Sponsored by Google Cloud
Choosing Your First Generative AI Use Cases
To get started with generative AI, first focus on areas that can improve human experiences with information.
While Microsoft has historically been synonymous with reliability and safety, deploying generative AI requires a more nuanced approach
Since the 1990s Microsoft has been the biggest and most trusted brand in enterprise IT. The adage: "No one gets fired for buying Microsoft," has long been a guiding principle in IT circles, reflecting the company's reputation as a safe choice. Even though the quality of their products has been variable, an IT decision-maker would rarely regret their decision to invest heavily in Microsoft's latest offering.
With reports casting doubt on its benefits, is Copilot 365 the exception to the rule?
In November 2023, Copilot for Office 365 became generally available. Launching with a minimum commitment of 300 seats at $30 per month, a year of Copilot involved a minimum investment of $108,000. For a mid-sized 1,000-person organization, the cost was $360,000 annually, on top of existing Office 365 licenses and internal IT and training costs.
Given the difficulty of estimating the efficiency gains of such a general-purpose tool, many companies decided they couldn't afford to wait and chose to take it on trust that the benefits would be realized down the line.
Now, many companies are still finding it hard to quantify the return on their investment in Copilot. Coupled with reports casting doubt on the economic impact of AI, many IT professionals who advocated Copilot rollout are facing difficult conversations in the boardroom.
The inability to point to the benefits of generative AI typically comes down to three key issues:
Assistance vs. augmentation
Ineffective rollout
Inability to measure benefits
Assistance vs. Augmentation: Finding the Right Approach
There are three fundamental approaches to using Generative AI to support your business:
Assist: A human asks for help when they have a question, or AI makes suggestions. (e.g., Copilot, Apple Intelligence).
Augment: Human Oversight of AI processes - AI transforms and creates the content and humans verify.
Automate: AI-powered processes with minimal human intervention.
Tools like Copilot provide an assist capability. This is the easiest for users to understand but the benefits are marginal, not transformational. Assistance is helpful when you have an ad hoc query, but limiting when you want help with complex repeatable work at scale. Prompt libraries and other tools help to an extent, but they still rely on the end user to do a lot of the heavy lifting.
Faced with this problem many companies will try full automation, but this is usually the wrong solution. Industries in heavily regulated industries require that the outputs are accurate and highly trusted and so automation without a human in the loop creates too much risk unless the problems they solve are very narrow.
The augment approach offers an elegant solution. As Steve Jobs put it, technology should be a "Bicycle for the mind". AI is called upon to do the heavy lifting such as drafting, summarizing, translating, and extracting information in a highly automated way, but the human remains in control providing the compliance and risk management clients expect.
Ineffective Rollout
This is another situation where too much focus on providing an assistant to users rather than augmenting their capabilities limits the benefits that can be realized. Technologies like Copilot and ChatGPT seem almost magical in their ability to attempt any task you can dream up and this leads some to believe you can put it in front of users and suddenly their productivity will dramatically improve. Sadly this is not the case.
At its best generative AI is like having a tireless assistant at your disposal - but we would not expect an employee to gain a human assistant and suddenly become an expert in delegation. Understanding how to get the best out of AI requires training, practice and a willingness to engage with new technology. Companies need to identify their superusers - those who know how to unlock an LLM’s full capabilities - and foster and reward them to ensure that their knowledge is captured and shared across the enterprise.
Inability to Measure Benefits
Even when benefits are realized, it may not be possible to quantify these in a way that supports the investment case for rolling out a tool. To effectively measure the benefits of AI adoption:
Establish a clear quantified baseline of your organization's throughput and capabilities.
Analyze core processes critical to your organization's performance.
Encourage the sharing of interesting use cases across the business.
Consider the impact on billable hours in industries that use that model.
Track intangible benefits to staff wellbeing.
For technology leaders, the implications are clear: The rollout of generative AI tools like Copilot 365 should not be automatic. While Microsoft's brand has historically been synonymous with reliability and safety, deploying generative AI requires a more nuanced approach. Simply relying on the "safe pair of hands" reputation won't suffice. Instead, focus on augmenting your workforce's capabilities rather than merely assisting them.
This means using a platform built for AI, augmenting users, fostering superusers, and measuring benefits against a clear baseline. By doing so, you can ensure that your investment in generative AI not only enhances productivity but also aligns with the high standards of accuracy and trust your clients expect. Ultimately, the goal is to empower your employees to leverage AI in a way that transforms your firm's operations, delivering measurable value and maintaining a competitive edge in your industry.
You can take action now: evaluate your current AI strategy and consider how you can shift from assistance to augmentation. Identify potential superusers within your organization and create a plan to nurture their skills. Most importantly, explore AI platforms that offer the flexibility and customization needed to truly transform your operations.
You May Also Like