Tellme Networks is a hosted service for American Airlines, Fedex, AT&T and Sprint. They automate your directory assistance inquiries [what city? response. what name? response. The number is….]. Microsoft plans to spend $800 million to own this business and integrate it with Jeff Raikes' Business Division. 

Brockmann rating on this deal is as follows:

Strategic fit [4/5]. Sounds great on paper. MSFT needs to play in advanced input methods and voice activated computers (didn't Gene Roddenberry invent that as part of the Star Trek franchise?) is something they feel strongly about. 

Timing [0/5]. Voice recognition is a massively processor-intensive function. Doing a hosted service makes tons of sense because it exploits dynamic processor allocation through P2P technology, and really balances the capacity available to do the service with the demand for the service, which of course is pretty dynamic and lumpy. Don't most people call the American Airlines arrivals and departures telephone number whenever there's a storm? Timing is irrelevant (that's why it's a zero) because this technology will not move out of the biga– server farms for some time.

Customer demand [0/5]. Users don't care. Of all the phone calls that I have made in the past five months, (probably over 500), only one used a voice recognition application instead of the usual auto attendant directory spell-by-name game. Furthermore, in studies performed at my former employers over half of call center managers and communications managers had no plans for voice recognition. Furthermore, Tellme customers are already Microsoft customers.

Potential [1/5]. Given the degradation in voice quality these past few years attributable to the mobile network service which makes it harder for a reliable utterance capture, the high processor demands of the utterance processing algorithm and the low demand for voice recognition in enterprise or in call centers, will keep this technology in the hosted approach for some time. Compounding the wider adoption of the capability is that it can only process limited utterances. You and I speak with emotion, but these are not reflected in the key punching we might do to prepare an extraordinary email. Generally, people would rather speak to a person than to a machine.

Overall: 5/20 = 20% = F grade.

There is value in integrating this application set with lots of other Microsoft Consumer properties – XBox service, LiveSearch, MSN, but how it ties to UC and Live Communications Server will be a disappointment for some time. 

Microsoft must be in its late stage. Five years ago, Intel bought up VPN companies and even the voice DSP company Dialogics, ran them for a time and sold them for a fraction of the purchase price. I fear MSFT is going through the same thinking, and this deal is a sign of that.

This post has already been read 0 times!