.webp)
A UK CIO we spoke to in June had a 22-year-old Oracle Forms system nobody left in the business actually understood. His board wanted a modernisation programme and would not stop asking why nothing moved. He had three consultancy proposals: £2.4m, £4.8m, and £7.1m, each for what the vendors described as the same work.
He was more confused after the tender than before he ran it.
The state of enterprise modernisation in 2026 is that the technology has moved faster than the pricing has. AI code assistants can now analyse legacy codebases at speed that was science fiction two years ago. Gartner reports that 45 percent of software engineers are already recording 10 percent-plus productivity gains from AI tooling. Yet vendors quote as if it were still 2022, and CIOs sign because the numbers look defensible against 2022 benchmarks.
This article is the honest read on 2026 modernisation cost. Real bands per approach. Where AI genuinely changes the maths. Which of the Five R's applies to which application. And the hidden costs that turn a £2m estimate into a £4m outturn.
The Five R's Framework and Real 2026 Costs
Five approaches, real per-application cost bands.
Retire (£5k to £30k per app in decommissioning cost). Turn the app off, archive the data, redirect users. Gartner puts 20 to 30 percent of the typical enterprise app estate in this bucket. It is the cheapest, fastest, highest-ROI programme win most CIOs skip.
Rehost or lift-and-shift (£30k to £120k per app, 6 to 12 weeks). Move the app to cloud infrastructure without changing the code. Fast, quantifiable, resolves data centre exit but does not fix technical debt. Best for apps with 3 to 5 years of remaining useful life.
Replatform or lift-tinker-and-shift (£60k to £250k per app, 3 to 6 months). Minor optimisation to use managed cloud services (managed database, load balancer, message queue) without full architectural rework. Best for apps that need to stay useful for 5 to 10 more years.
Refactor or re-architect (£150k to £800k per app, 6 to 15 months). Deep restructuring into cloud-native architecture (microservices, containers, serverless). Highest upfront cost, highest long-term value, hardest programme to run without slippage.
Replace with SaaS or low-code (£50k to £400k per app, 3 to 9 months). Substitute the app with a commercial SaaS or a low-code build. Best when the app does something standard (payroll, CRM, expense management) and your customisations are cosmetic rather than commercial.
The right R for each app is a portfolio decision, not a technology decision. See Microsoft's Cloud Adoption Framework for the classification methodology most enterprises now follow.
Where AI-Assisted Modernisation Actually Changes the Maths
Two years ago, refactoring a COBOL system was a multi-year manual project. In 2026, GitHub Copilot, Cursor, and Claude in Copilot ecosystems can analyse and document 80,000 lines of legacy code in under an hour. Discovery time (the phase where engineers work out what the legacy code actually does) is down 60 to 80 percent.
That does not make refactoring free. It cuts the analysis and documentation phases substantially. The code generation, integration testing, and business acceptance phases still take real engineering time. Realistic AI-assisted refactor cost lands 25 to 40 percent below 2023 benchmarks for the same scope.
Where AI helps most:
Legacy code analysis and documentation
Test generation for undocumented legacy behaviour
Straightforward refactoring (framework upgrades, language version bumps)
Data model documentation and mapping
Where AI still struggles:
Business logic that lives in stored procedures and triggers
Integrations with third-party systems where API docs are stale
Regulatory and compliance edge cases
Anything requiring judgement about intent versus implementation
The 45 percent productivity number Gartner reports is real for the tasks AI is good at. Aggregate programme productivity gain lands closer to 15 to 25 percent because significant work remains human.
The Hidden Costs Nobody Puts in the Business Case
Five costs that surprise CIOs six months in.
Data migration and reconciliation. Extract, transform, validate, reconcile. 15 to 25 percent of total programme cost. Rarely fully priced in the initial estimate.
Integration rebuilding. Every legacy app has three to twelve integrations to other systems. Each must be rebuilt or reconnected on the new platform. £15k to £80k per integration.
Licence overlap during transition. Running old and new in parallel for 3 to 12 months means paying two licences. £50k to £500k per year of overlap depending on estate size.
User training and change management. New UI, new workflows, retraining hundreds of users. 5 to 15 percent of technical build cost, budgeted late or not at all.
The 15 to 25 percent maintenance tail. The new system needs ongoing engineering. Assume 15 to 25 percent of build cost per year in year two and beyond.
Include these in the business case at day one. The Gartner CIO Board consult on legacy modernisation identifies data and integration cost as the two most-commonly underestimated line items across surveyed CIOs.
When to Rehost, When to Refactor, When to Replace
Three tests to run against each app in the portfolio.
