AI-powered portfolio and proposal platform for real estate agents.
How Propwise cut AI cost by about 30%
Before
The team could see provider spend and approximate cost, but not reliably connect that spend to models, sessions, and the product workflows generating it.
paiment revealed
GPT-4o was more expensive than the job required, and image processing was making an extra roundtrip even though those images already went with the first request in the session.
What changed
They switched from GPT-4o to GPT-5.1 Chat, then dropped the redundant image-processing call so images were processed once per session.
Result
About 20% from the model switch, about 15% more from removing the extra roundtrip. Conservatively, about 30% lower AI cost overall.