Data at Scale in the AI Era: What Leaders Are Learning the Hard Way (and What to Do About It)

25 May 2026 ·Blog ·DiGiCOR

“Data is exploding” is no longer a headline — it’s the operating environment. What’s changed isn’t just the volume. It’s the pressure: retention obligations, security risk, power limits, procurement volatility, and AI workloads all colliding at once.

In a recent DiGiCOR × Seagate executive roundtable in Melbourne, leaders across multiple sectors came together for an off-the-record, peer-led discussion under the Chatham House Rule. The format was deliberate: no presentations, no pitching — just practical, unfiltered perspectives on what’s breaking first and what’s actually working.

The uncomfortable question: is your data working for you... or against you?

One theme came up repeatedly: most organisations still struggle to separate data that creates value from data that creates liability. That matters, because the “keep everything” default isn’t just expensive now — it can amplify risk.

Non-attributable insight:
“If you’ve got it, it’s legally discoverable — so you want to delete it as soon as you legally can.”

Across industries, the discussion kept returning to the same reality: retention is strategy, not housekeeping. Whether you’re dealing with research datasets, customer interactions, media archives, or log telemetry, you need clarity on what you keep, why you keep it, where it lives, and what it costs to operate.

Five practical takeaways that kept resurfacing

Retention policies define the cost curve (not raw capacity)
Storage isn’t “cheap enough to ignore” anymore — especially when your retention posture silently expands over time. The cost isn’t only hardware. It’s power, floor space, backup/replication, security exposure, and operational overhead.
AI accelerates both value and “dark data”
AI doesn’t just create more data — it creates messier data: richer logging, more intermediate artefacts, more versions, and more “maybe useful later” retention. The result is often a growing pool of data that is available but not necessarily usable.
Metadata and classification are the only sustainable exit
When teams can’t confidently answer “who owns this data?” or “what is this dataset for?”, deletion becomes politically risky and operationally slow — so data persists by default. The consistent message: if you want to scale sustainably, you must invest in metadata discipline and clear accountability.
Real-world constraints are now the architecture
This wasn’t theoretical — it was grounded in constraints people are dealing with right now: pricing volatility, approvals lag, long lead times, and power limits becoming as important as performance specs. Even great technical planning can be derailed by procurement timelines and budget cycles.
Simplification is a competitive advantage
Many teams are carrying complexity they didn’t choose — too many platforms, too many copies, too many tiers that no one fully governs. The takeaway: consolidation and operational simplicity are no longer “nice to have”; they’re how you create headroom without multiplying risk.

A simple framework to take back to your organisation

Here’s a practical checklist that emerged from the conversation — useful regardless of industry:

Define retention as a business decision: what must be kept, what should be kept, and what should be deleted when legally allowed.
Make ownership explicit: assign data custodians and decision rights (not just storage admins).
Standardise metadata early: classification, purpose, retention category, sensitivity, and lifecycle.
Design for sustainability: cost, power, space, and operational effort — not only peak throughput.
Reduce platforms where possible: fewer moving parts means fewer surprises and easier governance.

Closing thought: every decision shapes the next one

The strongest closing message was this: infrastructure decisions made today become the constraints — or the capability — of your next two years. If you want AI readiness to be real (not marketing), you need retention clarity, metadata discipline, and scalable operations that won’t collapse under growth.

Want to join a future roundtable or continue the conversation? Reach out to DiGiCOR — we’re keeping these sessions intentionally small, practical, and peer-led.