Par Varadharajan S
Presales Head for Infrastructure solutions,
Wipro HOLMES Artificial Intelligence Platform
Par Tapati Bandopadhyay
Practice Head, Wipro HOLMES Artificial Intelligence Platform
retour au sommaire
COGNITIVE HYPER-AUTOMATION CAN HELP YOUR BUSINESS
All too often, business and IT leaders have encountered situations where a business loss has been traced back to a process failure which in turn has been traced back to an IT problem such as an application outage. Behind the scenes, multiple teams scramble to resolve the issue,
but they invariably work in silos and the precious initial hours are wasted on diagnostics and subsequent finger-pointing – “This is not a network problem!”, or “My app server is running fine, so the problem is not with us!”, and so on.
The unfortunate reality is: Most of the high-priority incidents/outages that result in significant business losses (with visibility ensuing for all wrong reasons, from top leadership and external stakeholders), do not happen in siloes. For example, typical issues such as application time-outs often happen as a combinatorial impact of database server loads and network fluctuations.
It is therefore imperative to have an integrated and real time view across multiple IT towers.
But who really is equipped to understand the entire landscape and provide a view in time to prevent such major disruptions? This is achievable through a cognitive horizontal capability
and enterprises must follow a step-by-step approach in building the capability that works
for their specific context.
BUT IT IS EASIER SAID THAN DONE
Enterprises that have been experimenting with automation and AI for years now have realized significant outcomes. Today, automation is recognized as a strategic, not just a technology,
lever. But building an automation strategy and roadmap is proving to be very challenging and
SO HOW DO YOU BUILD ORDER INTO THIS CHAOS?
By taking it one step at a time.
Step 1: Establish reliable monitoring mechanisms to enable event correlation
A critical ingredient for automation is to ensure timely alerts, flags and exceptions, when something goes wrong. This critical function is performed by end-to-end monitoring tools and many come with advanced event management and analytical capabilities that can enable us to eliminate, or at least significantly reduce, noise in the data and false positives. But a direct automated approach triggered from event management consoles may just end up creating more silos! That’s why it’s important to aim for actionable alerts from robust tools, ones that actually remediate the causes of anomalies. This approach helps in maintaining a trail of actions performed to ensure future reference, knowledge capture and service quality assurance.
Step 2: Enable self-heal/self-help with script-based automation to improve user experience
The second step is to look for ways to bring in further efficiencies to improve overall user experience. Self-help and self-service solutions not only help users in solving certain issues much faster but also give users a high degree of satisfaction.
Step 3: Implement dynamic, rules and threshold-based orchestration to solve routine issues
In a world without automation, it was almost impossible to avoid impacts on business as service engineers scrambled to get to tasks pending in their queue. The nature of the tasks could be something as mundane as restarting a service or cleaning disk space. With automation in
the picture, routine tasks can be taken care of by deploying an effective dynamic, rule based orchestration engine that is integrated with the IT service management solutions. It is necessary here to start off with common issues to encourage faster adoption by the teams on ground.
Step 4: Bring in cognitive & machine learning capabilities – start with supervised learning models
Cognitive systems enable context-aware interactions between the system and the end user.
It is important to start this journey with supervised learning models to ensure a high level of accuracy. For example, intelligent conversational interfaces for service desk operations are easy-to-conceive and deployable solutions with cognitive capability. These solutions can be deployed on top of any of the existing chat and communication channels such as Skype. Both IT and Business use cases can be made available through cognitive bots to give a seamless experience to end users.
Step 5: Deploy unsupervised machine learning algorithms e.g. deep learning, ANN based auto-optimizing intelligent automation
Deep learning based neural networks and recurrent neural networks, with non-linear functions and multiple hidden layers are capable of discovering patterns and classifier models that are more accurate and closer to reality but may have been hidden from the human eye. In a typical
IT operations environment, applications of these algorithms can significantly reduce response times, turnaround times (TATs) and cycle times, due to more accurate classifications, significant reductions in hops and reworks in resolution attempts.
Taking one step at a time, starting small and staying focused on the end-goal of hyper-efficiency and digital experience delivery, is the most effective, tried and tested approach for enterprises that are early movers in cognitive automation.
Meet Wipro at AI Paris !
retour au sommaire
Idriss Tom • Inscription Conférence
+33 (0)1 84 83 02 95
AI Paris est un congrès réalisé par Corp Agency, organisateur d’expositions et de conférences de haut niveau dans le domaine des nouvelles technologies.