AI for Engineering Is Leaving the Demo Phase

About this episode

Text-to-CAD. Autonomous simulation. Agentic workflows. AI copilots for PLM. Engineering teams “10x faster.”

Most of it still sounds like science fiction.

But what happens when you put two founders building real AI-native engineering software in the same conversation?

In this episode of AI Across the Product Lifecycle, Michael Finocchiaro speaks with Pradyut, co-founder of Bild, and Martin Bielicki, co-founder and CEO of Bench, about what AI is actually changing in engineering software, CAD data, simulation, product development, and manufacturing workflows.

Bild is building CAD data management that connects engineering to manufacturing. Bench is building an AI orchestration layer across CAD, simulation, PLM, and beyond.

The discussion cuts through the hype:

AI is already changing how startups code. QA and validation are becoming the bottleneck. Prompting matters less than context. Frontier model cost is becoming a real burn-rate issue. And engineering AI will not move as fast as software AI because CAD, simulation, manufacturing, sourcing, and PLM are different technical worlds.

The real unlock is not “AI replacing engineers.”

Timeline

00:00 – Introduction: Bild, Bench, and AI across engineering
00:29 – Pradyut introduces Bild
00:54 – Martin introduces Bench
01:16 – The OpenAI moment
01:55 – Bench was created because of the LLM breakthrough
02:25 – Bild’s early exposure to DALL·E
03:33 – How AI changed startup coding
04:01 – Cursor, Claude Code, Slack, Graphite, and same-day delivery
05:45 – Multi-agent development
06:37 – Why “looking good” is becoming commoditized
07:28 – Why old software stacks limit AI innovation
08:21 – Prompting vs context
09:53 – From prompts to loops
10:40 – Frontier LLM costs
11:20 – Token costs as the new AWS->12:15 – AI spend caps and productivity measurement
13:45 – Cheaper models and model routing
14:39 – Right model, right task
15:39 – AI and engineering org structure
16:19 – QA, validation, and human-in-the-loop checks
17:58 – How AI may reorganize hardware teams
18:50 – Multi-agent coding conflicts
21:16 – Where AI lives inside the product stack
21:42 – Bench: AI for context, planning, and judgment
22:43 – Bild: opt-in AI for CAD data and IP boundaries
24:19 – When will engineering have its OpenAI moment?
25:04 – Why engineering AI evolves use case by use case
26:35 – Faster adoption in consumer products?
27:04 – From text-to-CAD to DFM and manufacturability
29:09 – The coming CDFAM AI demo wave
30:05 – Advice for young engineers
30:43 – Don’t compete with agents. Build differentiated skills.
33:05 – Creativity roles and AI in physical sciences
35:06 – Why top engineers become more valuable
35:57 – Digital transformation reality check
37:21 – Prints, redlines, and physical sign-offs are still alive
39:31 – Can startups move faster than legacy vendors?
40:10 – Big OEMs asking for AI engineering visions
41:15 – Buying AI vs buying value
42:50 – Why transformation programs route back to incumbents
43:37 – Build vs Windchill
45:30 – Startup visibility vs legacy vendors
47:56 – Capability checkboxes vs real user experience
49:13 – AI agents for CAD and CAE workflows
49:33 – End-to-end orchestration and organizational readiness
51:49 – Keeping skilled engineers in the loop
52:26 – Trust but verify for hardware AI
53:39 – Where to meet Bild and Bench
55:31 – Closing remarks

Featuring Pradyut of Bild and Martin Bielicki of Bench. Hosted by Michael Finocchiaro.

#AI #EngineeringSoftware #CAD #PLM #Simulation #Manufacturing #DigitalThread #IndustrialAI #HardwareEngineering #Startups