BEST PRACTICE
What separates AI deployments that show up on the P&L
01
Boring beats flagship
Measurable P&L wins are often narrow, high-volume, and back-office — not customer-facing. Start where process is dull and repeatable.
02
Redesign, don’t bolt on
Only 21% of companies redesign workflows, yet it’s the no. 1 driver of EBIT impact, according to McKinsey.
03
Don’t chase 100%
Automating the last 10% can cost more than the first 90%. Right-size the scope for peak margin gain.
04
Design for durability
Build robust automations that survive model, data and systems change, not fragile deep links or APIs.
05
The AI is the easy part
BCG defines success criteria as 10% algorithms, 20% technology and data, 70% people and process. Focus on the 90% that isn’t AI.
How this plays out by function
Go-to-market
AI upskills average reps toward your best ones. Call and deal analysis moves win rates more than email drafting.
Customer support
Start narrow by volume, fix the knowledge behind it, design the human handover. Track genuine resolution, not deflection.
Finance
Don’t automate the workaround. The prize isn’t a faster reconciliation, it’s no reconciliation, because the data reconciles itself as it flows.
Proposal & tender
AI owns the data, drafts and formatting; experts spend their time on the 10% that wins.