Loading
Skip to content
Miracle Minds AI
  • Try “HIPAA”, “model risk”, “FinOps” or “agentic”.

Application Portfolio Assessment

The value of a portfolio assessment is in the decisions it forces, not the inventory it produces. Rank the calls by what each application blocks, for AI and for risk, and give every modernize or retire call a written plan for switching off the old system, or you will pay for both.

Part of Cloud, Infrastructure & Applications, in the Think mode. 6 sources, checked .

Share of US federal IT spending (over $100bn a year) typically reported on operating and maintaining existing IT (GAO, 2025)
80%
Share of US federal IT spending (over $100bn a year) typically reported on operating and maintaining existing IT (GAO, 2025)
U.S. Government Accountability Office, 2025(opens in a new tab)
Critical federal legacy systems flagged by GAO in 2019 whose modernization was complete by February 2025
3 of 10
Critical federal legacy systems flagged by GAO in 2019 whose modernization was complete by February 2025
U.S. Government Accountability Office, 2025(opens in a new tab)
Cloud leaders naming data quality for AI model training as a top cloud AI challenge, second only to security and compliance at 53% (Flexera, 2026)
40%
Cloud leaders naming data quality for AI model training as a top cloud AI challenge, second only to security and compliance at 53% (Flexera, 2026)
Flexera, 2026(opens in a new tab)

The problem

Most IT budgets go on keeping existing systems running. The US Government Accountability Office (GAO) reports that the federal government spends more than $100 billion a year on IT and cyber investments, and that agencies typically report spending about 80% of it on operating and maintaining existing IT, including legacy systems.1 The 11 federal systems GAO found most in need of modernization were 23 to 60 years old and cost about $754 million a year to run. Seven were operating with known cybersecurity vulnerabilities.1

Modernization stalls more often than it fails outright. Of the 10 critical legacy systems GAO flagged in 2019, agencies had completed three modernizations by February 2025. Of the 11 systems in the 2025 review, three had fully documented plans, six had partial plans and two had none.1 GAO's test for a credible plan is short: milestones, a description of the work, and what happens to the legacy system afterwards. It is a fair test for any enterprise portfolio.

AI raises the stakes. Applications are where the data AI needs is created, and where AI outputs have to land. In Flexera's 2026 survey, 53% of cloud leaders named security and compliance as the top challenge for cloud-based AI initiatives, followed by data quality for AI model training (40%).2 Old applications are also a security exposure: CISA lists the use of unsupported or end-of-life software in critical infrastructure as a bad practice that significantly elevates risk.3

A useful assessment gives each application one call (keep, modernize, replace or retire) with an owner, a run cost, a risk rating and a note on what AI needs from it: data access, an API, an event feed, or nothing. Regulators already expect the inventory. NIST CSF 2.0 calls for maintained inventories of software, services and systems, and of supplier services.4 EU financial entities keep DORA registers of their ICT contracts, which supervisors used to designate critical providers in November 2025.5

What you get

  1. 01

    Application inventory with business context

    Every application with its owner, users, business capability, run cost, hosting, vendor support status and the data it creates or consumes, built from your CMDB, contracts and interviews and reconciled where they disagree.

  2. 02

    A disposition call for each application

    Keep, modernize, replace or retire, with the reasoning, the dependencies and an estimated range of cost and effort for the change.

  3. 03

    AI dependency map

    Which applications hold the data, events and APIs your priority AI use cases need, which ones must accept AI outputs, and the minimum change each one needs to support them.

  4. 04

    Risk and support exposure

    End-of-life software, unsupported hardware, known vulnerabilities, scarce skills and vendor concentration, rated so security and risk teams can act on them.

  5. 05

    Sequenced modernization plan

    For every modernize, replace or retire call: milestones, a description of the work and the disposition of the legacy system, ordered by value unblocked and risk removed.

Where it goes wrong

The failure modes we design against from the first week, and the control for each.

