# 2026 State of AI Engineering — Barr Yaron, Amplify Partners

## Executive summary

The state of AI engineering is characterized by rapid maturity and increased complexity. Survey data from 1,048 respondents indicates that while open-weight models augment closed systems, the primary drivers for model choice are quality, agentic capabilities (like tool calling), and cost. Cost has become a 'first-class engineering constraint,' forcing teams to manage usage carefully. Furthermore, agents are rapidly evolving from summarization tools to systems with write access, necessitating robust control layers and sophisticated evaluation (eval) processes.

## Key takeaways

- AI Experience is Democratizing: The AI engineering workforce is maturing quickly; the median new engineer has nearly as much AI experience as a 10-year software veteran, indicating that AI skills are becoming foundational to modern development.
- Cost is a Primary Constraint: Three out of four respondents report adjusting their AI usage based on cost, establishing 'cost' as a first-class engineering constraint alongside quality and capability.
- Agents are Taking Action: Agentic workflows have shifted significantly: they are no longer limited to reading or summarizing, but are increasingly taking actions inside systems. Write access for agents has increased dramatically (from 52% to 89%).
- Evaluation Remains the Biggest Challenge: Across all layers of the stack, 'eval' (evaluation) remains the number one biggest challenge reported by engineers.

## Technical details

- Model Strategy: Teams are not choosing between open-weight and closed models; rather, they are mixing and matching. Open-weight models primarily serve as an augmentation to existing closed model usage.
- Model Selection Criteria: When selecting a model, the top considerations are quality, agentic capabilities (tool calling), and cost, with reliability being less frequently cited as a primary driver.
- Build vs. Buy Stack Decisions: Inference and model serving is the layer most often 'bought' (off-the-shelf), while prompt management, RAG, and eval tend to be built in-house due to specialized product logic.
- Modalities Adoption: While audio shows the highest 'intent to adopt' (56%), image generation saw a significant jump, with the share of respondents using generative AI for images doubling from 18% to 36%.

## Practical implications

- Prioritize robust, multi-stage evaluation (eval) pipelines as the primary engineering bottleneck.
- Design systems with explicit cost monitoring and usage tracking to manage AI expenditure effectively.
- Focus on building internal tooling for prompt management and product logic, while leveraging external services for core inference infrastructure.
- Implement granular permission controls and guardrails when integrating agents that have write access to production systems.

## Topics

AI Engineering, LLMs, Agentic Workflows, Model Deployment, Software Architecture, AI Engineering Survey (2026), Amplify Partners

Source: https://www.youtube.com/watch?v=RGe6EjucbzI