Sid Pednekar, an AI, technology, and data executive in asset management, compares two explicitly hypothetical firms in this 18-minute AI Engineer talk.
Spend judgment on the right bottleneck
- Define the business problem before choosing a model or copilot.
- Faster earnings summaries can lose context while leaving the real problem, late investment signals, untouched.
- Focus engineering and domain expertise on a few valuable end-to-end workflows rather than counting scattered pilots.
Redesign the work
- Let analysts define an investment thesis while automation monitors supporting evidence, contradictions, and exceptions.
- Measure where freed capacity goes: better decisions and deeper investigation matter more than additional generated reports.
- These workflows illustrate the argument; the talk reports no measured returns from the two firms.
Build a shared core
- Bring engineering, business, and domain specialists into the same working process.
- Reuse datasets, semantic definitions, known gaps, and lessons about which approaches work.
- Standardize shared infrastructure while allowing specialized workflows at the edges.
- Make organizational knowledge accessible through expert, workflow, or coordination agents as appropriate.
Make the system learn
- Feed corrections and successes back into tools, context, and shared practices, preserving the supporting evidence.
- Track business outcomes after deployment so saved time becomes useful capacity.
- Survey figures cited during the talk provide context, not a demonstrated causal test of this framework.
“The individual learned, but did the system?” — Sid Pednekar, 17:02