Somewhere in the long human habit of answering every problem by adding more, the enterprise data world has reached a turning point that its own builders are now naming. Dan Peacock of CryspIQ, writing for an Australian technology audience, argues that the rush to put artificial intelligence to work is revealing what all those extra copies, pipelines and stores truly cost. His vision of 2027 is one of restraint: fewer copies, one agreed meaning for each business term, and data foundations that can be trusted enough for machines to act upon. It is a forecast from someone with a stake in the outc
2027 Predictions: Enterprise Data Efficiency Emerges as AI Readiness Exposes Rising Data Costs
Twenty years finding cheaper places to keep more copies of our data
Mark
Why does this prediction matter beyond one vendor's pitch?
Mimi
Because it reframes data spending. Instead of asking where to store more, boards would ask what it costs to keep copying, securing and governing data nobody uses.
Mark
What's the headline number?
Mimi
Peacock expects compute costs to rise 100 percent within five years if AI growth goes unmanaged, especially AI with federated access across many systems.
Luke
Worth flagging that's his expectation, not a measured result. No methodology is given in what we have.
Mark
And the link to security?
Mimi
Every duplicate of customer data has to be secured and monitored. Fewer copies means less to protect and less to pay for.
Luke
Plausible logic, but it's an argument, not evidence. We don't see any breach or cost data.
Mark
What about AI agents?
Mimi
He says most still summarise, draft and retrieve. Few are trusted to act, because nobody will sign off on actions based on data they can't stand behind.
Luke
That's his diagnosis. Whether trust in data is the main barrier, versus model reliability, is asserted rather than shown.
Mark
Does the old infrastructure go away?
Mimi
No. He says warehouses, lakes and cloud platforms stay, with efficiency sitting beneath them.
Luke
And CryspIQ presumably sells in this category, so read it as a forecast with a stake attached.
Il Polso
- Organisations built a 'copy everything' data estate that now quietly drains budgets through storage, movement and security costs.
- Federated AI reaching into many systems at once could, by Peacock's estimate, double compute costs within five years, forcing boards to scrutinise data spend.
- Monolithic software suites are fragmenting into specialised AI apps, each carrying its own definition of the business and pulling against shared truth.
- Most AI agents still only summarise, draft and retrieve, and remain stuck short of acting because no one will sign off on data they cannot vouch for.
- The push is toward a governed, single-meaning data layer, with agents waiting on whether organisations invest in traceable foundations to let them act.
Somewhere in the long human habit of answering every problem by adding more, the enterprise data world has reached a turning point that its own builders are now naming. Dan Peacock of CryspIQ, writing for an Australian technology audience, argues that the rush to put artificial intelligence to work is revealing what all those extra copies, pipelines and stores truly cost. His vision of 2027 is one of restraint: fewer copies, one agreed meaning for each business term, and data foundations that can be trusted enough for machines to act upon. It is a forecast from someone with a stake in the outcome, and its real test lies ahead, in whether organisations will choose to own the meaning of their data before they hand it to agents.
Dan Peacock, CEO and co-founder of CryspIQ, has given a name to what he believes comes next for enterprise data: Enterprise Data Efficiency. In his 2027 predictions, published by IT Brief Australia, he argues that the scramble to put AI to work is exposing what it truly costs to copy, move and store company data securely.
Peacock calls this the end of the "copy everything" era. Each generation of data architecture solved its predecessor's problems by adding more storage, copies, pipelines and tools, and cloud platforms made storing everything so cheap that many organisations now pay to govern data nobody uses. His alternative is to use less data, better: define business meaning once, hold it in one governed copy, and let every report and AI tool read the same facts. Warehouses and lakes will remain, but efficiency will sit beneath them and reshape what they hold.
The sharpest argument concerns cost. Peacock expects AI, especially federated access reaching across systems, to double compute costs within five years, and he expects boards to start measuring data spend against its returns. He also sees security and cost as the same work, since every copy not made is one less to pay for and protect.
Finally, he predicts that AI agents will move beyond summarising and drafting only once organisations can stand behind their data. An efficient, traceable estate would give agents data they can vouch for. These are forecasts from a company selling in this space, not reported findings, and the thing to watch is whether agents begin to act.
Dan Peacock, CEO and co-founder of CryspIQ, has put a name to what he thinks the next phase of enterprise data will be: Enterprise Data Efficiency. His 2027 predictions, published by IT Brief Australia, argue that the rush to put AI to work is exposing what it really costs to copy, move and store company data securely.
Peacock frames this as the close of what he calls the "copy everything" era. In his telling, each generation of data architecture fixed the previous one's problem by adding more: more storage, more copies, more pipelines, more tools. The warehouse, the lake and the cloud data platform each followed that pattern. Cloud platforms made storing everything cheap, so organisations did, and many now pay to move, secure and govern data that most of the time nobody uses.
His alternative is to use less data, better. Business meaning would be defined once, held in one governed copy, and every report and AI tool would read the same facts. Peacock says warehouses, lakes and cloud platforms will not disappear. Efficiency would sit beneath them and change what they are asked to hold. He expects governance, quality and traceability to become central, with consequences for CFO reporting, trusted financial meaning and operational decisions.
The cost argument is the sharpest. Peacock expects AI, particularly AI with federated access reaching into many systems where they sit, to raise compute costs by 100 percent within five years. Each question can now touch many systems and many copies of the same data, and user demand keeps adding questions. Organisations that do not cut duplication, he says, face a cost explosion from unmanaged AI growth, and boards will begin measuring it as a return on data spend.
He also predicts the all-in-one software suite will break into smaller, specialised AI apps that replace traditional business functions rather than modify them. Each app brings its own working definition of the business. That makes the data layer the one thing worth owning, the place where meaning holds while apps change around it. A lake built to store everything now and work out meaning later cannot do that job, he argues, because when the consumer is an AI app there is nobody left to work it out.
Security and privacy get the same treatment. Every duplicate of customer data must be secured, monitored and accounted for, so Peacock calls fewer copies the cheapest cyber and privacy control available. Data minimisation and cost reduction, he says, turn out to be the same work: each copy not made is one less to pay for and one less to protect.
On AI agents, he observes that most enterprise agents still summarise, draft and retrieve, and very few are trusted to act. He says that is not a limit of the models; nobody will sign off on an agent acting on data they cannot stand behind. An efficient estate, with one meaning per term, quality checked as data arrives and every figure traceable to its source, would give agents data with a warranty and let them move from drafting to acting. The excerpt ends with a further prediction, that the data team's job will shift from building pipelines to owning meaning.
These are forecasts from the head of a company selling in this space, not reported findings. The thing to watch is whether agents move from drafting to acting as organisations invest in data they can vouch for.
Citazioni salienti
Fewer copies will be the cheapest cyber and privacy control available; every copy you stop making is one less to pay for and one less to protect.— Dan Peacock, CEO and co-founder, CryspIQ
Vuoi la storia completa?
Leggi l'originale su IT Brief Australia ↗