Data Readiness: The Cost Nobody Budgets For
The hidden dependency that quietly determines whether AI delivers or drifts
AI does not fix weak data discipline. It amplifies it. Every AI programme I have seen struggle can trace at least one of its problems back to the data estate it was deployed against—not the model, not the platform, not the vendor.
Fragmented repositories. Permissions that drifted years ago and were never reviewed. Records with no owner, no classification, and no retention rule. Files duplicated across platforms with contradictory versions. In a regulated business, these are not hygiene issues to address when convenient. They are risk events waiting to be triggered the moment you introduce a tool that makes everything faster and more discoverable.
Why This Is a Board Issue, Not an IT Issue
The Jersey AI Playbook is explicit: AI effectiveness depends on high-quality, governed data foundations. Fragmented systems, weak metadata, and unclear data ownership do not just undermine usefulness—they create risk. This is not an IT observation. It is a strategic dependency that belongs on the board agenda.
PwC's 2025 Channel Islands Wealth Management survey reinforces the commercial dimension: targeted technology investment can reduce breakeven mandate sizes and open otherwise uneconomic client segments. But that investment only delivers if the data it depends on is fit for purpose. AI applied to a disordered data estate does not improve outcomes—it makes disorder faster and more consequential.
The JOIC has been clear in its 2025–2026 strategic priorities: AI systems relying on large-scale data processing must address lawfulness, data minimisation, purpose limitation, and retention from the outset. These are regulatory obligations that your data estate must be able to demonstrate—on demand, with evidence.
The Three Data Risks That Quietly Kill AI Value
Oversharing risk
Collaboration platforms drift. Teams create workspaces quickly, access expands 'temporarily,' and old content stays discoverable indefinitely. When an AI assistant increases the discoverability of that content, oversharing becomes an active risk rather than a theoretical one. The AI is not creating the problem—it is making a pre-existing problem visible and exploitable.
Lifecycle risk
Retention schedules that exist only as policy documents do not protect you. Regulated businesses need lifecycle management that is operationalised—records retained on schedule, disposed with evidence, and demonstrably aligned to purpose limitation principles. Without it, you cannot demonstrate data minimisation to a regulator.
Ownership risk
If no one owns the data domain, no one owns the control, the quality, or the exception handling. In AI-enabled workflows, this means no one is accountable for the outcomes the AI supports—which is precisely what regulators and auditors will want to establish when something goes wrong.
A Practical Example: When Data Gaps Drive Cycle Time
A detailed process analysis conducted within a Channel Islands financial services context—anonymised here—reviewed twenty client onboarding cases across ten sub-processes. The findings were instructive beyond the firm itself.
The top three sub-processes (document receipt and review, initial enquiry handling, and application documentation) accounted for approximately 69% of total internal processing time. Across all twenty cases, the root causes were consistent: incomplete data submitted at the point of client engagement, manual validation steps that created rework loops, and fragmented ownership of handoffs between teams with no single accountable owner for the end-to-end journey.
The firm did not have a technology problem. It had a data completeness problem and a process ownership problem. Introducing AI into that environment without remediation would have accelerated the rework loops, not eliminated them. This is what 'data readiness as a prerequisite' means in practice: establish input discipline and ownership clarity before you deploy the capability.
Minimum Viable Data Readiness
Data readiness does not mean a perfect data estate. For most Jersey FS firms, minimum viable data readiness for an AI programme means:
- A small, consistent set of classification categories that staff can apply without ambiguity
- Permissions hygiene on the repositories your first AI use case depends on
- Lifecycle management that is operationalised—not just documented—for regulated records
- Named owners for key data domains and each priority AI-enabled workflow
- A plan to maintain discipline over time, not a one-off clean-up that decays within twelve months
The Data Debt Concept
Gartner's research describes 'data debt' as one of the most commonly underestimated forms of organisational debt—and as the category that AI programmes expose most rapidly. The metaphor is useful: data debt accrues interest. Every AI initiative funded against an unprepared data estate costs more, produces less reliable output, and carries higher regulatory risk.
A practical executive reframe: treat data readiness as the interest rate on your AI investments. The higher the data debt, the more expensive every AI initiative becomes. The return on investment in data hygiene is the AI value you unlock at lower cost and lower risk because the foundations hold.
Prioritise, Don't Boil the Ocean
The most common failure mode in data readiness programmes is attempting an enterprise-wide clean-up. A practical triage approach categorises repositories into three buckets:
| Bucket | Description | Action Priority |
|---|---|---|
| High risk, high value | Used in regulated workflows, contains sensitive data, accessed frequently | Immediate—before AI deployment |
| High risk, low value | Sensitive content with unclear purpose, no owner, or expired relevance | Early—lifecycle and disposal actions |
| Low risk, mixed value | General operational content with lower sensitivity | Later—can wait |
A 30-Day Data Readiness Sprint
Week 1: Map the top three workflows to their data sources and repositories. Identify which repositories those workflows depend on and what the current access model looks like.
Week 2: Run a targeted permissions review. Establish who can access what, and whether that access remains appropriate. Remediate the highest-risk exposures immediately.
Week 3: Define classification categories simple enough to apply consistently. Assign named owners for each workflow data domain. Confirm retention rules exist and are operationalised.
Week 4: Establish a readiness dashboard for leadership—simple, measurable, updated monthly. Document the baseline state as the evidence that governance existed before deployment.
A Readiness Gate Before You Scale
Before expanding AI access to any priority workflow, verify five conditions:
- Permissions on relevant repositories have been reviewed and remediated
- Classification rules are applied and understood by the teams using the data
- Named ownership is confirmed and the owner is actively engaged
- Retention rules are operationalised, not just documented
- A monitoring and exception process exists with a named reviewer
If any of these is absent, remediate before deploying. This single gate prevents the majority of avoidable data incidents in AI programmes.
Closing question: If you enabled AI assistance across your collaboration content tomorrow, could you be confident it would not expose sensitive client or regulatory information—by accident?
Next in the series — Part 4: Training isn't prompting. In a regulated firm, AI training is behaviour change, accountability under regulation, and escalation discipline.
Sources: PwC Channel Islands Wealth Management Insights 2025; Jersey AI Playbook, March 2026; JOIC 2026 strategic priorities; Gartner research 2025–2026.
Professional disclaimer: The views expressed are my own and are shared for general informational purposes. This content is intentionally non-confidential and does not include proprietary, client-specific, or employer-sensitive information. It is not legal, regulatory, investment, or professional advice; readers should seek appropriate advice for their circumstances.
Financial services leader (20+ years) | Lean Six Sigma Green Belt | London School of Economics (LSE) AI Leadership Accelerator
See where your firm stands
The free diagnostic scores your firm across the twelve operating-model instruments in about eight minutes.
Take the diagnostic →A version of this article was first published on LinkedIn as part of the Sequence body of work. This is the canonical version.
The views expressed are the author's own and shared for general information. This content is intentionally non-confidential and does not include proprietary, client-specific or employer-sensitive information. It is not legal, regulatory, investment or professional advice. © 2026 Paul Soares. Sequence is the trading name of Sequence Advisory Limited (in formation) · hello@sequenceadvisory.com