Many businesses remain stuck in experimentation and exploration mode when it comes to embracing AI. An IBM study from last year revealed that only a quarter of AI initiatives had achieved their expected ROI. Just 16% had successfully scaled AI across the enterprise despite rapid investment and growing pressure to compete.
These points are underlined in an IBM Think video featuring Matt Candy, Global Managing Partner for the Generative AI Consulting Business at IBM. He outlines a simplified roadmap that highlights key considerations for moving an AI project from pilot to production at scale and driving ROI. In this article, I explore some of the main points in the video.
Rethinking your business model v incremental progress
The companies that often see the biggest return from AI aren't just using it to drive incremental progress. The biggest gains come when companies take the opportunity to rethink their business with AI.
IBM research across EMEA indicates that many business leaders are trying to use AI to support strategic business transformations. Of the 66% who report significant productivity gains, nearly a quarter (24%) say AI is fundamentally changing their business models.
These sorts of transformational outcomes are typically associated with organizations that have an AI-first mind-set. They're not just bolting AI on top of how their business currently operates but using it to make more fundamental changes.
One of the biggest challenges, however, is trying to scale AI while at the same time trying to keep the business moving forward. This is likened to trying to change a tyre on your car while it is still moving.
Deciding where to start with AI
There are a few key ways that the video suggests companies should explore where to start with AI:
- Look for opportunities to increase productivity at scale. Review employee roles, business processes and the value your business delivers. Rethink how to do things better by augmenting roles and workflows with AI tools, allowing employees' creativity and potential to flourish.
- Consider how AI can improve the customer experience by making interactions more personalized and removing friction. Pre-empt customer needs and deliver on them to increase sales, customer loyalty and referrals.
- Explore the potential to introduce gen AI-powered digital products and services to drive new revenue. Entire industries are likely to be transformed and reinvented in this way.
- Find ways to use gen AI to enhance technology delivery. This includes using AI to improve the way you deploy and run back-end technology to optimize workflows, processes and employee experiences.
Use top-down and bottom-up thinking
Another recommendation is to define priority use cases for AI that could make the biggest difference to your business, and then look at them through both a 'top-down' and 'bottom-up' lens.
Consider a bank or financial services organization aiming to reduce operational costs while improving the customer experience. From a top-down perspective, the strategy would point to using AI to streamline service delivery, increase automation and enhance responsiveness to customers.
From a bottom-up lens this translates into role-specific applications of AI: supporting call centre agents with AI assistants for better, more personalized query resolution. Providing AI to the risk teams for faster analysis, to operations teams to automate workflows and to marketing teams to support more personalized engagment across channels.
The important point is that the people closest to the work have the best ideas about where AI can make a difference. Which is why you can't ignore the bottom-up approach.
When Excel spreadsheets were first introduced to the world, enterprises did not have a team at the top figuring out how every department in the organization should make use of them. A bottom-up approach enabled individual departments to explore what Excel could do and determine the specific ways in which it would work best for them.
AI control centre
To scale AI across the enterprise, you're likely to need a mix of technologies and models running across cloud, on-premises and edge environments to align with different use cases. But it's difficult to scale if you don't have consistent approaches across environments, processes and models.
This is why it's important to have a central control layer, to manage and monitor AI across the business. The AI control centre enables AI governance across the enterprise, supporting safe secure access to proprietary data and role-based access to different gen AI models, assistants and common capabilities. It also ensures the right model is used for the right task to optimize cost and performance.
Consistency and repeatable processes are essential for scalability. Without this, organizations risk duplicating effort, increasing complexity and building siloed AI initiatives.
Data layer: data is a competitive advantage
While major AI models are powerful, they are also widely accessible. For most businesses, the real competitive advantage will depend on combining their proprietary data with the right model to generate new sources of unique value.
However, access to proprietary data is a major barrier to AI adoption. In an IBM survey released last year 42% of executives said their organization lacks sufficient proprietary data to customize AI models.
One way to get around this problem is to form strategic partnerships and participate in industry-wide data-sharing initiatives. Enterprises can collaborate with non-competing companies, industry associations, or research institutions to access larger and more diverse datasets relevant to their sector.
Moving from idea to fully scaled AI
Even with the right strategy, use cases, governance, and technical infrastructure in place, it is still a challenge to move an initial AI concept from a pilot to a fully-fledged AI offering that delivers across the enterprise at scale.
IBM has helped clients set up dedicated gen AI labs made up of cross-functional teams to do this. They bring together strategists, designers, developers, data scientists and more, all experimenting and testing and working together to build and grow the solution.
By bringing together all the needed skills into one team, it's possible to make faster, more informed decisions and accelerate the journey from concept to a scaled AI solution.
The temptation with AI is to try to simply improve on what already exists, supporting incremental progress. But this is described as trying to build "faster horses", referring to a famous Henry Ford quote about what people would have seen as progress, had they been asked about it before the introduction of the motor car. In the same way, the real value of AI is likely to come not from making small changes to existing processes. It's going to come from stepping back and rethinking the problem or challenge and using AI to create a fundamentally new, better way of doing things.