Generative AI tools—such as ChatGPT, Gemini and Midjourney—can make your work better, help you complete tasks faster and enable you to achieve more. But, as you may have experienced, while mostly impressive, the outputs can seem a little… “experimental” at times. So, it should come as no surprise that replicating this process safely and consistently across a whole enterprise—while creating a real competitive advantage for your business—is another task altogether. In this article, I am going to explore what we’ve learnt helping leaders use Generative AI to create real business value.

GenAI: Should we follow the trend?

While the Generative AI market is expected to exceed $66 billion by the end of 2024, a little scepticism has started to spread. Some leaders are finding that the returns from GenAI applications aren’t justifying the costs of building and running them. It’s clear that GenAI can be an effective tool in growing revenue and reducing costs. The challenge is how. It isn't about following another tech trend. It’s about creating strategic tools to improve your company’s performance and position yourself for long-term success.

At Datasparq, we help enterprise clients across industries through this transition. Today, I’m going to share the methodology that we’ve defined to ensure that business value takes centre-stage—and help you avoid being led astray by the latest buzzword. So, to hear how we approach working on winning GenAI solutions, read on—you’ll find a clear guide to the key steps as well as a look at the bigger-picture, strategic aspects of GenAI.

"We’re really pleased to be working with machine learning experts like Datasparq; helping us to accelerate our ability to identify opportunities for GenAI and deploy fast, scalable solutions to solve our business challenges."

— Gabrièle Breda, Research & Innovation Director, ProductLife Group

Drive value with GenAI with enterprise use cases

1) Understand business priorities to evaluate GenAI solutions

The first step of the process is to get a deep understanding of your company's goals. This will allow you to map the strategic value of GenAI across departments and how to build competitively differentiating capabilities. Whether it's boosting revenue, winning market share or improving customer experience, initiatives should be aligned with your company goals. Armed with hypotheses, work with business stakeholders to validate the problem statements you are trying to solve and articulate precisely how solving these will benefit your organisation. These insights will allow you to create a longlist of use cases where GenAI can make a meaningful impact. For example, GenAI is usually a great choice when it comes to automating non-deterministic tasks to improve things such as speed and consistency, but it’s also important to consider where other approaches (solutions other than GenAI) might perform better. You might find there are easier, faster and cheaper ways to achieve the desired outcome (e.g. simply re-designing a business workflow). At this point, it’s crucial to make sure that the use cases you’ve come up with solve a meaningful business goal. Otherwise, you’ll be stuck in the precarious situation of having a solution in search of a problem.

2) Develop a clear business case to prioritise the right use case

While it feels like the number of use cases for GenAI is endless, business resources are finite. Death by a thousand pilots will deplete your company's resources and scatter your focus. Here's where prioritisation comes in. A strong business case is the foundation for any successful AI initiative. By clearly outlining the value proposition and potential ROI across the entire solution lifecycle—from ideation to deployment—you not only secure stakeholder buy-in, but also provide a business lead technology roadmap to navigate uncertainties or complexities that arise.

Building a winning business case starts with identifying the right metrics. Focus on KPIs that directly tie back to your business goals. These metrics should be clear and quantifiable, reflecting the expected outcomes of your GenAI project. For example, if you're aiming to improve customer service, KPIs could be increased customer satisfaction scores (which can impact lifetime value) or faster response times, leading to lower operating and opportunity costs. This value-driven approach will ensure your project's financial viability and should consider not only the upfront investment, but also ongoing maintenance, training and potential scaling costs. Balancing value with complexity will be key to securing ROI and buy-in across the business. 

Securing a business champion is essential to keep momentum & drive value. Crucially, involve business sponsors who help define and own the chosen KPIs. Their stake in the project is key to its success. Engaged business leaders, especially those actively driving AI initiatives, will make all the difference. Their support creates a smoother path when navigating the inevitable uncertainties and complexities that will arise during the project. For instance, they will be essential to help you plan the rollout, test the solution, and even generate buy-in from end-users. Finally, clear KPIs keep your team motivated. Simply put—everyone likes working on projects with a measurable, meaningful impact.

3) Keep the end-state in mind to create a solution that scales

a) Create a roadmap & implementation plan

Plans will change—but a high-level roadmap is essential.  If you've followed this guide, you already have a vision for your GenAI solution—its long-term goals and expected outcomes. Next, break down your project into key components, assessing the complexity of each element. Chunk the implementation into achievable milestones, enabling continuous progress and leaving space for adjustments to be made along the way. Monitoring KPIs throughout the process is crucial—it allows you to track progress and measure success against your goals. Finally, prioritise a user-centric rollout plan. A smooth transition and user adoption are vital for long-term success.

b) Align the solution to your company data & AI roadmap

Collaboration is key for success when it comes to building new enterprise capabilities with GenAI. Having a business sponsor champion the project is essential (as we mentioned before), but strong IT collaboration is equally important. Ensure your GenAI initiatives align with the broader enterprise digital and cloud roadmap, ensuring coherence and compatibility with existing systems and future tech plans. These tools should not be built in a silo!

Building with GenAI offers flexibility. While constructing a large language model (LLM) from scratch could be an option, it's expensive and requires very niche skill sets. Leveraging pre-built LLMs or refining them through prompt engineering can be a more cost-effective, faster path to value. Integrating these models into business tools requires robust technical infrastructure and seamless integrations for scalability.

Don't forget the people-side of the equation. Ongoing maintenance, training and potential infrastructure scaling necessitate the right skills to operate GenAI solutions at scale. This means upskilling your workforce and potentially acquiring new talent. Senior Leadership needs to identify required capabilities based on prioritised use cases and develop a plan to build and sustain these skills.

