Last week I broke my arm and needed surgery to put the pieces back together again. That led me to re-explore some technologies I'd looked at years ago — speech recognition technologies. Given that it was my arm and not my credit cards that were injured, I ordered a new copy of Dragon Naturally Speaking (DNS) v.9 and had it within a day. I had heard others talk about how effective it was (for example, see David Pogue's "New York Times" review of DNS last summer), but I was skeptical. Earlier versions (much earlier) of the Dragon products were interesting but frustrating. So, too, was IBM's ViaVoice product. Then again, I noted that my general practitioner's office used speech recognition to transcribe the doctor's notes of his meeting with me. So, maybe these tools were ready for prime time.

This wasn't an issue of intellectual curiosity. Impending deadlines for research materials coupled with a useless (and painful) left hand drove me to re-evaluate working with speech recognition technology. DNS installed smoothly and quickly. Training took about 20 minutes and then it took DNS another hour to go through all my documents to understand my writing style. Was I surprised at how effective DNS had become! (Maybe it was necessity driving my impression — after all, it was still making mistakes that I had to correct.) But in the end, I compared my productivity without this tool with my productivity with it and realized I'd be lost right now without it. (I'm dictating this posting using DNS.)

This leads me to think about policies for standardization and user-installed software. I installed DNS on my company-supplied system — probably a violation of company policy. I never bothered calling IT support — after all, I knew they had better things to do than test whether DNS worked properly with all corporate-approved applications used by everybody in all departments. For all I know, DNS is incompatible with our corporate CRM system, but so is a useless left hand! (I am an analyst, not someone in sales and marketing, so CRM system support is irrelevant to me.)

Clients often call asking for estimates of total cost of ownership (TCO) and return on investment for a variety of technologies. DNS cost me $99. I have no idea what the true TCO of this product is (our TCO analysts are not going to like this comment), and I'm confident nobody's ever going to try to figure it out. Return on investment? It was a smart move given that I can now return to writing!

Given the surgical repair I had, it will probably be at least a couple of weeks before I can type again with my left hand. I wonder if I'll stay with DNS once the splints are off.

What's your policy on products like this? Have you assigned individuals to go off and test these products? Do you have a set of standards under which you will deploy these products? Do you block individuals from making this type of decision on their own? Do you bother calculating costs and returns? Or do you just let it happen, without explicit management?