Today is the day. After months of posting blogs that explain all the technologies and concepts we are working on here at Market Logic, I can proudly announce the official release of our cognitive marketing assistant. It’s been an exciting time building this, to say the least, and with new technology breakthroughs every 2-3 days, an especially rewarding one.
That being said, let’s take a look at the business motivation for our new marketing assistant. Talks with one of our largest clients about the idea of upgrading cognitive search dossiers to a conversational interface focused on three business needs:
With these needs acknowledged, the stage was set: build a conversational assistant that understands what the marketer wants to do, recommends the best practice to get the job done, and delivers relevant insights at the right time, every step of the way. Our assistant would be able to learn to run a business process in a constantly optimized form and sequence, to transfer machine learning models from one domain to another, and to automatically learn to identify intent without supervision. It would have to answer complex, multistep questions, while emulating a friendly and supportive friend or coach. It would also have a self-optimizing way of running a process from start to finish, while proactively recommending information and insights based on the topics at hand, even if the user isn’t actively searching for them. Easy!
As you know (assuming you’ve read my other blogs), the requirements we defined were not solvable with off the shelf technologies from any of the many chatbot technologies, so we had to build from scratch. Our new assistant isn’t a simple task guidance chatbot that performs discrete lookups, like booking an airline ticket or ordering a pizza – it has to intelligently and efficiently guide a user through, let’s face it, corporate process monsters that can easily run for several weeks. As a result, the logical technology building blocks we’ve created represent the latest and greatest from the artificial intelligence academic research community, while seamlessly merging all the NLP and understanding technology we already built into our insights platform architecture stack.
We solved these challenges by creating several technologies that sit behind a beautifully designed Assistant UI that includes: a conversational area where you can talk to the assistant, another screen area that presents the answers and relevant information pieces, and a third, multipurpose area that can be switched between previous tasks and conversations so you can jump from one specific “process project” to another, WITH a task-related notebook to document all relevant information along the way.
And yes, it’s also mobile, so your marketing assistant lives wherever you need it to be, in your pocket and on your desktop. That means it adapts the user experience on a mobile device to make sure the answers are so relevant you don’t need several results, just the one, and then drafts your ideas with voice, no typing.
Let me wrap up today by saying we also had to make it really easy to actually design each client’s marketing processes so business consultants can easily adapt your assistant to suit your processes and sub-tasks without going back to development. That means my team can focus on the fun work of making it faster, more effective, more efficient and more engaging to use.