The talent mix depends on your approach. For solutions built on existing models or SaaS offerings, a data engineer and software engineer might suffice. However, for more complex solutions, technical and talent needs can vary widely and potentially extend beyond the technical team to include expertise across departments. This could include for instance, product, commercial, legal and other business functions. The remit of these contributors should be clearly defined as you develop your business case.

C) Keep the end-user in mind and co-build with them to enable adoption at enterprise level

Don't underestimate change management. A robust change management plan is crucial for successful GenAI adoption. This means a clear communications plan, training strategy, business process redesign to support how employees do their work, manage issues, track benefits and realise value.

Upskilling is key. Equipping your workforce with GenAI skills is critical. Provide hands-on training to ensure employees can leverage these new tools effectively. While GenAI applications can be user-friendly, users might still need to optimise prompts, understand technology limitations and integrate these tools seamlessly into their business as usual (BAU) workflows.

Plan for continuous learning. Your rollout plan should include clear usage guidelines and ongoing education about potential risks and new developments. Fostering a culture of exploration encourages innovation and effective GenAI integration into your business processes, while of course putting the relevant safeguards in place.

There are lots of opportunities, but clear pitfalls, too

1) The unspoken truth about enterprise GenAI use cases

GenAI solutions are iterative; open source LLMs are a black box. Building impactful and usable GenAI solutions is an iterative process. These tools are tailored for specific tasks, requiring a deep understanding of training data, potential gaps and desired outcomes. Moreover, GenAI technology differs significantly from “traditional” AI. Unlike conventional models where methods to explain behaviour exist and can be experimented with (for instance, SHAP values), LLMs are black-boxes to users, abstracting much of their complexity.

Accessibility comes with trade-offs. While GenAI's accessibility is a plus, its underlying technology has limitations users might not see. Not knowing what training data the model has used, along with developers making hidden changes to LLMs create challenges for businesses in output reproducibility, scalability and bias. For a deeper dive into these issues, see this article by my colleague Michèle Pettinato.

Don't underestimate complexity. Many users, impressed by applications like ChatGPT, underestimate the complexities of collecting a clean dataset for developing, testing and maintaining robust GenAI solutions at scale. This is especially true for out-of-the-box models where limitations and the need for customisation are often overlooked. There’s an enormous difference between getting ChatGPT to return a sensible answer to a single user and automating this process to safely and consistently replicate results at an enterprise level.

The answer? A strong evaluation framework. To mitigate these challenges, businesses must implement a robust evaluation framework that effectively assesses and explains GenAI outputs.

2) Evaluation framework: A crucial component to materialise value form GenAI

Given the inherent complexities of GenAI for enterprise use cases, a structured evaluation framework is essential. At Datasparq, we advocate for a framework with three core components:

A) Value assessment

This component focuses on measuring the tangible business value derived from a GenAI use case—think increased efficiency, cost savings or improved customer satisfaction. Additionally, incorporating social value metrics, such as ethical considerations, ensures a well-rounded evaluation.

B) Explainability

To address the black-box nature of GenAI, explainability tools are crucial. For traditional data science use cases, you can use SHAP values to explain model output. Unfortunately similar techniques don't exist yet in the industry to help explain why the LLM is producing certain outputs. What you could build instead are processing flows (i.e. architectures and prompt sequences) with break-points at each step, where you can examine the output from the LLM and even ask the model to provide reasoning for each decision. By demystifying the decision-making process, businesses will be able to build trust and reliability in their GenAI agents or applications; especially in industries like finance or the medical sector.

C) Enterprise GenAI guardrail

An enterprise GenAI guardrail serves as a company owned post-processing service to assess and mitigate risks associated with GenAI outputs. This tool (usually built on your enterprise cloud) could check for biases and industry-specific issues, ensuring compliance with company standards. Additionally, it could run offline tests to monitor changes in the underlying LLM, such as detecting output drift by analysing the consistency of responses to a standard set of prompts over time.

Implementing such a framework empowers enterprises to not only leverage GenAI effectively but also ensure that the technology aligns with their strategic objectives and ethical standards. This holistic approach is key to navigating the complexities of GenAI and achieving sustainable business value.

Conclusion: How businesses can navigate GenAI

GenAI is more than a tech trend; going forwards, it’s a strategic necessity. As the market for generative AI continues to surge, its potential to drive business growth and reduce costs becomes increasingly evident. As I’ve set out in this article, embracing GenAI isn’t about following the latest fad—it’s about finding the right use cases for your business that will pave the way to long-term success.

Remember:

1. Prioritise business value: Align GenAI initiatives with your company goals.
2. Build a strong business case: Articulate clearly the value proposition and ROI of the solution while defining an iterative rollout plan.
3. Develop a comprehensive evaluation framework: Everything is measurable and you should assess outcomes continuously to refine your approach and pivot if needed.
4. Foster a culture of innovation: Identify solution champions, create an enablement plan while communicating changes.

By following these principles, you’ll leverage GenAI effectively to drive innovation, improve efficiency and achieve sustainable business growth. Something we didn’t touch on in this article is AI governance. While it’s not something you necessarily “own” when working on deploying individual solutions, it’s key to consider as part of a wider data and AI strategy. I’ll be writing more on this in the coming months.

At Datasparq, we help organisations unlock the power of GenAI. Our methodology ensures that business value remains at the forefront of every initiative. By understanding your priorities, developing a robust business case and implementing a comprehensive evaluation framework, you can navigate the complexities of GenAI and unlock its full potential. 

If you’re interested in learning more about how GenAI—and the team here at Datasparq—can help your business, send us a message. We’d be delighted to help!