns it? Data without clear ownership tends to be data without clear quality standards. It is a useful proxy for whether the data can be trusted.

None of these questions require a data scientist to answer. They require an honest conversation about the current state of the organisation.

Sequencing Is Everything

This is not an argument for perfection before progress.

Data quality is an ongoing discipline, not a destination. Organisations that wait for clean data before exploring AI will wait indefinitely. The argument is for sequencing: understanding the data landscape before committing to an AI initiative so that the initiative can be scoped around what actually exists.

There is also something worth understanding about how data quality improves. Data that is collected but not actively used in decision-making or embedded in business processes can retain hidden inconsistencies, ambiguities, and structural weaknesses that go unnoticed while the data sits dormant. These issues often only become visible once the data is put to work. At that point, the organisation is compelled to understand and resolve them. In other words, using data is often part of how you improve it. Waiting for perfect data before starting is not just unnecessary. It may prevent the very process that makes the data better.

Some use cases require rich, well-structured historical data. Others are viable with considerably less. The Prospera Wealth Management work is a good example. The problem was specific and well-understood: manually extracting data from multi-page provider documents was creating a bottleneck. The data existed. It was just locked in an inaccessible format. Once that was addressed, natural language processing could do its job. The result was up to ten times the processing capacity without increasing headcount.

That outcome was only possible because the data question was answered first. Not perfectly. But clearly enough to know what was viable and what needed to change.

What Readiness Actually Means

When we run an AI Readiness Assessment, data is one of five dimensions we look at. The others are technology infrastructure, capability, operational readiness, and leadership and governance. But data tends to be the one that determines pace more than any other.

The assessment does not need to take months. A structured review of the current data landscape, mapped against a specific AI objective, can be done in a matter of weeks. What it produces is clarity. Where the gaps are. What needs to be addressed. What is viable right now and what needs to be sequenced later.

That clarity, at the beginning of an AI programme, is worth considerably more than the most sophisticated model money can buy.

Where to Start

If you are exploring AI and want to understand whether your data is in a position to support it, our AI Readiness Assessment is the right starting point. It maps the current landscape across data, technology, and organisational readiness, and creates a clear picture of what needs to be in place before AI can reliably deliver value.

If you want to understand where the AI opportunity sits at the same time, our AI Opportunity Discovery works alongside that process. Together they give you a complete view of where you are, what is possible, and what needs to happen in what order.

For a more detailed picture of how AI actually delivers value inside organisations, including real examples and a practical framework for adoption, our guide Is AI the Answer? is free to download.

The invisible dots are there in most organisations. Connecting them is where the real work starts.