  • An inventory nobody keeps up

    Spreadsheets built for one assessment go stale within months. Control: build the inventory into the systems teams already maintain (CMDB, contract register, cloud tags), give each record an owner, and manage it through the application's life cycle, as NIST CSF 2.0 expects.4

  • Ranking by age instead of by what it blocks

    An old system that works and holds no data AI needs can wait; a newer one that blocks a priority use case cannot. Control: score each application on value unblocked, risk and run cost. GAO's own scoring combined age with vendor support, legacy languages, cybersecurity risk and operating cost.1

  • Modernizing without retiring

    The new platform goes live while the old one keeps running just in case, so costs rise instead of fall. Control: approve no modernization without a dated plan for the legacy system's disposition, one of the three elements GAO requires of a credible plan.1

  • Treating unsupported software as a backlog item

    End-of-life software in critical services is a risk decision, not a hygiene task, and CISA treats it as especially serious when it is reachable from the internet.3 Control: list every unsupported component and give each one a dated fix or a documented risk acceptance signed by the accountable executive.

Standards and rules

  • NIST Cybersecurity Framework 2.0, Asset Management (ID.AM)

    Primary text on nvlpubs.nist.gov(opens in a new tab)

    Inventories of software, services, systems and supplier services must be maintained and managed through their life cycles, which is the baseline a portfolio assessment should leave behind.

  • EU financial entities keep a register of their contractual arrangements for ICT services, and supervisors use it to identify critical providers, so the application inventory and the register should agree.

  • US Interagency Guidance on Third-Party Relationships: Risk Management (2023)

    Primary text on federalregister.gov(opens in a new tab)

    Banks are expected to maintain a complete inventory of third-party relationships and reassess their risk periodically, so vendor-hosted and SaaS applications belong in the portfolio.

  • CISA Bad Practices: unsupported software

    Primary text on cisa.gov(opens in a new tab)

    Using unsupported or end-of-life software for critical infrastructure and national critical functions is named a dangerous practice, especially in internet-facing technology.

How to measure it

  • Share of applications with a current owner, run cost and disposition call
  • Run cost of applications marked retire that are still live, by quarter
  • Unsupported or end-of-life components in critical services
  • Share of IT spend on running existing systems versus building new capability
  • Priority AI use cases whose blocking applications have a funded, dated plan

Questions buyers ask

How is an application portfolio assessment different from a CMDB?

A CMDB records what exists and how it is configured. A portfolio assessment adds judgment: what each application is worth to the business, what it costs, how risky it is, and what should happen to it next. The best assessments start from the CMDB, fix its gaps, and write the decisions back into it so the inventory stays useful after the assessment ends.

What does AI need from our existing applications?

Usually three things: access to the data they hold, an interface such as an API or event stream to read and write it, and a place in the workflow to act on AI outputs. Many older applications have none of these. The assessment finds which applications sit on the path of your priority use cases and the minimum change needed, which is often an integration layer rather than a rebuild.

Should we retire or modernize a decades-old core system?

It depends on what it blocks and what it costs to keep. If it is stable, supported, cheap to run and holds no data your priorities need, keeping it can be the right call. If it runs unsupported software, depends on scarce skills or blocks priority use cases, plan its replacement, and decide upfront how and when the old system will be switched off.

Sources

Every figure on this page comes from the sources below. We checked each link on .

  1. 1.Information Technology: Agencies Need to Plan for Modernizing Critical Decades-Old Legacy Systems (GAO-25-107795)(opens in a new tab)U.S. Government Accountability Office, July 17, 2025 · files.gao.gov
  2. 2.Flexera Finds Cloud Value is Rising While AI Waste Grows(opens in a new tab)Flexera, March 18, 2026 · flexera.com
  3. 3.Bad Practices(opens in a new tab)Cybersecurity and Infrastructure Security Agency · cisa.gov
  4. 4.The NIST Cybersecurity Framework (CSF) 2.0 (NIST CSWP 29)(opens in a new tab)National Institute of Standards and Technology, February 26, 2024 · nvlpubs.nist.gov
  5. 5.The European Supervisory Authorities designate critical ICT third-party providers under the Digital Operational Resilience Act(opens in a new tab)European Banking Authority, November 18, 2025 · eba.europa.eu
  6. 6.Interagency Guidance on Third-Party Relationships: Risk Management(opens in a new tab)Board of Governors of the Federal Reserve System, FDIC and OCC (Federal Register), June 9, 2023 · federalregister.gov