AI Workforce Transformation: Augment, Automate, Protect
- Maturity
- L4
- Domain
- Grow & Keep
- Analytics
- predictive
- Operations AI task-exposure
- 55–65%
- Sponsor
- Chief Transformation & AI Officer
- Confidence
- High
The situation
Task by task and role by role, what is Aurora's AI exposure — what should be automated, augmented or protected — and how fast can the workforce be reshaped to capture the value while staying controlled?
The recommendation on the table
Reframe the AI programme as augmentation-first workforce reshaping
Protects both value and the workforce; governed by the augment-vs-automate rule, it avoids the underdelivery and conduct risk of cut-first programmes.
Trade-offRequires the CTAIO to hold the line against a faster-looking automate-led path.
The evidence
Aurora's AI programme was being run as a technology programme with a workforce afterthought; the reality is the reverse, because AI value is realised by people — by what work humans stop doing, what they do better with AI, and the new control work AI creates. The Banking flagship, BNK-03 maps AI exposure to the task and role level, decides augment-versus-automate by a transparent rule, sequences reshaping against reskilling capacity, quantifies the value net of reskilling and governance cost, and — distinctively — accounts for the risk and model-governance roles AI creates, not just those it removes. It found exposure concentrated on the largest populations, reshaping pace lagging exposure by ~30%, intent skewing automate-led where it should be augment-led, and governance roles dangerously understaffed.
Aurora — AI Workforce Transformation
Task by task and role by role: what to automate, augment or protect — and whether the bank can reshape fast enough while staying controlled.
| Protect | Augment | Automate | |
|---|---|---|---|
| Operations | 1 | 3 | 4 |
| Analytical | 2 | 4 | 2 |
| Technology | 2 | 4 | 2 |
| Advisory | 4 | 2 | 1 |
| Control | 3 | 3 | 2 |
Key takeawayOperations skews automatable; advisory skews protect; most banking work is augment, not automate.
- Augment52%52%
- Automate31%31%
- Protect17%17%
Key takeawayAugmentation-first by design — automate only where redeployment and control checks pass.
Key findings
AI exposure is not a niche programme: at 55–65%, operations is a fifth of the bank, and reshaping pace currently lags exposure by about 30%. The binding constraint is capacity to reskill and redeploy, not ambition — the transformation is rate-limited by the very engineers and trainers who must deliver it.
What we can’t claim
Intent skews automate-led in operations with no redeployment plan, and AI-governance roles are under half staffed while automation accelerates. The most consequential and least-watched risk in the whole transformation is that control coverage falls behind the obligation — turning a productivity win into a regulatory exposure. The credible value number is net of the cost of staying controlled.
Recommendations
Reframe the AI programme as augmentation-first workforce reshaping
high priorityProtects both value and the workforce; governed by the augment-vs-automate rule, it avoids the underdelivery and conduct risk of cut-first programmes.
Trade-off
Requires the CTAIO to hold the line against a faster-looking automate-led path.
Gate automation on reshaping pace and control coverage, and report value net
high priorityEnsures the bank never automates faster than it can redeploy people or govern models, and gives the board a defensible net number.
Trade-off
Adds gates to the transformation and budgets reskilling and governance explicitly — a forward investment.
Analytical framework
How we reached this
Predictive — a task/role AI-exposure map and a reshaping roadmap that captures value while staying controlled, augmentation-first.
ConfidenceMedium
Analytical framework
How we reached this
Predictive — a task/role AI-exposure map and a reshaping roadmap that captures value while staying controlled, augmentation-first.
Methods applied
Statistical techniques
Algorithms
Data sources
Outputs generated
Why this confidence
Exposure assessment is defensible and transparent; forward projections are bounded by external AI/regulatory uncertainty and adoption risk, and reported at the lower bound with stated governance guardrails.
The reasoning
Business context
The flagship and clearest expression of Banking's thesis. It consumes the BNK-01 baseline and BNK-02 productivity intelligence, and is sponsored by the Chief Transformation & AI Officer because its outputs gate the bank's AI-capital and operating-model decisions.
Expected value
The largest value pool in the Banking portfolio — automation savings plus augmented-capacity revenue, net of reskilling and governance cost. Its credibility depends on honest netting and on augmentation-first framing that protects both value and the workforce.
Workforce landscape
Exposure lands where the headcount is: operations (55–65%) is a fifth of the bank. Reshaping pace lags exposure by ~30% — capacity, not ambition, is binding. Intent skews automate-led in operations with no redeployment plan, and AI-governance roles are <50% staffed while automation accelerates.
The analytics journey
Level 4, predictive. The exposure assessment is a transparent, rubric-based task/role model (defensible to regulators); the projections — reshaping pace, net value — are forecasts with explicit confidence, bounded by genuine external AI/regulatory uncertainty and adoption risk. Augmentation-first: AI raises capacity, the bank decides what humans do with it.
Under the hood
Roles are decomposed into tasks scored on structure, judgment, regulatory sensitivity and client-trust content; interpretable exposure scoring yields a disposition (automate/augment/protect); time-series projects reshaping pace; linear programming optimises reskill-vs-redeploy; similarity matching maps people to destination roles. Every automate disposition carries a redeployment obligation and a control-coverage check — automation without both is blocked.
Confidence & evidence
Why you can rely on this
The inconvenient truth
Intent skews automate-led in operations with no redeployment plan, and AI-governance roles are under half staffed while automation accelerates. The most consequential and least-watched risk in the whole transformation is that control coverage falls behind the obligation — turning a productivity win into a regulatory exposure. The credible value number is net of the cost of staying controlled.
Method
Confidence is a deterministic read of KPI strength, target and benchmark coverage across this project — shown on an illustrative reference dataset, computed the same way it would be on live data.
Take this further