HMRC’s £2bn IT Shift: What UK SMEs Should Copy

HMRC is lining up more than £2 billion in technology procurement over the next two years. That headline is easy to scroll past—until you realise what it’s really saying: even one of the UK’s biggest, most process-heavy organisations is still wrestling with legacy systems, and it’s willing to spend serious money to get out.

For UK small businesses, this matters for a simpler reason than “public sector news”: the same problems show up at SME scale, just with smaller budgets and fewer people to absorb the pain. If HMRC—collecting £858.9bn in tax revenues in 2024–25—still struggles to modernise, your finance system that “sort of works” and your spreadsheet-based reporting definitely aren’t going to modernise themselves.

This post sits in our Technology, Innovation & Digital Economy series for one reason: the UK economy is moving toward data-driven operations, automation, and AI-assisted decision-making. HMRC’s plan is a loud signal of where the direction of travel is heading—and what practical steps SMEs can take to modernise faster, with less drama.

What HMRC’s £2bn tech plan really tells us

Answer first: HMRC’s spending plan is a reminder that “digital transformation” isn’t a one-off project—it’s years of disciplined work to replace brittle systems, consolidate data, and run services more reliably.

According to HMRC’s procurement pipeline (reported by Business Matters), the early focus is a data warehouse transformation expected to be worth around £410m, combining “run and modify” of existing platforms with migration and eventual decommissioning of legacy data warehouse systems. A big organisation can’t switch off core reporting overnight; it has to run old and new in parallel.

The pipeline also points to a £350m public cloud computing procurement (replacing an existing AWS agreement), and a £306m “Digital Platforms Run and Change Products” contract to support live applications—including legacy platforms. In other words, HMRC isn’t only buying “new shiny tools”; it’s funding the unglamorous work: keeping critical services stable while it changes the engine underneath.

The National Audit Office (NAO) added a sobering note in November 2025: HMRC is taking longer than planned to exit legacy systems and hasn’t yet achieved expected efficiencies from its digital services programme. That’s the part most organisations recognise.

Snippet-worthy reality: Legacy systems don’t fail all at once—they quietly tax your time, your reporting accuracy, and your ability to change.

Legacy systems: HMRC’s problem is your problem (just scaled down)

Answer first: SMEs rarely call it “legacy,” but if your business depends on spreadsheets, disconnected apps, and manual re-keying, you’re already paying the legacy tax.

HMRC’s legacy estate is complex and costly, but the pattern is familiar:

• Data lives in multiple places, with no single source of truth
• Reporting takes too long, so decisions lag behind reality
• Changes are risky because nobody’s fully sure what depends on what
• Workarounds multiply: exports, re-uploads, manual checks, “temporary” spreadsheets

For a small business, this shows up as:

• Month-end dragging on for days (or weeks)
• Sales and marketing teams arguing about “whose numbers are right”
• Customer service searching inboxes for order history
• Cash flow surprises because invoicing and banking aren’t joined up

If you’ve found yourself saying “don’t touch that sheet, it breaks everything,” you’re living in a legacy environment—just not one that cost £2bn to create.

The myth: “We’ll modernise when we’re bigger”

Most companies get this wrong. Waiting to modernise until you’re bigger usually means you modernise when your operations are more complex, your customer expectations are higher, and the cost of downtime is worse.

Modernisation for SMEs isn’t about copying HMRC’s spend. It’s about copying the intent: clean up data, reduce manual work, and build systems you can change without fear.

The practical lesson: data first, then automation, then AI

Answer first: AI tools only work well when your data is consistent and your processes are standardised—HMRC’s “data warehouse transformation” is the clue.

A lot of SME AI conversations jump straight to chatbots and content creation. Useful, yes—but the bigger win is often boring: joining up operational data so automation can happen safely.

HMRC’s plan starts with data foundations for a reason. When your data is messy, AI becomes a confidence trick: it produces fluent output that may be wrong, and nobody can audit it.

A simple SME version of “data warehouse transformation”

You probably don’t need a data warehouse team. You do need these three outcomes:

1. One source of truth for customers, products, and invoices (even if it’s just your accounting platform + a proper CRM)
2. Clean identifiers (consistent customer names, order IDs, VAT treatment, SKUs)
3. Automated flows between systems (so you stop copy/pasting between tools)

Once those are in place, AI becomes genuinely helpful:

• Categorising transactions and suggesting reconciliations
• Drafting customer responses using your order and policy data
• Forecasting cash flow based on invoice cycles and seasonality
• Flagging anomalies (duplicate supplier invoices, unusual refunds, odd margin changes)

Good AI is built on boring consistency. If your inputs are chaotic, your “insights” will be too.

