Skip to content

Research and development

What I am exploring next

Current research and development, connected to practical finance work. Each track explains what I have tried, what it demonstrates and what still needs testing.

01 · Active

2 tracks

AI workflows with financial checks

AI can carry out several steps in a finance process, provided each important step has clear checks and human approval.

  • Multi-step orchestration ·
  • Tool use behind a deterministic gate ·
  • Confidence-threshold routing ·
  • Evaluation sets that measure quality ·
  • AI coding and implementation assistants including Claude Code, Codex, and Antigravity
Next validation Build an evaluation set for the lowest-confidence document classes.

Connecting purchases, stock and salesAssessing acquisition opportunities (M&A)

Making methods easier to reuse

Documented methods, procedures and prompts help other teams repeat work that has already proved useful.

  • Reusable frameworks, procedures, runbooks and prompt libraries ·
  • Versioned markdown and Git-based knowledge capture ·
  • Obsidian-style project notes and learning logs where useful ·
  • Claude Code, Codex, and Antigravity for structured implementation workflows
Next validation Publish one reusable framework or runbook as a worked example.

Assessing acquisition opportunities (M&A)Business Central remediation and integration readiness

02 · Deepening

3 tracks

Consistent data for group reporting

A shared data model helps group reports treat companies, scenarios and currencies consistently.

  • Dimensional design ·
  • Conformed dimensions ·
  • Grain selection ·
  • Scenario and currency modelling ·
  • Staging, canonical and semantic warehouse layers
Next validation Move the blueprint into a managed store fed by the standardisation programme.

Group financial intelligence modelFinance data standardisation programme

Power BI and DAX reporting

Dashboards connected to accounting systems help finance and operations see what needs attention.

  • Published Power BI inventory reconciliation reporting ·
  • Direct connection to Dynamics 365 Business Central / Navision data ·
  • DAX measures for reconciliation, exceptions, and operational review ·
  • Dashboard usability for finance and operational stakeholders
Next validation Publish a sanitised case study of the inventory reconciliation report.

ERP-connected inventory reconciliation dashboardGroup financial intelligence model

Turning documents into usable data

Tested Python processes turn varied financial documents into organised data that people can check.

  • Regex-first hybrid parsing with LLM residual handling ·
  • SHA-256 content-hash caching ·
  • Cost-aware model routing ·
  • Typed code, focused tests and meaningful coverage ·
  • Exception logging for reviewer follow-up
Next validation Package one pipeline with tests, documentation and redacted inputs.

Connecting purchases, stock and sales