Anticipate & Handle Objections¶

Do not get defensive. Objections are signs of engagement.

Common Archetypes¶

Objection: "How do we know why it chose to deny this loan? We can't use a black box."
Handling: Introduce local interpretability immediately. "That's exactly why we use LIME. For every single denial, the model outputs a receipt showing the top 3 reasons, like 'Income too low' or 'Recent late payment'. Would you like to see an example receipt?"

Objection: "You said the accuracy is 85%. That means it's wrong 15% of the time. We can't risk that."
Handling: Anchor against the current baseline. "You're right, it is not perfect. However, our manual human review process is currently running at 65% accuracy. This model represents a 20% absolute improvement, saving 40 hours of manual work a week."

Objection: "What if a customer has a spelling mistake in their name, lives abroad, and uses a VPN? Will it break?"
Handling: Acknowledge and redirect. "Excellent edge case. We haven't explicitly trained for that exact combination. However, our fallback protocol automatically routes low-confidence predictions (under 50%) to a human agent. The model doesn't handle everything, it handles the 80% routine traffic."

KSB	Description	How This Addresses It
S5	Deployment, value assessment, and ROI	Translating model performance into business impact
S6	Communicate through storytelling and visualisation	Presenting ML results to non-technical stakeholders
B4	Consideration of organisational goals	Framing technical results in terms of business objectives
B1	Inquisitive approach	Exploring creative ways to explain model behaviour