|
| 1 | ++++ |
| 2 | +Talk_date = "" |
| 3 | +Talk_start_time = "" |
| 4 | +Talk_end_time = "" |
| 5 | +Title = "Code Was Never the Bottleneck: A Flow Engineering Approach to AI Adoption" |
| 6 | +Type = "talk" |
| 7 | +Speakers = ["steve-pereira"] |
| 8 | ++++ |
| 9 | + |
| 10 | +The software industry is currently obsessed with "AI-assisted coding." We are |
| 11 | +equipping developers with powerful generative tools to write code faster than |
| 12 | +ever before. But if writing code was ever the constraint, it hasn't been for |
| 13 | +decades. |
| 14 | + |
| 15 | +If your developers write 50% more code, but your QA process, deployment |
| 16 | +pipeline, or requirements gathering remains static, you haven’t accelerated |
| 17 | +delivery—you’ve just created a traffic jam. |
| 18 | + |
| 19 | +In this data-driven session, Steve Pereira will apply the principles of Flow |
| 20 | +Engineering to AI adoption. We will move beyond the hype to rigorous systemic |
| 21 | +analysis, using modeling and visualization based on Amdahl’s Law and Little’s |
| 22 | +Law to prove why optimizing non-bottlenecks (like coding speed) yields |
| 23 | +diminishing returns and often degrades system performance by flooding the |
| 24 | +system with WIP. |
| 25 | + |
| 26 | +Whether it’s using LLMs upstream to clarify ambiguity in requirements, or |
| 27 | +downstream to automate compliance, the highest leverage points for AI are |
| 28 | +rarely in the IDE. Stop guessing where to put AI. Measure your flow, find your |
| 29 | +constraint, and apply intelligence where it matters. |
| 30 | + |
| 31 | +Key Takeaways |
| 32 | +1. System over Silo: Understand why optimizing local developer efficiency with |
| 33 | + AI often harms global system throughput. |
| 34 | +2. Constraint-Driven Adoption: Learn to prioritize AI investments based on your |
| 35 | + actual bottlenecks (e.g., Testing, Requirements) rather than market hype. |
| 36 | +3. Evidence-Based Tactics: Leave with a method to justify AI initiatives using |
| 37 | + hard data and collaborative modeling rather than vague promises of |
| 38 | + productivity. |
0 commit comments