Generative AI is dominating discussions everywhere from the boardroom to family dinner tables. The technology represents a seismic platform shift, and — as when Steve Jobs unveiled the iPhone back in 2007 — it’s hard to envision the full scale of the impact it will have on enterprise and our daily lives.
But embracing the GenAI wave is not easy for businesses. The technology is currently so fast-moving that there’s a new development every month, forcing enterprises and regulators to develop their approaches and implementation strategies in real time. Leaders across industries are confronting a huge new challenge. So, in partnership with Salesforce Ventures, we recently brought together a panel of experts to share their insights on navigating the new era.
Fig. 01
The brilliant panel included Laura Hutton, Chief Customer Officer at Quantexa; Seraphina Goldfarb-Tarrant, Head of AI Safety at Cohere; Daniel Svonava, Cofounder of Superlinked; and Dimitri Masin, Cofounder of Gradient Labs AI. The group explored applying AI at enterprise grade, strategies for adapting, and how our companies and societies can safely adopt AI. We were reflecting on the valuable, practical insights that arose in the discussion, and thought it would be helpful to share what we learned more widely. Here are our key takeaways:
Applying AI at Enterprise Grade: Don’t try GenAI without having the technical foundations in place
We all know that Enterprise GenAI looks extremely seductive. Just thinking about its potential productivity gains and streamlining opportunities could be enough to make any company start implementing. However, we were struck by one warning from our experts. They said: If your company was not fully equipped to roll out Business Intelligence or data science, then it is very unlikely to be ready to embrace Enterprise GenAI.
The panellists recommended that leaders take an honest look at their businesses before doing anything, and then to ensure that they have established all the building blocks necessary for rolling out GenAI. Fix the foundational technical elements, then make the leap.
Make sure that there is a clear business use case for your company adopting GenAI
It is appropriate that enterprises today are asking: how do I build a company incorporating GenAI? Or, how do I build AI into my product? But many companies are starting to implement AI based more on its novelty than any clear utility. Our panellists stressed that it is vital to avoid the mistake of adopting GenAI for its own sake, without a clear business use case or for the wrong use case. You wouldn’t use a hammer to tighten a screw — and similarly it’s not a good idea to use GenAI for deterministic decisioning like issuing loans.
Put data sovereignty and trust at the centre of any GenAI strategy you implement
The panel also dived into another key area: the importance of looking at how data governance will be impacted by the new wave of Enterprise GenAI.
The experts advised every business to centre customer concerns around data sovereignty and intellectual property when implementing any new GenAI strategies. Companies have invested huge amounts of money and time in establishing data governance processes, and it’s important to ensure those guardrails remain in place through any new processes. For example, what happens if running GenAI through your systems means that employees can now query any piece of your enterprise data?
On a related topic, GenAI hallucinations are real, and our experts cautioned against pushing LLMs into regulated processes that require a totally consistent, guaranteed accurate output in each instance. If you’re in healthcare, for example, you don’t want your “GP-GPT” to come up with two different diagnoses on the same set of inputs — you either need guardrails, or a different underlying technology.
Building with AI: Align output with your brand and constantly stay adaptable
We took two key points away from our panel’s discussion on building with AI.
The first is to remember that incorporating GenAI into your enterprise does present a very real risk of damaging a company’s reputation or operations. It’s important to ensure that any AI output aligns with your business goals and brand voice, or you could quickly end up in a messy situation — like if your customer service bot gives a user incorrect or off-key advice.
The second point was on implementing in a fast-changing technology landscape. Our panellists advised businesses trying to build with GenAI to remain adaptable and not to plan too far ahead — especially around technical specifications of a system.
Building too specifically on a certain solution is a risk in an industry with lightning-fast pace of change, so it’s best to build core AI competency while remaining open to a modular approach. In the GenAI era, it is necessary to be more willing than ever to be ready to adapt, drop internal research projects, and pivot towards the ‘latest thing’ — all while tightly defining your own core competency and right to exist.
AI Safety is our responsibility — don’t leave it to the regulators
Another line of discussion that stayed with us was around AI safety. The panel explored how alongside the incredible potential of the “human+AI” synergy, the technology’s potential harms need to be taken seriously.
The panel’s view was that regulation from bodies such as the EU is definitely set to bring AI safety and fairness into every product team’s discussions. However, they agreed that we are unlikely to face a major regulatory clampdown on AI in the near future.
In this context, the experts discussed how it is incumbent on everyone, including enterprises, to get approaches to GenAI safety right so that the technology can reach its full potential. For example, to make GenAI palatable to society and regulators, we need to think through features like explainability, transparency and fairness. If we don’t do this, we all risk poisoning the well.
In summary, our panellists agreed that businesses can get ahead on GenAI integration through: meticulous preparation, clear use cases, vigilant data governance, adaptability, and a collective commitment to safety.
Thanks again to everyone who joined us at Dawn HQ for the evening, and please do stay tuned for forthcoming events and further thought-provoking discussions.