Building Internal AI Capabilities: How to execute A.I. at scale

AI-driven organizations will be structured differently than the organizations of yesterday. Just like companies reorganized in response to the Internet starting in the 90s, AI will force new organizational structures and departments. Don’t expect this change to happen overnight though.

At first, the new AI-driven organization may resemble a digital twin of the traditional org structure. Data and machine learning expertise spokes will be embedded into all lines of business. Each spoke will deeply understand the domain of their business group, be empowered to make decisions, and execute on their own behalf. An oversight hub would be created to harmonize the individual spokes to the needs of the greater organization. This hub will lead overarching strategy and disbursement of data resources, technology resources, and success measurement. AI talent will be embedded into the business units and the oversight hub shall direct and/or support the spokes. Each spoke shall represent every function of the business.

Each spoke will include data scientists, machine learning engineers, AI program managers, and software engineers who have domain proficiency and process expertise — with each of them understanding what data is most valuable to their line of business. 

Being embedded, they’ll mostly be self-managed to direct, build and manage AI projects for the business unit. This may include providing predictions or automating processes that support the line of business leadership objectives. The spoke will then report back into the hub to ensure global alignment to key priorities and are held accountable to performance. The spoke will seek data, technology, and budget resources for new programs. The hub will collaborate with the spoke to ensure project success.

 

The hub and spoke system allows for wholeness between strategy, resources, and execution functions so that they collaborate freely to harness the collective intelligence of the entire organization. The system also empowers lines of business to conceptualize and execute the projects that will provide the most material impact. According to the HBR report, Building the AI Powered Organization, those surveyed stated the average timeframe for this realignment is 18-36 months.

You’ll also need high-performance computing and infrastructure that supports the massive computational requirements of AI models. In the past, businesses got this by processing small data sets with large systems of record. Things have now reversed. Today, we process big data sets with minimal on-premise footprint. This is most economically provided by cloud technologies and consumption-based usage models. An operational IT backbone will need to be established. To support the development and run AI solutions, you’ll need to be able to store and access critical data, integrate AI with other applications, provide reliable operations, and ensure privacy and security. Companies must ensure their massive amounts of high-quality data is cleaned, tagged and available as AI algorithms will learn from this data.

These are key technologies that are required to start and scale AI programs:

CLOUD/GPU

AI requires high-powered computational infrastructure. The parallelism offered in GPUs makes them 100X faster than the traditional CPU and is now the gold standard to train and run models. GPUs are not cheap and purchases can run between $20,000 – $40,000 per core. So cloud-based services like Google Cloud, Amazon Web Services, or Microsoft Azure become more attractive.

END-TO-END MANAGEMENT

AI is not a standalone general intelligence application. It’s not plug-and-play either. It requires IT orchestration of servers, network, storage, and software. Ideally, you can find one interface to manage all these layers.

Want more insights into how you can scale up your internal AI efforts? Download our free whitepaper here