Per Sedihn, CTO at Proact IT Group AB
As the hype cycle around artificial intelligence (AI) intensifies, many companies have had great success in using small test cases to understand the benefits they could enjoy from the technology. However, if AI is to reach its full enterprise potential, now is the time to move beyond experimentation towards delivering real business value.
Taking the step towards fully-fledged AI programmes requires organisations to tailor their approach towards their specific business needs. It requires they step out of comfort zones and embrace new methodologies. Currently, most companies restrict themselves to looking at solutions from a Software as a Service (SaaS) viewpoint. With SaaS everything is pre-built for businesses, providing fully functional, on-demand applications that are well suited to standard functions, such as Office 365, CRM, ERP, web conferencing or Salesforce. This model does not suit itself to developing a bespoke AI platform.
Instead, a Platform as a Service (PaaS) approach provides a framework for customised software creation and delivery. It goes a step further than SaaS by providing an operating environment that includes an operating system and application services. Businesses using PaaS can concentrate on building a particular application, while a third party manages the deployment platform, software updates, storage and infrastructure on their behalf. In a PaaS approach, resources can easily be scaled up and down as business requirements demand and multiple users can access the same development application. This not only adds speed and flexibility to the process; it can also reduce costs and simplify the challenges typically associated with rapidly developing or deploying an AI application.
The PaaS basics
Launching an AI platform should be process of continual improvement. It’s important to start small and kick off at pilot stage in a test bed or sandbox to understand how the process works. This will allow you to demonstrate fast results without the confines of IT governance and processes that typically strangle such projects in their early stages. Then, if successful, you can begin building in processes and policies that ensure the project gets off the ground.
Flexibility is vital, so starting off in the public cloud will help you to build the foundations without requiring massive resource or expense. You can then make more informed decisions on adding additional resources based on the initial success of projects. Furthermore, architecting your cloud environment with multicloud capabilities will help you avoid getting locked into one provider, which otherwise can see costs spiral and won’t allow you to reap the potential benefits of other cloud choices.
Business understanding is crucial
The decision-making process for AI should be held outside of IT. You need to build a project team that understands the business and your business goals, as opposed to the project being governed by traditional IT processes, if you are to truly foster innovation and realise the full value of AI.
Before launching into an AI application or pilot process it’s important to take a step back to consider broader questions behind your strategy, such as:
- What will happen if the pilot is successful?
- How much data will we collect?
- What type of data?
- How will it be prepared / cleaned for use?
- Are there any regulatory considerations?
- What security policies do we need?
- What does good look like?
Facing these questions upfront will help you understand the requirements of the project and be better placed to shape and architect the right solution for your business. This will also inform where the project should be located and paint the picture of how it will evolve over time. Business and IT need to work together in the early stages to ensure that all practical considerations are taken into account. With these practicalities covered, you can develop systems that enable the project to ‘fail fast’ and deliver an attractive proof of concept.
Steps towards getting AI live
Many factors will impact upon how successfully and smoothly your AI project is rolled out, from business readiness through to having the right people and skills in place. The most pressing you need to focus on are as follows:
- Executive buy-in: Getting senior executives on board with your AI plans as soon as possible in the process will be instrumental to its success. They can help attract the right talent, provide the budget you need to grow and experiment, and affirm its importance to all internal stakeholders. For these reasons, it’s vital to invest time in convincing them of the value of the project and having them input into governance from the outset.
- Address your skills gap: Skilled technical employees, such as data scientists, are instrumental in getting your AI projects off the ground. As AI projects grow, you want to be sure that you have the resources in place to scale with requirements. However, there is already a shortage of AI skills and, as demand rises, this is set to become more challenging. This is why executive buy-in is essential, as they can help make the role more appealing by giving assurances about the longer-term vision of AI for the organisation.
- Aligning business goals: AI can help make better business decisions, automate services and provide superior service to customers. The need for AI applications should therefore be driven by the business, as opposed to technology and IT teams or the CIO. The business needs to brief IT on the benefits they are looking to reap as the need for AI will influence what IT infrastructure and services are required. It’s important that the two have direct lines of communication and are aligned on how needs can best be met.
- Consider Edge: If your AI project is likely to rely on sensor-based data, then it will be vital to embrace opportunities to process more data at edge locations in order to reduce bandwidth requirements and lower latency. Additionally, local data processing could limit the risk of security threats.
Collaboration with trusted partners is key
Kicking off your AI project can be intimidating, so it’s important to build a network of partners and service providers that understand and have experience in AI and machine learning. Working collaboratively with the right partners and being open to testing solutions from various partners and vendors will be vital in your path to AI success.
Proact has been providing data centre infrastructure solutions and services to enterprises for almost 25 years, and managed cloud services for more than a decade. Through our partner network we’re able to help our customers to assemble and orchestrate the technologies they need to run a high performance AI platform. This includes our collaboration with NetApp on its ONTAP AI package, which is pre-architected to provide simple adoption of NetApp storage along with Cisco networking and Nvidia GPUs. This platform uses flash to enable the super-low latencies required for intensive algorithm training workloads.
Having an expert partner network in place can help you to shape your AI environment to ensure flexibility and adaptability is built into its core. AI is a fast-moving area and the requirements for enterprise will likely change significantly over the next decade. Knowing that you can rely on external expertise to ensure your approach is current and aligned with industry best practice will provide the confidence that your AI platform is designed to deliver real business value.