AI and Automation

AI and Automation in Housing Finance: Faster Routine Work, People Still Approving

Housing finance teams are asked to do more with the same people. Automation and AI can take on routine work, but only with clear rules about who decides. This is how we bring them into a housing finance system step by step.

November 12, 2026
7 min read

The most useful automation in housing finance is often the least visible.

Before any AI feature is switched on, a modern finance system can already remove a large share of routine work through standard automation. AI then adds help where judgement is involved: spotting the unusual, drafting the routine and answering questions in plain language. In both cases people keep the decisions that matter.

Our standard approach: automate the routine first, add AI where it helps, and keep people approving postings and payments.

Automation available from day one

  • Recurring and automated journals for regular charges, allocations and reversals.
  • Automated accruals from received-but-not-invoiced purchase orders.
  • Allocations that spread shared costs across schemes and cost centres by agreed rules.
  • Invoice capture and matching: invoices read from documents, matched to the order and receipt, and only exceptions sent to a person.
  • Approvals and notifications routed by value, cost centre and delegation, with escalation when someone is away.

These are configured, not coded, so they are maintained through the normal system setup.

Where AI helps

Anomaly checks

Flagging unusual journals, duplicate-looking invoices or spending that breaks a pattern, for a person to review.

Assisted journals

Suggesting coding or drafting routine journals from supporting documents, ready for a person to check and post.

Questions in plain language

Asking "which schemes are over budget this quarter?" and getting an answer from governed data.

Microsoft Copilot

For Microsoft 365 users, summarising reports and drafting commentary from approved finance data in Power BI and Teams.

Predictive forecasting and ESG (environmental, social and governance) reporting can follow, once the core reporting and forecasting cycle is stable and trusted.

People keep the decisions

Every AI use case is agreed before it is switched on, with clear limits:

  • Postings, payments and changes to supplier details stay with authorised people. AI suggests; a person approves.
  • Each use case has an approved purpose and a defined set of data it can use.
  • Suggestions and approvals are logged, so the audit trail shows who decided what.

Data protection and security

AI features follow the same security, privacy and audit controls as the rest of the finance system. Our standard approach is that the housing association's data is not used to train general AI models without an explicit, authorised agreement.

Introduced through tested releases

Oracle updates its cloud applications every quarter, and new AI features arrive through those updates. We review each release, assess what is relevant, test it in a separate environment and agree with the finance team before anything reaches live use. New capability is adopted when it is useful, not just because it is available.

A staged path

  1. First: a stable close, exception reporting, budget holder forecasting and automated board reports.
  2. Next: forecast drivers and scenarios, anomaly checks and ESG data.
  3. Later: Copilot-assisted analysis and wider automation, as the organisation grows.

Each stage has its own business case and measurable acceptance before it starts.

HR: routine requests handled by AI

At a UK repairs and maintenance company in the housing sector, 80% of routine HR requests are handled by AI.

Finance: month-end close from 18 days to under 5

At a large UK housing group, Oracle Fusion finance with PCL One cut month-end close from 18 days to under 5.

Less routine work, the same accountability

PCL One implements Oracle Fusion for housing providers and introduces automation and AI through the same tested release process as every other change.

The aim is more time for judgement, not less judgement.