5 Essential Tips for Your First Generative AI Project

The hype around Generative AI continues to gain steam, with many organizations embarking on a journey to begin evaluating these tools for internal use and adoption. This has led many organizations to become marred in analysis paralysis, struggling to determine where to start and how to define success.

While not uncommon with new technologies that drive radically different organizational models, it can be unsettling for an organization looking to make quick progress and ensure a place ahead of their competition fueled by the application of Generative AI.

We have previously explored how to define our policies that govern Generative AI use, what enterprise readiness will look like for these tools, and potential use cases, including marketing and incident response.

Once we have explored how Generative AI empowers our organization, we can look at structuring a team to apply Generative AI technologies quickly, which drives learning and further successful adoption.

1. Team Building for Generative AI Projects

First, we begin by identifying the team that will take on this initial project. My principles for team building during these bleeding-edge projects are:

• Small – Five or fewer people ensures quick decisions can be made, everyone is accountable for taking action items, and communication overhead is kept low.
• Empowered – The team members should have clearly delegated authority from leadership to decide and act to achieve the agreed objectives. Team members should have the necessary authority and resources to move quickly, adjust priorities for supporting team findings and accept risks that would normally be managed by or communicated to other teams.
• Passionate – Team members that have invested personal time reading and exploring this exciting market. Their passion will be a powerful asset in overcoming technical problems, applying high levels of creativity, and bringing forward constantly new ideas.
• Time-boxed – Urgency can often bring out higher levels of creativity. Urgency comes from time-boxed projects where teams know they have a set time to deliver maximum value while ensuring their time and priorities are protected. Leadership should put reasonable time limits on the project team and come together regularly to assess progress and determine if the investment continues or if the team has moved to other work priorities.

2. Identifying and Engaging with the Executive Sponsor

Second, we must identify and engage with our executive sponsor. This leader will play a key role in ensuring the team has adequate resources for execution, as well as support for fast-tracking company policy and decision-making. Projects that take on the implementation of new technologies will often run into roadblocks driven by existing processes that do not account for new capabilities.

Our executive sponsor is the one willing to take on the risk while understanding the potential rewards to fast-track decisions to enable the team to move quickly, learn and pivot as needed. Given the newness of Generative AI, we will have to invest time in educating this executive and his leadership team on the opportunity, pitfalls, and learnings.

3. Prioritizing Use Cases for Generative AI

Third, we move to prioritize the potential use cases where Generative AI will provide us a competitive advantage. I am a fan of simplistic methods for prioritizing potential use cases, a simple comparison of the potential return of a use case (savings or revenue) against the ease of data availability for the specific use case.

The most important part is that the potential return is a relative measure to compare the use cases against one another, not absolute monetary value. At this stage, the ability to get granular with specific monetary returns will be limited, and our goal is to prove viability and approach over any specific financial returns.

4. Developing a Communications Plan

As our project team organizes and moves forward on testing one or more Generative AI technologies against our prioritized use cases, we must develop a communications plan to keep the wider organization aware of the project status, learnings, and any changes in objectives.

The project team should identify a member that is accountable for communication on status to the organization. This will help with setting proper expectations with various stakeholders.

5. Managing Initial Generative AI Projects

These projects are never easy. The team is learning a lot and pivoting often while managing inputs and requirements put upon them by outside teams. Some best practices for managing these types of initial projects include:

• Iterate & Improve – The team should establish culture and behaviors early for how progress is measured, viewed, and decisions made. This culture should value experimentation, iteration, and incremental improvement.
• Data Sensitivity – Infosec considerations, compliance, and third-party agreements can often take short projects and extend the time in unpredictable ways. The lower the sensitivity data the team can start with, the lower the risk to the organization and the faster they can move with trialing new technology and vendors.
• Align on Expectations – Many projects where new technology is being explored get held up due to misalignment in definitions of the pilot, proof of concept, or production-ready deployment. The earlier the project team can define specific outcomes in terms of data availability, user readiness, and integration with other business systems, the more effectively the team can manage expectations and timelines.

Conclusion: Exploiting the Potential of Generative AI

Generative AI is going to move like a bullet train through our organizations. If centralized IT organizations do not deploy these capabilities quickly, business units and functional teams will. The most successful organizations at exploiting these trends will be the ones that can knock down barriers that slow experimentation and adoption to allow the most valuable tools to be deployed quickly and broadly in the organization.

Find your passionate team members, give them executive support to experiment quickly, and communicate learnings broadly.