AI-assisted reporting
Context before commentary.
How can a report explain the movement behind a metric?
Bring risk data and supporting context together so risk owners have a better starting point for their analysis.
Explore the walkthroughI observed risk owners preparing branch reports that often restated metric movements, with little supporting analysis or external context.
I designed, coded and rolled out a ChatGPT-assisted reporting workflow. I owned the workflow, external-news integration and prompt development, while risk owners retained responsibility for the final report.
- Supported by the team
- More branchesWith less report preparation time
- Analysis and drafting
- ChatGPTRisk data combined with supporting context
- Final review
- Risk ownersReview, refine and finalise the narrative
My contribution
My role in discovery and rollout.
01
Observe
Understand how risk owners use the data and where their reporting needs more context.
02
Design
Map the reporting workflow around the inputs, branch needs, and risk-owner review.
03
Build
Code the workflow that brings the data together and prepares the report draft.
04
Add context
Find and integrate relevant external news alongside the risk data and existing reports.
05
Develop prompts
Create prompts that help ChatGPT analyse the inputs and draft the narrative.
06
Roll out
Put the workflow into use with teams so they can support more branches in less time.
The outcome
More branches supported, less time preparing reports.
AI supports the analysis and drafting. Risk owners in each discipline review, refine, and finalise the report.
01 / The approach
How context reaches the report.
Data, previous reporting and relevant events inform a draft. The risk owner checks the analysis and remains accountable for the conclusion.
- 01
Gather the inputs
Combine current metrics with previous reports and incident context.
- 02
Add external context
Bring relevant news alongside the internal risk data.
- 03
Prepare a draft
Use ChatGPT and the reporting prompts to draft a contextual narrative.
- 04
Review and finalise
The risk owner checks the figures and interpretation before use.
02 / Explore in practice
Explore a branch report draft.
Select a risk metric, review its supporting inputs and open the sample draft to see how the narrative connects them.
The ownership and outcome statements describe my work. The walkthrough uses fictional metrics, sample headlines and prewritten drafts. It does not call a live AI service.
Example 01 · Fintech
Branch Risk Reports
ChatGPT-assisted drafts for branch reporting, reviewed and finalised by risk owners across disciplines.
This walkthrough uses fictional metrics, sample headlines, and prewritten drafts to illustrate the workflow.
Inputs for the AI-assisted draft
Risk data
Current metrics and recent trends from the branch dataset.
Last report
Previous commentary to provide context for the latest figures.
Incidents
Incident retrospectives to support the analysis of a change.
External news
Relevant external news to give the risk owner wider context.
ChatGPT analyses the inputs and drafts the narrative
Prompts bring the risk data, previous report, incidents, and relevant news together. ChatGPT prepares a contextual draft in the report template.
Risk owner reviews the analysis
The risk owner checks the figures, supporting context, and suggested interpretation.
Risk owner finalises the report
Each discipline’s risk owner refines the draft and remains responsible for the final narrative.
From a metric movement to a contextual draft
Choose a risk metric, then open the sample draft to see how the reporting workflow brings the trend, incident history, and external news into the narrative.
Arrears by month. Apr 2.4%, May 1.9%, Jun 2.2%, Jul 2.1%, Aug 2.0%, Sep 3.0%. Latest is 3.0%, +1 pt versus Aug.
The starting point: metric-only commentary
Arrears are 3.0% this month, up 1.0 percentage point from August.