The Hype Problem
Every vendor now claims "AI-powered" something. Most of it is marketing. Some of it is real. Here's how to tell the difference.
Where AI Actually Helps
1. Plan Analysis
AI can read plan documents and identify:
- Ambiguous language
- Missing definitions
- Logical contradictions
- Gap in coverage
This is pattern recognition at scale—something AI does well.
2. Gaming Detection
AI can monitor transaction patterns and flag:
- Unusual deal clustering
- Timing anomalies
- Split patterns
- Credit disputes
This is anomaly detection—another AI strength.
3. Scenario Modeling
AI can simulate plan changes against historical data:
- Cost projections under different assumptions
- Payout distributions
- Attainment predictions
This is computational simulation—AI handles complexity well.
4. Document Generation
AI can draft:
- Plan summaries
- Rep statements
- Exception documentation
- Audit reports
This is language generation—AI's core capability.
Where AI Doesn't Help (Yet)
1. Strategic Decisions
"Should we cap accelerators?"
This requires business judgment that AI can't provide. AI can model the impact of capping vs. not capping—but the decision is human.
2. Exception Approval
"Should we approve this exception?"
AI can flag precedent and calculate impact, but the judgment call requires human context and accountability.
3. Conflict Resolution
"How do we handle this dispute?"
Disputes involve relationships, politics, and judgment that AI can't navigate.
4. Change Communication
"How do we tell the sales team about this change?"
AI can draft language, but communication strategy requires human understanding of culture and context.
The Integration Model
The best AI applications in SPM follow a pattern:
1. AI does the analysis (fast, comprehensive)
2. Humans make the decisions (judgment, accountability)
3. AI implements at scale (consistent, traceable)
4. Humans handle exceptions (relationship, context)
This isn't AI replacing humans. It's AI augmenting humans—doing the computational heavy lifting so humans can focus on judgment.
The Build vs. Buy Question
Should you build AI capabilities or buy them?
Build if:
- Your comp structure is highly unique
- You have strong data science resources
- Integration with existing systems is critical
- You want full control over the logic
Buy if:
- Your needs are relatively standard
- You want faster time-to-value
- You prefer proven algorithms
- You don't have AI expertise in-house
Most organizations should buy for commoditized AI (gaming detection, scenario modeling) and build for strategic AI (custom analysis, proprietary rules).
Tags