When it comes to enterprise software, operational efficiency trumps bells and whistles every time.
Why then, when it comes to optimising those applications, particularly for the Internet of Things (IoT) are enterprises trying to boil the ocean?
A recent GlobalData survey of 1,000 IoT professionals revealed a heavy reliance on hardcore business intelligence software.
Forty percent of those surveyed ranked business intelligence platforms well above all other means of analysing data.
More broadly throughout companies, major do-it-all business intelligence software platforms have already given way to numerous smaller, more discrete ways of deriving value from enterprise data, be that a direct SQL query, a predictive data modeller, an auto-generated data discovery visualisation, or a live, interactive executive dashboard.
The reasons for this are simple: business intelligence software is reactionary and static.
Its users rely heavily upon basic reporting mechanisms that, in turn, rely heavily upon laborious queries and reports – a very costly venture to both build and maintain.
This reluctance to follow the broader market away from business intelligence platforms within IoT is concerning, given a subtle shift noted in the same survey concerning when, during its lifecycle, an IoT deployment fails.
In 2016, no failures were noted post deployment.
In 2017, however, that number shot up to twelve percent. And what was the number one reason for both failures and abandoned projects?
Obviously there is a significant mismatch between opportunity and expectation at play here.
The Internet of Things is supposed to let users anticipate and respond to the unknown, and to build transparent business systems capable of evolving to meet changing needs.
Why then, are IoT practitioners relying on centralised analytics reporting? If IoT analytics can only point out what happened in the past, how can users expect to anticipate the future?
Fortunately, there is an answer to these questions already at hand in the form of artificial intelligence (AI).
With even the most simple AI machine learning framework and model at the ready, IoT practitioners can solve two pressing problems: detecting anomalies and predicting desired outcomes.
The same technology can automate the distribution of cucumbers in rural Japan or automate the optimisation of a fleet of trucks in North America (both real use cases).
No expensive dashboards are required.
The survey supports this idea of using IoT to solve discrete problems, showing that 43 percent of IoT buyers see the improvement of operational efficiencies as the number one reason for investing in IoT.
It also showed that AI is already viewed as the best tool at hand to centrally automate and optimise business processes.
The problem, however, lies within the idea of centralisation.
IoT happens, not within a business information report, or a data warehouse, or within an all-encompassing predictive business model.
IoT deployments should employ tools like machine learning frameworks, not centrally, but instead at the edge, close to the device itself.
And like today’s enterprise software, those analytics endeavours should be brief and to the point, and focused on solving specific challenges.
This is the same idea espoused by some branches of AI itself such as deep learning, where the combination of numerous, smaller decision-making algorithms can give rise to a larger, seemingly intelligent system.
The idea is simple — don’t try to build an expensive monolithic analytics system centrally.
Instead seek global visibility while optimising locally via discrete, AI-driven outcomes.
This approach will not solve the full set of potential problems, but it will have an impact on the business and affordably, helping to prove the value of IoT one problem at a time.