Where SMEs can use AI tools immediately (without ripping everything out)

Answer first: Start with “thin layers” of AI—small, targeted automations that sit on top of existing systems—then consolidate platforms once you’ve proven value.

HMRC is funding “run and change” alongside legacy support because big systems can’t pause. SMEs should copy this approach: improve while you operate.

1) Finance ops: faster month-end and fewer surprises

If you want one place to start, start here. Finance touches everything.

Practical AI-assisted wins:

• Auto-categorisation of transactions with human approval
• Invoice chasing sequences that adapt to customer behaviour
• Extracting data from supplier PDFs and matching to POs
• Cash flow forecasts that update weekly, not quarterly

What to measure (pick two):

• Days to close month-end
• % of transactions manually categorised
• Debtor days / overdue invoice value

2) Customer service: reduce response time without sounding robotic

AI helps most when it has context: order history, SLAs, delivery status, returns policy.

A sensible setup:

• A shared inbox/helpdesk as the system of record
• Canned responses that AI personalises (not improvises)
• A “human sign-off” rule for refunds, complaints, or legal topics

What to measure:

• First response time
• Time to resolution
• Repeat contact rate

3) Marketing: speed up production, keep strategy human

For lead generation, AI can speed up output, but your positioning still matters more than volume.

Good uses:

• Turning a case study call into a blog + email + LinkedIn post
• Generating variant ads while you control claims and compliance
• Summarising call notes into pain points and objection handling

Guardrails that prevent mess:

• A short “brand notes” doc (tone, offers, banned claims)
• A facts-only source (pricing, deliverables, terms)
• A review step for anything that mentions results or numbers

Don’t copy HMRC’s mistakes: avoid these transformation traps

Answer first: The biggest risk isn’t choosing the wrong tool—it’s running transformation as a side project with unclear ownership and fuzzy definitions of “done.”

The NAO’s warning on slower, more expensive progress is a familiar story across sectors: modernisation drifts when it’s not tied to measurable operational outcomes.

Here’s what I’ve found works better for SMEs than big “transformation programmes”:

Define outcomes in operational language

Not “implement a new CRM.” Instead:

• “Reduce quote-to-invoice time from 3 days to 1 day.”
• “Cut month-end close from 10 working days to 5.”
• “Get a weekly margin report we trust by Monday 10am.”

Keep a ‘legacy exit list’ (yes, even for spreadsheets)

Write down the workarounds you want to retire:

• The spreadsheet that merges Shopify + Xero
• The manual VAT check step
• The shared mailbox rules nobody understands

Every month, kill one.

Buy less tech; fix more process

AI tools won’t rescue a broken process. If approvals are unclear, data fields are optional, and nobody owns customer records, you’ll just automate confusion.

A clean process is the cheapest “AI upgrade” you’ll ever make.

A 30-day AI modernisation plan for UK small businesses

Answer first: In one month, you can map your data, automate one high-friction workflow, and set up basic governance—without replatforming your whole business.

Here’s a pragmatic plan you can actually finish.

1. Week 1: Map your systems and data

   • List every tool used for sales, ops, support, finance
   • Identify the “master” for customers and invoices
   • Note every place data gets re-keyed by a human

2. Week 2: Pick one workflow to fix (choose the one with daily pain)

   • Examples: invoice chasing, lead follow-up, support triage, bank reconciliation
   • Define success metrics (time saved, fewer errors, faster responses)

3. Week 3: Add AI with guardrails

   • Human approval for money, legal, or complaints
   • Audit trail: keep prompts, outputs, and changes logged
   • Limit access to sensitive customer and payroll data

4. Week 4: Make it repeatable

   • Document the new workflow in one page
   • Train the team (30 minutes, recorded)
   • Schedule a monthly review to improve and expand

If you do this consistently, you’ll modernise faster than businesses that wait for a “perfect” replatform.

What HMRC’s spend signals for the UK digital economy

Answer first: Public sector investment tends to pull suppliers, skills, and standards in a direction—and SMEs that align early get a compounding advantage.

HMRC’s procurement pipeline includes major spending on data platforms, cloud services, and ongoing “run and change” support. That points to a UK-wide reality: resilience, data governance, and automation are becoming table stakes, not optional extras.

For SMEs, the opportunity is straightforward: you can adopt the benefits of modern infrastructure—cleaner data, faster reporting, smarter workflows—without paying enterprise prices.

The question worth asking isn’t whether your business should “do AI.” It’s whether you’re willing to keep paying the legacy tax in time, errors, and missed opportunities while everyone else modernises.