Key building blocks of AI have been developed years ago, but are now being leveraged more intensively thanks to machine-learning capabilities, especially deep learning, which provides the ability to train computer programs to recognise patterns and to make predictions. The major achievements thus far are in greatly improving the ability of computers to process natural language (enabling voice interfaces) and to recognise elements in an image (computer vision). More is expected by applying a similar approach in data analytics for decision-making rather than (advanced) controls or commands.
Deep-learning capabilities are developing now because all the key conditions are finally met, i.e. access to huge processing power (amplified by the cloud infrastructure) and access to large data sets to seek correlations.
Internet giants are set to dominate consumer services and will reap the benefits for consumer services through the platform model more than through VPAs which are merely serving as another user interface (UI)
Large OTT players combine all the necessary elements for developing AI: they are ahead technologically (direct R&D investments and/or buying start-ups), they have their own processing infrastructure and they control very large datasets. For consumer data, they are indeed almost the only ones with so many different dimensions. They are all popularising AI through their VPA solutions (on smartphones or through a dedicated appliance), even though their usage remain modest for now.
Current solutions are today far from what could be developed tomorrow. VPAs are indeed merely another UI rather than advanced AI. AI is essentially for now an enabler of speech recognition, giving them access to the same usual services, with no specific business model (except the one-off purchase of the appliance). The evolution will be to offer progressively personalised services using content, user history and data to provide differentiated answers better fitting the needs of the user, more precisely than analytics would do with their ‘best guess’ answer. This may still not transform those services in killer apps, but will reinforce the marketplace/platform approach of large OTTs, creating some switching costs/barriers to entry for newcomers and customer lock-in, in addition to the usual bundling of services around a killer app and/or device.
The jury is still out on vertical markets, but OTT players may find their way … through VPAs
The key question for other stakeholders then is to understand if there is also room for them outside the platform business of OTT players. Traditional industry can also benefit from AI, leveraging technology from specialist tools providers such as IBM’s Watson. However, they may often have insufficient data to really use AI capabilities (analytics may be just enough) and therefore take time before adopting it extensively. OTT players do not have data in pure B2B markets and are therefore unlikely to compete with internal developments of vertical players. The latter are unlikely to share them (security, confidentiality) when targeting optimisation of processes, but may partner to develop new services and associated monetisation on a case by case basis. OTTs may indeed also over time capture more data thanks to B2B2C solutions/services/apps developing around connected machines and smart services leveraging the OTT ecosystem (technical enablers like OS or app store). As a UI enabled with AI, a VPA may also prove to be interesting in the quest for data and AI in new verticals. End users will likely indeed want some continuity of service rather than ‘train’ new VPAs for each machine.
To delve deeper on this theme
Check out our last reportEn savoir plus
How could ehealth move from the innovative pilots to the next stage?
Ehealth encompasses health IT system, connected medical devices and connected care services. It holds potential to bring social costs down, with the promise lying in reduced readmission and long-term care costs, as well as improving general population health. Despite those benefits, no financing model has yet been clearly defined, neither has a single business model achieved success globally. Given current government spending cuts and growing public debt, insurers, medtech and digital players, as well as authorities are all looking into how to finance these ehealth solutions.
Smart mobility : From the development of the self-driving car to its integration in the smart city
What are the investments prospects and stakes? By definition, it is commonly accepted that automation could be framed at six levels, in regard to how an automated driving system works in the dynamic driving tasks on a sustained basis.