Enterprise AI: Why organizations fail in scaling AI and How to do it right?

Organizations know that scaling AI is a must to extract optimal value from their investments. To scale AI, they need to cross a few barriers and address key structural issues.

According to market research firm IDC, spending on cognitive and AI systems globally will reach $57.6 billion in 2021. 2019 has been a good year for AI with many business thinking of it more from a strategy point of view than mere technology adoption. Organizations convinced with the benefits of AI are looking for much deeper and widespread use of AI. They want to scale AI, yet the reality is that many organizations are failing to do it.

A report by Accenture finds that 84% of C-suite executives feel the need to scale AI to achieve their growth objectives, and 76% acknowledge they struggle when it comes to scaling AI in their organizations.

Scaling AI

What does Scaling AI mean and why is it challenging?

Many organizations have been successful in AI pilots or use cases. But are finding it hard to move towards scaling AI. It is because scaling AI means deploying AI across the organization and realize its full potential. This requires the transformation of the operating model of a business, a series of top-down and bottom-up actions and commitment of a big budget.

Why the disappointing results? What is stopping businesses from scaling AI across their organizations and how to cross those barriers?

Many factors are at play here –

1. Insufficient data infrastructure – Scaling AI in enterprises would require the collection of required data and building a robust data infrastructure. The existing data infrastructure in organizations is designed for basic analytics or data storage purposes and not for AI deployment at scale. Many businesses have been only collecting structured data that can be used for extracting business intelligence. AI can process both structured and unstructured data but organizations must first build the right Data transformation pipelines capacities to collect, transform, annotate and then store the data for downstream analytics.  Each of those processes not including that required for AI needs different data infrastructure, workflows and computing capacity for storing, processing and analyzing huge volumes of data.

2. Dedicated Resources – Building on the previous point, the activities involved with Data processing and Analysis almost always require a separate but parallel infrastructure from the core infrastructure that is supporting a product platform, a service platform or both. This increases the need for allocating resources both technical and human to ensure optimal output and productivity. For the companies that have limited IT budgets but a growing demand for Data & AI needs, this additional cost always comes at the cost of their core products or services.

3. Siloed work culture – Scaling AI in enterprises requires coming together of business, technology and data. The organizational data needs to be unlocked to ensure its free flow across the organization. This cannot happen in a siloed work culture and organizations must build an interdisciplinary team to drive AI in the organizations.

4. Multiple Data Streams – A business is a combination of multiple Workstreams, Teams, Applications, Clients, Products & Services. Each of these areas is capable of generating data that have an impact on other areas laterally. To effectively leverage cross-functional data and derive insights to power AI, is another hurdle that businesses need to overcome.

5. Effort to Value Proposition – Getting an AI project live takes time, at times several months. Longer periods mean companies or teams entrusted with the responsibilities losing momentum and focus. Organizations not only need to find agile ways to implement but also need to build a mechanism to monitor projects, fix accountability, focus on small but incremental gains in an Agile manner and incentivize progress.

6. No AI Governance Model – Scaling AI in an enterprise also requires creating an AI governance model. This would require C-suite buy-in, alignment with the business strategy and structuring of roles and responsibilities for execution. A hub-and-spoke model wherein the hub takes responsibilities for strategy & planning and small teams in various departments across organizations taking care of execution can be an efficient way to manage your enterprise AI implementation.

7. Resource Constraints – Getting high skilled resources in new technology initiatives is a challenge. While the resource pool is increasing, it is still highly competitive and expensive to get experienced resources who can lead AI implementation projects.

Artificial Intelligence and other technologies are set to disrupt the way businesses operate. Scaling AI today gives organizations a huge head start not only in picking the low lying fruits of automation and intelligence but also helps them build capacities for the future to drive transformation and innovation.

Qualetics, our on-demand Diaas (Data Intelligence as a Service) platform helps you solve many of the barriers in scaling AI such as time, resources and infrastructure. To know how Qualetics works click here.