Want to Scale GenAI? Start with Five Type of Ownership

It seems that GenAI has replaced the weather as people's favourite conversation starter. Almost every organisation is grappling with what the rapidly evolving capabilities of generative models can do for them. Many of them have run hackathons or launched pilot programs as a means of getting started with these technologies. However, there is a big difference between playing around and production deployments. To be successful with the latter, ownership is key.

Firstly, there needs to be ownership of the business problem one wants to solve. While it is tempting to let technologists drive, it is important to start with a clear understanding of what business problem is being targeted. This means being smart about how you scope the project, the KPIs, and determining what success looks like. This is to prevent solutions being developed simply because of the technology's novelty, and ensures resources can be prioritised effectively.

Secondly, there needs to be ownership of the end-to-end solution. This means looking beyond the model itself to all the elements that make the solution work. For instance, although selecting and training models may be the fun part, it's crucial to establish robust data pipelines and suitable endpoints for model outputs. Once the solution is deployed, continuous monitoring is also essential. This typically necessitates a strong partnership with the IT organisation.

Thirdly, there needs to be ownership of any impacted processes. This is important both in terms of understanding how things are done today – and therefore, how they can be improved using GenAI – as well as anticipating where friction or conflict might arise when it comes to solution deployment. For example, what would happen to commission plans if an AI solution is used in certain stages of the sales process, obviating sales managers' involvement?

Fourthly, there needs to be ownership of the ethical implications. Whether an AI solution will be impacting employees or customers, it is important to understand this upfront and assess if AI is being used responsibly. With GenAI in particular, it is critical to understand where solutions could go wrong. For example, could models be used to generate output that is offensive, incorrect, or misleading? Creating transparency for users in how models work is essential.

Finally, there needs to be ownership if things go wrong. Air Canada recently provided a cautionary example when, after their chatbot gave incorrect information, they claimed that the chatbot was a "separate legal entity that is responsible for its own actions." The judge did not agree. Given the degree to which GenAI solutions are propagating, it is critical organisations take accountability and work to rectify mistakes lest they face a consumer backlash.

With GenAI firmly on this year's agenda, executives across industries are feeling the pressure. Yet, while it is tempting to focus on the technology's capabilities, the key to success will lie in leaders' ability to instill a culture of ownership. Ownership of the problem, ownership of the solution, ownership of the process, ownership of the ethics, and unwavering ownership of the outcome. These are the cornerstones to allow organisations to harness GenAI for real impact.

– Ryan

Cover image by DALL·E 3.