InsightsPerspective

Technology Rationalization
Why applications, infrastructure and data have to be rationalized together

Most enterprises rationalize one layer at a time, which is why the savings in the business case so rarely reach the invoice.

One retirement, three layers
APPLICATIONSDATAINFRASTRUCTUREretired

The application is retired. Its data and its servers are still running.The same decision, carried through. The data is archived and the servers are off.

Illustrative

The application was retired in the spring. The project closed on schedule, the announcement went out, and the business case recorded the saving.

Eighteen months later its database is still running, and it is still backed up every night. Four interfaces continue to deliver files to it. A storage volume holds eleven years of customer records that nobody has classified. Two service accounts with broad privileges still exist, and the vendor contract renewed automatically in the autumn. The application has left the portfolio report. Most of its cost, and all of its risk, remain in the estate.

The scenario is a composite, but few technology leaders will find it unfamiliar. It is what happens when an enterprise rationalizes its applications without also rationalizing the infrastructure and data underneath them. It goes a long way toward explaining why rationalization programs have a reputation for promising more than they deliver.

Figure 1One retirement, ten things to close

Application

  • The applicationRetiredRetired
  • Four inbound interfacesStill delivering filesRemoved
  • Two service accountsStill activeRevoked

Data

  • Production databaseStill runningDecommissioned
  • Eleven years of customer recordsNever classifiedArchived under a retention rule
  • Nightly backupsStill runningEnded

Infrastructure

  • Three virtual serversStill runningSwitched off
  • Storage volumeStill allocatedReleased
  • Vendor contractRenewed automaticallyEnded at term
  • Patching and monitoringStill requiredStopped

1 of 10 closed. The saving is in the business case.10 of 10 closed. The saving is on the invoice.

Illustrative

Three programs, one estate

Most large enterprises run three rationalization efforts, usually under three different leaders.

Application rationalization

Deciding which applications to keep, consolidate, modernize or retire.

Usually led byEnterprise architecture or the portfolio office

Infrastructure rationalization

Reducing the servers, storage, networks, platforms and cloud services the organization pays to run.

Usually led byInfrastructure, cloud and FinOps teams

Data rationalization

Removing duplicate, outdated and unowned data, and consolidating the databases and platforms that hold it.

Usually led byThe data office, where one exists

Each of these is sensible on its own terms. The difficulty is that all three describe the same estate. An application is a claim on infrastructure, on data, on licenses and on people’s time. A decision about the application is therefore a decision about everything it runs on and everything it holds, whether or not anyone intended it that way. If nobody carries that decision through the other layers, it has been announced but it has not been carried out.

Technology rationalization is the practice of treating the three as one discipline. Every decision about an application, a platform or a data set is followed through each layer it touches, until the cost and the exposure have actually left the estate.

What each program leaves behind

Run separately, each program leaves a characteristic residue.

Application rationalization leaves infrastructure and data. The portfolio shrinks on paper while servers, databases, backups and contracts carry on as before. Savings are counted when the decision is made, and nobody checks them against an invoice.

Infrastructure rationalization leaves the question of need. Rightsizing and commitment discounts make a workload cheaper to run. They cannot tell you that the application on it has forty users and a replacement already in production. Flexera’s 2026 State of the Cloud Report estimates wasted cloud spend at 29%, the first increase in five years.

An optimization team can spend a year lowering the cost of systems that should no longer exist.

Data rationalization leaves almost everything, because it is the hardest of the three to do alone. Nobody will delete a data set without knowing which applications, reports and retention obligations depend on it, and that knowledge sits in the application layer. So the data stays. Storage grows, the same customer exists in six systems, and the AI program inherits all of it. Gartner has predicted that through 2026, organizations will abandon 60% of AI projects that are not supported by AI-ready data.

Why it has become urgent

Three pressures have moved this from an efficiency topic to a board topic.

Funding. Boards expect investment in AI, and in most enterprises the money has to come out of the run budget. Rationalization is one of the few ways a CIO can release run budget without cutting service.

Growth that nobody approved. Gartner expects that by 2027, 75% of employees will acquire, modify or create technology outside IT’s visibility. Zylo’s 2026 SaaS Management Index found that large enterprises add 21 applications a month on average, and that the average organization leaves 36% of its SaaS licenses unused. An estate that grows at that rate cannot be kept in order by a study every three years.

Exposure. Every system that should have been retired still has to be patched, monitored and defended. The implementation examples for NIST’s Cybersecurity Framework 2.0 tell organizations to “periodically identify redundant systems, hardware, software, and services that unnecessarily increase the organization’s attack surface.” It is a security instruction, and it covers all three layers in a single sentence.

Rationalize by decision, not by layer

The remedy does not require a reorganization. It requires changing the unit of work from the layer to the decision. Five practices do most of it.

