Use the ‘AI Sandwich’ for a Human Approach to New Technology



This research is part of the research conducted under AI Leadership: 100% Human pillar of Leadership Labs.



AI can bring quick, unexpected insights to leaders. But without a human approach, your work could end up wholly homogeneous. Here’s how to use the technology to whip up something bold, not bland.

The expectations for what AI tools can achieve on their own are pretty high. So much so, that some people predict it will replace the human hand altogether in plenty of strategic areas.

But, in fact, the opposite is true. If anything, it makes the human touch more important.

Harpreet Khurana, Chief Digital and Data Analytics Officer, explains how to get the best out of new tools using a considerate approach—where the use of AI is sandwiched between human intervention.

AI Sandwich

The top slice: Human insight

Imagine you are writing an industry insights report: As a starting point, you might want to review research from third-party data sources and combine those with your own proprietary data. 

It might be tempting to let the AI drive this part of the analysis. But data provenance and confidence in the quality of external and proprietary data is key. For example, how was the proprietary data collected and compiled—was it through 1:1 conversations or an extrapolation of previous work?

Just as the right raw ingredients impact hugely on the quality of your final dish, your analysis is only as good as the data that feeds it—and you want to be able to trust what you’re serving up.  


The filling: AI analysis

Just as you use kitchen tools to more easily chop your vegetables, AI can similarly help add efficiency to your analysis.  

As the chef of your industry analysis report, your time might be better spent letting AI ingest your raw input and looking for patterns and connecting the dots. 

The AI could take a first pass of all the different cuts of the data, rather than manually teasing them out. Meanwhile, you could be looking at gathering additional perspectives from adjacent industries, or connecting with industry experts.

Even here, it’s important at this stage to add a human checkpoint to avoid becoming over-reliant on what the AI has churned out. This could be a deliberate process to avoid a completely generated output, or the simple act of interrupting what comes out of the algorithm from going directly to clients or colleagues. 

In short, necessitate a human touch and sanitize the AI generated analysis. 


The bottom slice: Human decision

Now it’s time to plate your dish. You can use AI assistance in adding the final touches, but you do the ultimate taste test alone. You ultimately craft the output, put it into context and ensure that it drives valuable outcomes.

That’s why human judgment becomes even more important when you introduce AI tools. You need to care about the ingredients that go into the meal, even if the middle part of the process is partially done by an algorithm. It’s important to humanize in order to tailor to end user preferences.

The AI might have helped visualize the different lenses and cuts of data, but it’s only you that truly understands the audience who will read the piece and what they’ll get out of it.

That’s why the sandwich is so important. The AI can save you a lot of time, but it can’t interpret nuance. It can churn out readable content in a matter of minutes, but it can’t understand the interplay between customers and companies. It can’t translate and download human dynamics.

AI is currently output oriented, rather than outcome oriented – it might provide you with certain deliverables, but it’s not focused on the long-term goals or strategy of what you’re trying to achieve.

The human touch is therefore still the most important part of your sandwich.


Humanize, don’t homogenize

When we talk about a responsible approach to AI in terms of a sandwich, it’s not simply the process of AI-human-AI, it’s also a sandwich of perspectives

To get the best out of automation, you also need a cross-section of input. If anything, using a machine to produce content or come up with answers to a critical business problem prompts you to think more about the human element. 

Does the output match with your values? Your legal privacy policy? Has it gone through the right set of layers or filters that represents your organization? If it is simply the output of one person’s prompt, it will be fairly limited.

Instead, using AI needs thought and preparation, through the right seasoning of data sources, sanity checks. In essence, it needs human oversight. If you get lazy on this part, everything becomes a sea of sameness. 

And if all the sandwiches look the same, who will want to eat at your table?



Harpreet Khurana is Russell Reynolds Associates’ chief digital and data analytics officer. He is based in New York.



Join us as we explore the new frontiers of leadership in an AI-powered world.