Pain point inventory
Start with the bottleneck list from workflow analysis: latency, rework, judgment-heavy routing, data extraction delays, exception volume, and decision lag.
Method
Five-step matching sequence
Capability matching is not a lookup table. The goal is to identify structural fit between a workflow's failure modes and what specific AI capabilities are actually good at doing — then find the overlaps where one capability clears multiple bottlenecks.
Capability taxonomy mapping
For each pain point, identify which AI capability class applies: prediction, extraction, classification, generation, or orchestration. Some pain points map to multiple capabilities.
Multi-problem resolution scan
Identify capabilities that appear across multiple pain points. A single orchestration layer that routes, summarizes, and escalates may resolve five bottlenecks simultaneously.
Data and signal availability check
Validate that each matched capability has the required inputs: historical data volume, label availability, document structure, or real-time signal access needed to function reliably.
Build vs. buy decision framing
For each capability match, assess whether commodity models are sufficient or whether the use case requires proprietary training, fine-tuning, or a custom knowledge layer.
Outputs
Artifacts produced by the process
Capability heat map
Visual matrix showing which AI capabilities address which workflow pain points, with overlap highlighted.
- Pain point vs. capability grid
- Multi-resolution opportunities flagged
- Confidence rating per match
Signal readiness assessment
Evaluation of data availability for each capability match — what exists, what needs work, what is missing.
- Data quality and volume baseline
- Label and annotation requirements
- Real-time vs. batch access constraints
Leverage multiplier ranking
Ranked list of capabilities ordered by how many pain points each one resolves.
- Cross-bottleneck resolution count
- Implementation complexity estimate
- Governance and risk exposure flag
Build vs. buy recommendation
Decision framework for each capability match — off-the-shelf model, fine-tuned, or proprietary build.
- Commodity model fit assessment
- Proprietary data advantage potential
- Time and cost delta for custom builds
Engagement Cadence
How the process runs in practice
Typical timeline: 1-2 weeks
- Days 1–3: pain point inventory review and capability taxonomy mapping
- Days 4–6: multi-problem resolution scan and data readiness check
- Days 7–10: build vs. buy framing, leverage ranking, and synthesis
Output: a prioritized capability map with clear data requirements and build recommendations, ready to feed directly into model strategy decisions.