^{CHAPTER 05}
Your data foundation – the work that comes first

^{05. YOUR DATA FOUNDATION – THE WORK THAT COMES FIRST}

AI built on fragmented, inconsistent data does not underperform. It misleads. That distinction matters. Underperformance is a problem. Confident outputs based on flawed inputs create decisions that feel grounded and are not.

Most organisations underestimate the state of their data before they start an AI initiative. The gaps are usually not dramatic. They are structural and quiet. Master data is inconsistent across systems. Cost centre definitions do not align between ERP and the planning tool. Entity hierarchies changed two years ago and were never fully reconciled. Data arrives from source systems with a 48-hour lag. These are not edge cases. They are the default state of mid-market finance data.

Five warning signs that your data foundation is not AI-ready:

1.

Planning data and ERP data require manual reconciliation to agree

2.

Master data (cost centres, entities, products) differs between systems

3.

Analysts spend time correcting actuals before they can use them

4.

Forecast inputs arrive from business units in different formats or templates

5.

There is no clear owner for data quality in the finance technology stack

None of these disqualify an organisation from pursuing AI. They are the work that needs to happen first. Master data governance, lineage tracking, reducing data latency between ERP and FP&A, and standardising input processes are not glamorous projects. They are the foundation that determines whether AI creates value or accelerates confusion at scale.

The sequence matters. Data readiness is not a parallel workstream to AI adoption. It is the prerequisite. Organisations that skip this step and move straight to AI tooling will produce outputs that cannot be trusted, workflows that require manual correction, and a team that reverts to spreadsheets because the AI-generated numbers are unreliable. The technical debt accrued by skipping foundational work does not reduce over time. It compounds.

Start with a data readiness assessment before any AI project begins. Prioritise reducing latency between source systems and the planning layer. Establish clear ownership for master data. These investments pay back directly in the reliability of everything that follows.

^{05. YOUR DATA FOUNDATION – THE WORK THAT COMES FIRST}

Key takeaway

Data foundation work is not a prerequisite that delays AI adoption. It is the difference between AI that creates real value and AI that produces unreliable outputs at speed.

Action

Map your current data flow from ERP to FP&A. Identify the three biggest points of latency, inconsistency, or manual intervention. Treat those as your first AI readiness projects.

94% cite data security and privacy as a critical to moderate priority when considering AI.

Dresner Research Insight – The Office of Finance in 2026