A summary of the online debate that I mentioned last week has appeared in Speech Technology magazine online: http://www.speechtechmag.com/issues/industry/12716-1.html.
Of most interest is the participant poll. A good number of attendees (60% of those who answered the question) already have speech in their contact center and presumably know what they are talking about. We aren't told the absolute numbers (although the granularity of response percentages suggests there were a decent number) but some interesting data points emerge:
I don't think this is unhealthy. On the second point, there's clearly still a big opportunity for innovators in the integration and implementation spaces -- design tools that drive down the time taken to build speech apps, and solutions that better cover integration with data backends and the rest of the call center. Plus lower TCO has really got to start happening all round. (It is always good, too, to have confirmation that the limitations of the underlying recognition technology are only a small part of the overall equation.)
On the first point -- cost saving vs customer retention -- I think this means that fewer companies will be interested in investing more time in great voice user interfaces, and will stick to simple functional dialogs, probably directed, menu-driven style, because that will meet their goals, cheat-sheets or not. That's unfortunate. But it plays against the 20% figure of customer reluctance to use automated services -- reducing that reluctance factor can only happen if a broader investment is made in better voice UIs. So driving down the cost of good (beyond just acceptable) spoken dialog deployments should really be a goal for us all if the industry is to overcome the barriers that these figures tell us are in the way.
Update (6 Jan 2006): the debate itself was recorded and is available here.