  1. Give every decision a ledger.For each application to be retired or consolidated, list what it runs on and what it holds: servers, databases, storage, backups, interfaces, accounts, licenses, contracts and data sets. The decision is complete when every line is closed.
  2. Take the evidence from the estate.Surveys record what owners believe. Code, logs, configuration and billing records show what is running, who uses it and what it costs.
  3. Decide the data before switching anything off.Each data set is migrated, archived under a retention rule, or deleted. This is the step most programs postpone, and the one that most often halts a retirement in its final week.
  4. Count a saving when the invoice stops.A saving is real once a contract has ended or a resource has been switched off, and finance has confirmed it. Until then it is a forecast.
  5. Keep it running.New purchases, renewals and acquisitions rebuild sprawl every month. Rationalization has to work at the same pace, as an operating routine.
Figure 2When a saving counts
  1. The decision is madeWhere most programs count the saving
  2. Data is migrated, archived or deleted
  3. Interfaces and access are removed
  4. Infrastructure is switched off
  5. The contract ends
  6. Finance confirms the invoice has stoppedWhere the saving is real

What HFS Research found

HFS Research has named Axiarete AI a Services-as-Software™ Hot Tech. Its analysts, Joseph Montrosse and David Cushman, start their report from the same observation: “many rationalization studies are obsolete before they even get to a readout.”

HFS Hot Tech badge awarded to Axiarete AI
“Axiarete AI makes application rationalization an AI-driven capability instead of a perpetual consulting exercise.”HFS Services-as-Software™ Hot Tech: Axiarete AI, HFS Research

HFS gives two reasons for the designation. The first is that Axiarete builds its understanding of an estate from the enterprise’s own code, logs and configuration. The second is AxiareteForge, the execution arm that carries out the consolidation, modernization and remediation the platform recommends. HFS also notes that Axiarete works at the level of the whole landscape: applications, infrastructure and data.

15%+cost takeout across applications and infrastructure
20%+of the portfolio confirmed for reduction
4–5 yearswhat a manufacturing IT leader told HFS the same work would take without AI-generated recommendations

The two percentages are client-reported results from a Fortune 500 semiconductor manufacturer with about 500 manufacturing applications. All three figures are published by HFS Research.

The first figure is the one that matters for this argument. The saving was measured across applications and infrastructure together, which is what a decision produces when it is carried through the layers.

How Axiarete does it

Axiarete maintains one continuously current graph of an enterprise’s applications, data and infrastructure, assembled from the estate itself. Because the three layers sit in one model, redundancy is identified across all of them, including the duplication that one layer passes to the next.

Its agents then do the work that usually stalls. They sequence each retirement by dependency, migrate, decommission, and track the realized saving into the run-rate. Every change goes through the customer’s own pipelines, approval gates and change calendar, and a named person approves it. Where a program needs more capacity, AxiareteForge adds forward-deployed engineers who work alongside the agents. Our CIO guide describes the approach in full, and a comparison of ten vendors sets it beside the alternatives.

Where to start

A new study is not required. Take the last ten applications your organization declared retired, and ask five questions of each.

The ten-retirement test

  1. Is the infrastructure switched off, or only the application?
  2. Where is the data now, and who decided what happened to it?
  3. Do any of its interfaces, service accounts or keys still exist?
  4. Has the contract ended, and on what date?
  5. Has finance confirmed the saving?

The answers take a few days to collect. They show how much of your rationalization program has reached the invoice, and they often identify the next quarter’s savings without any further analysis.

Common questions about technology rationalization

What is technology rationalization?

Technology rationalization is the discipline of reducing and simplifying an enterprise’s technology estate across applications, infrastructure and data together. Each decision to retire, consolidate or modernize is carried through every layer it touches, so that the cost and the risk are removed along with the application.

What is application rationalization?

Application rationalization is the process of assessing every application in the portfolio and deciding, on evidence, which to keep, consolidate, modernize or retire. Our CIO guide to application rationalization covers it in depth.

What is infrastructure rationalization?

Infrastructure rationalization reduces the servers, storage, networks, platforms and cloud services an organization runs. It removes what is unused, consolidates what is duplicated and resizes what remains. It delivers most when it follows application decisions, because the largest infrastructure saving is a workload that no longer needs to exist.

What is data rationalization?

Data rationalization removes duplicate, outdated and unowned data, and consolidates the databases and platforms that hold it. Each data set is kept, migrated, archived under a retention rule or deleted. It depends on knowing which applications and reports use the data, which is why it rarely succeeds as a separate program.

Which should come first: applications, infrastructure or data?

Start with the application decision, because applications determine which infrastructure and data are needed. Then finish the data and infrastructure work for that same decision before starting the next one. HFS Research makes a related point about sequencing: rationalization comes before modernization “because there’s no sense in modernizing what won’t exist soon.”

If you would like to run the ten-retirement test against your own estate, we will run it with you, using evidence from your systems.

Request a Private Briefing