END-TO-END PRODUCT DESIGN / DEUTSCHE BANK
Risk is messy.The decision can't be.
I designed dbAssess end to end and took it through to worldwide deployment at Deutsche Bank. AI researched Inherent Risk; assessors examined its reasoning and made the final call without juggling spreadsheets, email and separate tools.
THE QUESTION / A RATING WORTH QUESTIONING
One change. Several questions.
IR rises from Important to Significant while likelihood falls. The AI explanation is available beside the earlier rationale; Agree remains one action.
ILLUSTRATIVE INHERENT RISK / A CHANGE THAT NEEDS EXPLAINING
PREVIOUS / ANNUAL 2025
ImportantAI-RESEARCHED / ANNUAL 2026
Significant ↑THE QUESTION
The rating rose.
Likelihood fell.
Why?
The AI’s explanation is available; Agree remains immediate. A different rating needs a reason.
ILLUSTRATIVE ASSESSMENT / THE FIGURES COME FROM THE RECONSTRUCTED PRODUCT SCREENS
THE WORK THAT DISAPPEARED
The judgement was the job. Chasing evidence wasn’t.
BEFORE dbASSESS
Previous ratings and rationales sat apart from controls, findings and events. Assessors used spreadsheets and email to hold the working picture together.
WITH dbASSESS
One experience brought together the queue, AI-researched Inherent Risk, prior rationale, findings, granular drivers and the assessor’s decision. No parallel working file required.
Less evidence assembly. Fewer workarounds. Greater confidence.
AI RESEARCH / THE ASSESSOR’S JUDGEMENT
The AI researched the risk.
Would you agree?
AI independently researched Inherent Risk and proposed a higher rating even as likelihood fell. The earlier rationale, generated explanation, controls, findings and events sit together for inspection. Agree Indicative Rating remains one immediate action; Provide My Own Rating requires an alternative position and a reasoned explanation. The design supports scrutiny without making it compulsory.
RESEARCH / RECOMMENDATION / JUDGEMENT
A researched rating is a starting point.
- THE AI RESEARCHES
The system conducts its own research into the risk.
- INDICATIVE IR + EXPLANATION
It generates a proposed inherent-risk rating, likelihood, severity, impact category and structured rationale.
- THE ASSESSOR CAN EXAMINE
Previous rationale, controls, findings, events and drivers are available to inspect.
- THE HUMAN DECIDES
Agree with the indicative rating, or provide a different position with explanation and evidence.
01 / DECIDE WHAT NEEDS ATTENTION
A busy grid. A much faster way to find the change.
I kept current, indicative and proposed ratings visible together across Inherent Risk, Control Environment and Residual Risk. Breaking those comparisons into separate views would have slowed prioritisation. Directional arrows and filters let assessors identify movement without opening every record.
THE TRADE-OFF / Preserve the comparison, even at the cost of a dense first screen.
02 / THE AI'S RESEARCH
A recommendation
with its working visible.
AI independently researches the risk and generates an indicative Inherent Risk rating with a structured explanation. The previous rationale, controls, findings and events sit alongside it. That research can be examined before Agree Indicative Rating, but the UI does not require it; assessors no longer need to rebuild the working picture in spreadsheets or email.
03 / THE DECISION ON THE RECORD
Don't separate the rating from its reasoning.
The previous rating, the newly proposed position and their written rationales remain visible together. That replaced the manual comparison assessors had maintained across different files and made the basis of the judgement easier to examine.
DESIGN CHOICE / Previous position and new judgement, side by side.
04 / AGREEMENT VERSUS CHALLENGE
One click to agree.
A proper way to disagree.
Agree Indicative Rating takes one action. The AI research and previous rationale are visible, but reading them is not enforced. Provide My Own Rating requires an alternative position and rationale, with evidence attachments available. Both paths stay in the assessment—no separate spreadsheet or email justification.
THE COST OF THE CHOICE / Agreement is fast; disagreement demands a reason.
05 / KEEPING THE INVESTIGATION CONNECTED
Less room for the table.
More room for judgement.
I kept granular investigation inside the parent assessment. The narrower view costs screen space, but preserves the connection to the judgement being examined. Assessors could filter driver-level records without opening a separate working tool or rebuilding their context.
THE DETAIL DID NOT REQUIRE A DETOUR
Open a driver. Keep the assessment in sight.
The selected driver's historical rating and rationale appear alongside the granular list. Assessors can move into the underlying evidence and back to the overall decision without reconstructing where they were.
WHAT CHANGED / DEPLOYED WORLDWIDE
Less assembly.
More judgement.
dbAssess was built and deployed worldwide at Deutsche Bank. The connected workflow replaced spreadsheet-and-email comparisons and separate evidence-gathering tools. Efficiency improved, workarounds fell away, and trust in judgement rose because the AI research, supporting evidence and final human call stayed visible together.
EFFICIENCY
Less time gathering and reconciling evidence across disconnected sources.
TRUST IN JUDGEMENT
Previous rating, AI-researched explanation and human decision in the same record.
FEWER WORKAROUNDS
No parallel spreadsheets or email chains required to complete the assessment.
Product screens use illustrative assessment data; deployment and workflow consolidation are confirmed.
Scroll to inspect the screen. Select Close to return.
Scroll to inspect the screen. Select Close to return.
Scroll to inspect the screen. Select Close to return.
Scroll to inspect the screen. Select Close to return.
Scroll to inspect the screen. Select Close to return.
Scroll to inspect the screen. Select Close to return.
Scroll to inspect the screen. Select Close to return.