UNPAK3D Podcast Ep. 20: From Hype to Production: Inside Writer’s Playbook for Deploying Enterprise AI w/ Waseem Alshikh, CTO @ Writer
Nov 26, 2025
Waseem Alshikh, Co-founder and CTO of Writer, shares insights on deploying generative AI in enterprises. He discusses the importance of predictability and domain-specific models over general hype, emphasizing their role in regulated industries. Waseem reveals how Writer's full-stack approach ensures security and performance. He presents a contrarian view on 'small' models and highlights the need for effective change management to achieve real ROI. He also introduces the concept of 'Artificial Governance Intelligence' and the rising trend of multimodality in AI interactions.
AI Snips
Chapters
Transcript
Episode notes
Train With Synthetic Data, Ground With Customer Context
- Use synthetic data for base model training and keep customer data for grounding in a separate context layer.
- Connect customer data via knowledge graphs and an orchestration layer to provide safe, contextual answers.
Design Models For Practical On-Prem Deployments
- Optimize models for deployment speed and cost, not just parameter count, so enterprises can run them on a few GPUs on-prem.
- Aim for 70B–18B parameter sweet spots to balance performance, speed, and reduced hallucination.
Sell Use Cases, Not Model Names
- Focus sales on concrete use cases and time-to-value rather than model brand names.
- Treat customers as partners and plan change management to unlock real ROI from AI.