AGI Isn’t Here (Yet): How to Make Informed, Strategic Decisions in the Meantime

Since ChatGPT launched in November 2022, the ubiquity of words like “inference,” “reasoning,” and “training data” indicates how much AI has taken over our consciousness. These words, previously heard only in the corridors of computer labs or in the conference rooms of huge technology firms, are now heard in bars and the subway.

Much has been written (and much more shall be written) about how to make AI agents and co-pilots higher decision-makers. However, we sometimes forget that, at least in the near term, artificial intelligence will improve human decision-making moderately than completely replace it. example is the enterprise data corner of the AI ​​world, which incorporates players (as of this text’s publication) from ChatGPT to Glean to Perplexity. It’s not hard to imagine a scenario where a product marketing manager asks his AI tool in text-to-SQL, “What customer segments gave us the lowest NPS score?”, getting the answer he needed, perhaps asking a few follow-up questions, “…what if “Will you segment it by geographic location?” and then use that knowledge to fine-tune your promotion strategy planning.

It’s human-enhanced artificial intelligence.

Looking even further, there’ll likely be a world in which a CEO can say, “Design a promotion strategy for me, taking into account existing data, industry best practices on this issue, and what we learned from the last launch,” and the artificial intelligence will create a person comparable to a good human product marketing manager. There may even come a world where AI is self-directed and decides that a promotion strategy can be a good idea and starts working on it on its own to share it with the CEO, i.e. act as an autonomous CMO.

Overall, it’s protected to say that until artificial general intelligence (AGI) emerges, humans will likely stay on top of vital decisions. While everyone is expressing their opinions on what AI will change our working lives, I wanted to come back to what it won’t change (anytime soon): Good human decision-making. Imagine that your enterprise intelligence team and a group of AI agents are preparing an evaluation for you regarding a latest promotion strategy. How can you utilize this data to make the best decision possible? Here are some time-tested (and lab-tested) ideas that I live by:

Before you see the data:

• Decide on go/no-go criteria before looking at the data: People are known to change the goalpost in the moment. It might sound something like this: “We’re so close, I think another year of investing in this will bring us the results we want.” This is the style of thing that drives managers to proceed projects long after they grow to be feasible. An easy tip from behavioral science might help: set decision criteria before looking at the data, and stick to them as you review the data. This will likely lead to a much wiser decision. Decide, for example, that “We should continue a product line if >80% of survey respondents say they would pay $100 for it tomorrow.” At this point, you are impartial and can make decisions like an independent expert. When the data comes in, you know what you are looking for and you may stick to the established criteria moderately than introducing latest ones in the moment, reverse engineering them based on various other aspects, reminiscent of the appearance of the data or the mood of the room. To read more, check equipment effect.

Looking at the data:

• Ask all decision makers to document their opinion before sharing it with each other. We’ve all been in rooms where you or one other senior person declares, “This looks great – I can’t wait to implement it!” and one other nods enthusiastically in agreement. If someone on the team who is close to the data has serious concerns about what the data says, how can they express those concerns without fear of being rejected? Behavioral science tells us that once data is presented, no discussion ought to be allowed beyond the asking of clarifying questions. After presenting the data, ask all decision makers/experts in the room to quietly and independently document their thoughts (you’ll be able to make this as structured or as unstructured as you would like). Then share each person’s written thoughts with the group and discuss areas of disagreement. This will be certain that you are truly leveraging the broad expertise of the group, moderately than stifling it because someone (normally in a position of power) has influenced the group and (unconsciously) discouraged disagreement outright. To read more, check Asch’s conformist studies.

When making decisions:

• Discuss “mediational judgments”: cognitivist Daniel Kahneman taught us that every major yes/no decision is actually a series of smaller decisions that collectively determine that major decision. For example, replacing L1 customer support with an AI chatbot is a big yes/no decision that is made up of many smaller decisions, reminiscent of “How will the cost of an AI chatbot compare to the cost of humans today and as we scale?”, “Will Artificial intelligence will the chatbot have the same or greater accuracy as humans? When we answer one big query, we implicitly think about all the smaller questions. Behavioral science tells us that making these implicit questions explicit might help improve the quality of choices. So make certain to clearly discuss all the smaller decisions before you begin talking about the big decision, moderately than jumping right into, “So should we move forward?”
• Document the rationale behind your decisions: We all know bad decisions that unintentionally lead to good outcomes and vice versa. Documenting the rationale for your decision “we expect our costs to drop at least 20% and customer satisfaction to remain unchanged within 9 months of implementation” will allow you to truthfully consider your decision during your next business evaluation and learn what you probably did right and incorrect. Building this data-driven feedback loop can provide help to elevate all decision-makers in your organization and start separating skill from luck.
• Set your “termination criteria”: related to documenting your decision criteria before seeing the data, define criteria that, if still not met in the quarters after launch, will indicate that the project is not working and ought to be terminated. This might look something like “>50% of customers interacting with our chatbot request to be redirected to a human after spending at least 1 minute interacting with the bot.” This is the same concept that moves the goalpost that you just shall be “gifted” a project as soon as you green light it and begin to develop selective blindness to the signs of its poor performance. If you define the termination criteria in advance, you shall be committed to the mental honesty of your old, unbiased self and make the right decision to proceed or end the project as soon as the results appear.

If at this point you are considering, “That sounds like a lot of extra work,” you will find that this approach will quickly grow to be second nature to your management team, and any beyond regular time they spend on it provides a high return on investment: providing all the knowledge specialist in your organization is expressed and sets up barriers to limit the negative consequences of choices and draw conclusions from them, whether it would go well or badly.

As long as there are humans in the loop, working with data and evaluation generated by humans and AI agents will remain an extremely invaluable skill set – especially navigating the minefields of cognitive biases when working with data.