Business value test. Does this app produce competitive advantage, or is it commodity capability? Commodity capability rarely justifies refactor; SaaS replacement is usually better.
Remaining useful life test. How many years does this app need to keep working? Under 3 years: rehost. 3 to 5 years: rehost or replatform. 5 to 10 years: replatform or refactor. Over 10 years: refactor or full replace.
Technical health test. Is the code understood by current staff? Does it have automated tests? Is it on a supported framework? Poor technical health increases the cost of any approach and often tips the decision toward retire or replace.
Score each app against all three. The portfolio that emerges usually shows 20 to 30 percent retire, 30 to 40 percent rehost or replatform, 15 to 25 percent refactor, and 10 to 20 percent replace. That mix matches what Gartner sees across CIOs surveyed for the Consult the Board series.
How to Phase the Programme Without Breaking Operations
Three-year phased programme structure for a mid-to-large enterprise (200-plus applications in the estate).
Year 1: Retire the deadwood, rehost the routine. Complete the application portfolio review. Retire the 20 to 30 percent with no business value. Rehost 15 to 25 percent of routine applications. Cost: 20 to 30 percent of total programme.
Year 2: Replatform mid-value applications, start refactor on top 3 assets. Move 15 to 25 percent onto managed cloud services. Begin refactoring the two or three most business-critical applications. Cost: 40 to 50 percent of total programme.
Year 3: Complete refactor of top assets, replace commodity apps with SaaS. Finish refactoring the core business-critical applications. Replace 10 to 20 percent with SaaS. Cost: 30 to 40 percent of total programme.
For a 200-application estate, total programme cost typically lands £8m to £25m over three years, delivering 40 to 60 percent operating cost reduction on IT infrastructure by year four.
What We Have Seen Fail in Enterprise Modernisation
Four failure modes we see repeatedly.
Big-bang refactor of everything. Board approves a three-year rewrite that runs to five years and delivers half the scope. Almost always fails. Portfolio-phased delivery works.
Skipping the retire step. Teams jump straight to rehosting apps that should have been decommissioned. Wastes 15 to 30 percent of the programme budget.
Underestimating data migration. Data is dirtier than legacy documentation admits. Reconciliation takes twice as long as planned. Budget accordingly.
Business owner absence. IT-led programmes without business ownership at the executive level slip because scope questions do not get decided. Every phase needs a business sponsor.
You can see our work at our portfolio. If you want a candid conversation about your specific estate, get a modernisation assessment from WhiteStone.
Frequently Asked Questions
How do I decide between rewrite, replatform, and refactor?
Score each app against three tests: business value (competitive vs commodity), remaining useful life (under 3 to over 10 years), and technical health (understood, tested, supported). Commodity plus poor technical health tips toward replace. High business value plus 5 to 10 year useful life tips toward refactor. Everything else lands in rehost or replatform.
How much does a legacy ERP modernisation cost?
For a mid-market UK enterprise, ERP modernisation lands £800k to £4m depending on approach and integration count. SaaS replacement (moving to NetSuite, Dynamics 365, or SAP S/4HANA) at the lower end. Custom refactor of a bespoke ERP at the top end. Add 15 to 25 percent for data migration and integration rebuilding.
How long does an enterprise modernisation programme take?
For a 200-application estate, plan for three years with three phases: retire and rehost in year one, replatform and start refactor in year two, complete refactor and SaaS replacement in year three. Larger estates (1,000-plus applications) take four to six years phased across business units.
Does AI actually cut modernisation cost?
Yes, for the specific tasks AI is good at (legacy code analysis, documentation, test generation, straightforward refactoring). Discovery phase is 60 to 80 percent faster. Aggregate programme cost lands 25 to 40 percent below 2023 benchmarks for the same scope. Business logic in stored procedures, third-party integrations, and compliance edge cases still take human engineering time.
What is a realistic year-one ROI?
Retiring the 20 to 30 percent of apps with no business value delivers immediate cost savings (licences, infrastructure, support). Realistic year-one savings: 10 to 20 percent of the legacy IT run cost, often enough to fund most of the year-two modernisation work.
The One Thing to Remember
Enterprise modernisation is a portfolio decision, not a technology decision. The mistake CIOs make is treating the whole estate as one problem. The reality is 20 to 30 percent should be retired, 30 to 40 percent rehosted or replatformed, 15 to 25 percent refactored, and 10 to 20 percent replaced with SaaS. Get that mix right and the programme funds itself.
If you want a candid assessment of your specific estate, browse our custom software development services or come to the modernisation call.
.webp)