Sign up for our daily and weekly newsletters to stay updated with exclusive content on leading AI developments.
Since the launch of ChatGPT in November 2022, terms like “inference,” “reasoning,” and “training-data” have become incredibly common. Once confined to computer science labs and major tech company meetings, these words are now heard everywhere, from bars to subways.
A lot has been written about making AI agents and assistants better decision-makers. However, it’s essential to remember that at least for now, AI is here to enhance human decision-making, not replace it. For instance, in the world of enterprise data, tools like ChatGPT, Glean, and Perplexity help professionals make informed decisions. Picture a product marketing manager using a text-to-SQL AI tool to ask, “Which customer segments have given us the lowest NPS rating?” and getting the necessary answers. She might further inquire, “…and if we segment it by geography?” and use these insights to shape her promotional strategies. This scenario demonstrates how AI supports and augments human capabilities.
Looking further ahead, we can imagine a future where a CEO could ask AI to “Design a promotion strategy based on existing data, industry best practices, and lessons from our last campaign,” and the AI comes up with a strategy comparable to what a human expert would devise. There might even be a time when AI autonomously decides to create a promotional plan and presents it to the CEO, acting as an independent Chief Marketing Officer.
Overall, until Artificial General Intelligence (AGI) becomes a reality, humans will remain integral to significant decision-making processes. While there’s much speculation on how AI will change our professional lives, I want to focus on what won’t change soon: the importance of sound human judgment. Imagine your business intelligence team, supported by AI tools, preparing an analysis for a new promotional strategy. How do you use that data to make the best decision? Here are a few tried-and-true methods to ensure wise decision-making:
Before seeing the data:
– Set decision criteria in advance: Humans often shift their goals based on immediate data. This tendency might lead executives to continue investing in projects long past their prime. To avoid this, decide on your go/no-go criteria before reviewing the data. For instance, determine that “We will pursue the product line if more than 80% of survey respondents say they’d pay $100 for it tomorrow.” When the results come in, you’ll stick to your pre-set criteria, avoiding the impulse to change your goals based on current sentiment.
While reviewing the data:
– Document independent opinions: Before any group discussion, have all decision-makers document their thoughts independently. This practice prevents groupthink, where strong opinions (especially from authoritative figures) can unduly influence everyone else. Once thoughts are recorded, share and discuss differing opinions to get the most from the team’s collective expertise.
While making the decision:
– Discuss intermediate judgments: Big decisions consist of smaller, underlying decisions. For example, deciding whether to replace customer support with an AI chatbot involves questions about cost, accuracy, and scalability. By explicitly discussing these smaller questions, you improve the overall decision quality.
– Document decision rationale: Clearly document why a decision was made, such as expecting a 20% cost reduction and stable customer satisfaction within nine months. This record allows for honest assessment later, helping to refine decision-making processes over time by distinguishing between skill and luck.
– Set kill criteria: Decide upfront under what conditions a project should be terminated. For instance, if more than 50% of customers interacting with a chatbot ask for a human within a minute, it’s a sign the project is not performing and might need to be reconsidered. This prevents project proponents from ignoring signs of failure due to emotional attachment.
Adopting these practices may feel like extra effort initially, but they quickly become second nature. The time spent ensures all expertise within the organization is utilized, setting boundaries to minimize risks and providing a valuable learning experience from each outcome.
As long as humans are involved, the skill to work effectively with data and insights—whether from humans or AI—remains crucial, especially in avoiding cognitive biases.
Sid Rajgarhia, a member of the investment team at First Round Capital, has spent the last decade focusing on data-driven decision-making in software companies.