In our first piece on chat bots, we outlined four big challenges, one of which was the automation of data science. Today, I’m going to do a deeper dive on automation, and look at the return of portals.

First time around (in the nineties), portals were meant to bring together the sudden deluge of information that was hitting us via the Internet. Companies like Yahoo, AltaVista, and AOL would show us relevant news, weather and various snippets of data. Of course, these gateways to the web were also filtering and advertising selected content, and the space became a battlefield because of the quantity of visitors passing through, and the value of that space to advertisers. At the same time, large organisations could see the benefits of offering a collection of selected services to their enterprise (and partners), and the major software vendors rapidly moved to develop portals and portal-building tools that incorporated their own eco-systems – companies such as Salesforce, SAP, Oracle, Microsoft and IBM. End-users quickly learned to juggle multiple portals, and for a brief period there was talk about portals of portals, creating a meta-layer above the rapidly growing collection of systems that were intended to make our life easier. Instead, we settled on Google as the gateway to the un-managed, un-discovered, didn’t-get-around-to-bookmarking-it portion of the web, and a small number of enterprise or vertically focused portals that fit our corporate culture, lifestyle and workflow.

The battle to own that valuable gateway real-estate has continued though, within and without the enterprise. Google has a strong position in the web, but it’s not unassailable. Bing made a determined push to regain share in 2009, introducing a host of new concepts and a much prettier front-end, to the point where it had progressed to a 20% share of the US desktop search market by 2015. Within enterprises, the big software vendors added more functionality to their portal offerings, making them appear bloated, especially in comparison to startups like Backbase and Jahia that launched lean portals based on Service Oriented Architectures (SOA) and web-based technology such as AJAX and RESTful APIs. Recent portal trends have seen an increased focus on business process integration, personalisation, better user experiences, authoring, integration with social, and mobile availability.

Users, though, have their own ideas about what they want. When I worked at Microsoft (2008-2013), there was a proliferation of internal SharePoint portals, but the one app that was indispensable to most of my colleagues was Lync (later merged with Skype). The ability to use Instant Message (IM) to communicate in near-real time, conversationally, either one-one or in a group, was very powerful. Combined with presence (so you could time your call for when the person was actually there), and the capability for VOIP, webinars and conference calls, and you had something special. The rise of WeChat, WhatsApp, Snapchat, Slack, Digsby, Adium, Meebo et al, plus Microsoft’s acquisition of Yammer, and Apple’s own message service – and the focus it received during this week’s WWDC show -  shows the growing popularity and importance of these tools. Even so, when I need to buy a birthday present for my brother, or I want to check out a new competitor’s website, I still reach for my browser and do a search, interacting with the resulting website to perform a transaction, communicate, or participate in a process. What if I didn’t need to do that? What if I could type a request into my IM system and Cortana executed the transaction?  Or if I could ask Siri to perform a complex task while I was driving my car (or typing on my MacOS!)?  Welcome to the new portal - the AI chatbot.

Which brings us back to where we started. The surge in AI Chatbot activity is because chatbots are the new portal, and will (very soon) control access to the internet and apps that you see – or at the very least the things you’re offered in the first instance. I recently talked to Mark McNally, CEO of UBIX, about the capabilities required to power AI chat bots under the covers – not from a Natural Language perspective, but in terms of raw intelligence to perform the necessary tasks. UBIX are focused exclusively on automating data science. They have an AI engine (called UBIX Auto-Curious) along with a product they call the UBIX Advanced Insight Portal (AIP). I asked him to explain what that is:

“The Advanced Insight Portal is essentially a way of giving business users the ability to ask and answer complex questions, and generate insights, without having to wait weeks for a data scientist to become available to help them.”

Mark gave me an example where they had worked with a client, and after analyzing the top ten or so questions that the customers had of their data, were able to build a semantic map which linked to the UBIX Advanced Insight Portal that was then able to answer thousands of questions, without any additional input from the subject matter experts.

Mark explained, “If you take a basic question like ‘Why is this oil pump failing?’ there are potentially hundreds of factors and parameters that could be attached to the question that would produce thousands of distinct analytic queries. You might want to know if abrasion is a contributing factor, or if weather in the Permian Basin correlates with recent failures, or if fracking has increased the amount of debris in the pump. Each of those questions potentially requires access to new data sources, or changes the basic question enough that most companies would require a lengthy round-trip to a specialized team that can build such queries. What UBIX does is automate the data science portion – it understands the meaning and context of the data, and what additional information makes sense to be combined – and it can go and find those items for you. What’s more, it learns by your choices, so that it can make better suggestions for you, or the next person that asks a similar question.”

Now we’re getting to the root of the AI Bot challenge. Once the ‘simple’ problems of understanding (NLP) and producing a reply (NLG) have been conquered – with apologies to everyone working in that field [it’s damned hard, but you seem to be getting close!] , the real issue is: can we ask meaningful questions and get a reasonable answer? Or at least, can we get a sensible next step along the process to the eventual answer. If I ask “How do I cure testicular cancer?” I’m probably not going to get a meaningful response. However, by utilising a system like UBIX with its Advanced Insight Portal, what we might get is an answer such as “Would you like to look at causal factors that have been demonstrated for other types of cancer?”, and it may offer access to specific databases of gene information that could be searched for patterns. In fact, that's exactly what UBIX have done with one of their (currently confidential) academic customers – reducing the time it takes to find gene mutations in a strain of cancer from 18 weeks to less than 3 days.

As expected, Apple used WWDC to announce the opening of Siri to third party apps, and Siri’s availability on the Mac. In the same week, Microsoft’s acquisition of LinkedIn was a surprise, but makes a lot of sense if you think about the services Cortana could now start adding, with deep integration to the business social-network of choice for 400 million individuals: “Find me anyone with Python development skills, who has a qualification in Robotics, lives in Seattle and is only one connection away from me”, or “Show me the CTO’s that are talking about blockchain in financial services, with links to Bank of America.” I predict that Cortana+LinkedIn will be a very powerful combination, even if Microsoft did pay too much for it ;)

Make no mistake, AI chatbots are the new portals, for home and work. Third party integrations are good, a logical next step, and the LinkedIn acquisition makes a lot of sense in that light, but the revolutionary changes won’t happen until we automate more of the complex data science, and give everyone access to deep insight. I asked Mark McNally why his company was named UBIX, and he said that when they launched the company several years ago they had a belief that to make a real difference AI technology had to be ubiquitous – hence UBIX (with a nod to Phillip K Dick’s novel Ubik, too). That world seems to be getting much closer! Next time out, we will look at the how AI uses narrative... let me tell you a story...