I've never delivered photos direct from the camera to a client, whether film or digital. So my choice of whether, when using a digital camera, to capture and store raw data vs rendered images (TIFF or JPEG) in-camera has been/is wholly based on what my cameras are capable of and the workflow I need to obtain it.
My first several digital cameras were only capable of storing rendered images in JPEG format. I learned how to set the cameras to obtain what I wanted, with the various notions of whether I wanted/needed the out of camera results to be 'finished' renderings or renderings suitable for further image processing. (Note that I had been doing digital image processing for at least 15 years prior to there being an affordable digital camera to work with, so I knew at least the minimums of what to expect from that.)
When I obtained the first camera I owned that included the ability to store raw data, I spent time learning how to do raw processing. To me, it represented a simplification of the shooting workflow (no need to choose different settings for different notions of what I needed out of the camera) at the cost of additional time and effort in rendering post facto. This seemed a worthwhile trade off—we are typically pressed for time when making exposures moreso than when processing what we've captured. The early raw processing tools were somewhat slow and clumsy to use, so it was often a bit of an annoyance, but even so results improved due to there being more data to work with: greater headroom and broader dynamic range, more adjustability, etc. The cost and downside were the speed of processing and the amount of storage require to support saving the data in intermediate steps presented serious constraints.
Then the tools improved in quantum leaps. Computers doubled, redoubled, redoubled again and again in power, reducing compute time to a tiny fraction of what it was at the beginning. Storage prices dropped by orders of magnitude over a very short time. Parametric editing without having to save chains of intermediate work reduced storage requirements wholesale as well. The raw processing applications now automated a huge amount of the work needed to the point where I could actually get away with many image rendering needs simply by running them through the image processing workflow at the defaults. The constraints in time, space, and cost of post facto rendering of photographs have become almost trivial compared to what they once where, to the point where it is in most cases of no consequence to the workflow.
The end result of this is that I tend to save exposures as raw data only and do all the rendering in post facto processing. This is the most similar to my film camera workflow, where my a priori decisions are limited to exposure, focus, framing, and timing ... all the other work is post facto. That said, I still have a couple of cameras that require more time, and others that only store rendered data. For example, the Light L16 camera's rendering app, Lumen, allows post-facto choosing of simulated aperture and some degrees of freedom in determining best focus plane on original capture data, and it takes a good bit of time to take the settings you've applied and render out even a raw file processable in other image processing apps.
So ... The choice of whether to save captures as raw or JPEG should be driven by the camera capabilities, the needs of the image you are trying to produce, your skills in using the available tools, and the timeline (and possibly other requirements) of the client you are supplying the images too. Making a pronouncement of doing something one way in all cases and all situations ignores these criteria and choices, and is likely an unhelpful thing to do.
G
