The attached image samples are taken from some tests that I ran today, comparing Topaz Sharpen AI with the deconvolution sharpening in Photoshop CS6 ACR. This is a reduced size JPEG of the whole image. The original is a full frame RAW file, captured with a Nikon D800E and Nikkor 24-70mm f/2.8G. This lens has significant field curvature at 24mm, and I suspect that's the reason why some areas of the frame are soft, even though the center is reasonably sharp.
Here are the descriptions of the 100% crops, from left to right. The first was processed with no sharpening in ACR or Photoshop. The second image was sharpened in ACR (35, 0.7, 100, 0). For the third one, no ACR sharpening, and Topaz Sharpen AI was used in Stabilize mode (0.4, 0.3, 0). The fourth one was converted in Capture One 12.0.2 with all sharpening disabled, and then exported to Photoshop, where it received the same treatment as the third image. I didn't attempt to perfectly match the tonal balance and colors of the third and fourth images - but there are other obvious differences.
To my eyes, the Topaz software is extracting (or synthesizing?) image detail that I could not obtain with the sharpening tools in Photoshop and ACR. I did try other settings and other tools in the Adobe product. I am showing the results from the Stabilize mode in Topaz, because it produced a better result than the Sharpen and Focus modes, for this particular shot.
Because Topaz Sharpen AI works in a different way than other methods, it might require changes to your workflow. I suspect that it may be wise to apply it after adjusting highlights, shadows, tones, colors, blurring, and noise reduction. Since the software employs pattern recognition, you want to prepare its input data so that the desired patterns are recognizable, and the undesired ones (e.g., noise) are attenuated.
Topaz Sharpen AI is not a one-click panacea. It may take some trial-and-error to find the appropriate mode and settings for an image. Like any sharpening tool, bad settings can produce artifacts and ugly, crunchy results. It won't save shots that have tragic levels of blur. Masking may still be needed to control the level of correction in different areas of the frame. But I am finding that, on some images, this product can produce remarkable improvements.