Speech recognition makes everything better. Add speech-based input to vanilla calendaring, to-do lists and note taking, and you have a great office productivity tool. Turn this into a service available on cell phones, and you have the makings of an exciting start-up or two. Several of these services can transcribe voicemails and on-the-go voice memos that are then emailed. Great idea. Here’s one that’s even better: transcribing and emailing agent conversations back to the customer.
Contact centers routinely record conversations (“this conversation may be monitored…”) for agent training purposes, as well as for compliance and other legal requirements. In my last post, I described how contact centers have even started mining these conversations for specific keywords that match up with customer purchasing preferences.
This is all internally helpful to the contact center, but can’t speech recording and even automated transcription technology be made available directly to the customer?
As we dialogue with agents, there’s lots of useful information and thinking-out-loud brainstorming that we may not remember or have time to write down. I’ve had more than a few calls with bank, insurance, mutual fund and credit card contact centers where my note-taking couldn’t keep up with the fast flowing conversations.
Speech recognition technology would seem to hold the answer. In its current state, it’s very, very good at detecting isolated words. We experience discrete word-spotting when navigating through speech-based menus (“press 1 or say yes…”). And speech software has gotten much better at spotting keywords within a continuous voice stream. That’s the flavor of speech rec that contact centers use when mining conversations. Another speech-rec application in the contact center: voice-based routing.
Translating continuous speech to text is also much improved, but still not completely fool-proof.
Sounds likes crowd sourcing could help! And in fact, to improve accuracy, one of the start-ups supplements its automated speech recognition software with a staff of quality assurance workers.
I would be willing to pay extra for a transcription of my contact center calls.
But even if speech recognition at this level is impractical for cost reasons, I’d settle for a service that should be achievable—receiving a secure URL to a media server that contains my digitized conversations.
I suspect there are security and privacy concerns (read HIPAA, Graham-Leach, and Sarbanes-Oxley) that would make providing external customer access to voice records difficult from a legal perspective.
Still, it seems that more contact centers should be offering this service in areas where financial and personal data is not the issue—say, a tech support call. Can anyone tell me why this isn’t being done?
