Organizations must consider how to use large language model (LLM)-powered tools like ChatGPT in more strategic ways.
By Bryan Kirschner, Vice President, Strategy at DataStax
For all the deserved enthusiasm about the potential of generative AI, “ChatGPT is not your AI strategy” remains sound advice.
That said, it’s still worthwhile to think about how to use large language model (LLM)-powered tools like ChatGPT in more strategic ways.
New research from Microsoft on use of its Copilot AI assistant points to ways that everyone—from the most junior individual contributor to the CEO—can lean in to doing so in any organization.
One of the researchers’ observations was that “LLM-based productivity tools may sometimes provide a new option for information workers that did not exist before: the ability to do a certain set of tasks far faster but with marginally lower quality.”
At first glance that might sound like a tradeoff.
But if we step back and recognize that knowledge worker tasks such as building business cases, exchanging emails, and putting together slide decks are simply means to an end—in this case, “high-quality decisions”—we can look for a win-win.
The key is making intentional choices about where the quality comes from.
The price of going back to the drawing board
One option is a way of working that we’ve likely all experienced. In “Designing Jobs Right,” strategist Roger Martin describes it like this:
“Whether a CEO has delegated a mission to the president of a business unit, or a business unit president has handed over an initiative to a category manager, or a category manager has entrusted a brand manager with a project, the sequence of events is eerily consistent. The subordinates do an enormous amount of work to prepare the project for review by their bosses. They wait until the work is as thorough and bulletproof as possible and then present it for approval.”
Even before generative AI, this approach, common as it is, had its downsides. For staff who did get “sent back to the drawing board,” a feeling of failure is almost inescapable–accompanied by dread that now they must work even harder to prepare for the next review.
And it puts managers and executives into something of a bind for how they add value–so, as a result, sometimes: “…bosses have no interest in sagely nodding and saying, ‘Great work!’ That is a dumb job. They want a real, value-adding job. And when they haven’t been given one, they tend to create one that isn’t terribly helpful: nitpicking. What about this? Have you thought about that?”
If the boss really does identify a material flaw in the team’s thinking, they have indeed protected the quality of a final decision, but at the price of demoralization. If they truly only poked some holes that fundamentally don’t matter, they’ve triggered both demoralization and needless rework.
And the mindset with which people might be primed to approach that rework can be dangerous now that we’re in the age of generative AI, because “we need more facts to bolster our case” is a risky, even counter-productive way to use it.
Spark a fruitful conversation with AI
Generative AI has boundless capacity to tell you what you want to hear–including confidently presenting 100 percent fabrications (“hallucinations”) as facts. If a team presses it for more facts in search of the perfect, bulletproof case (while the boss taps it for more “gotchas”), we’ve spun up a “worst case scenario.”
Conversely, generative AI is very good at helping generate new ideas, faster, and providing instant–even if imperfect–feedback and examples that can still spark a good discussion even if they happen to be made up. (Think about hypothetical scenarios: we talk through them as humans all the time.)
These “skills” are a great fit for a different way of working that also predates generative AI. As Martin recommends:
“Instead of waiting until the 11th hour to give bosses a dumb job, give them smart jobs along the way. Come back early and say, ‘Boss, I’m defining the problem you gave me as one of streamlining our go-to-market approach to make it more cost-effective and responsive to end customers. Does that definition resonate with you? How might you modify or enhance it?’ That is a real job that bosses can do and will enjoy doing, and it will help your strategy effort.
“When you have possible solutions, come back and say, ‘Boss, based on the problem definition that we refined, I’ve come up with the following three potential solutions. Are you so allergic to any of them that it isn’t worth pursuing? Or is there another possibility floating around in your mind that I should be considering?’ Again, that’s a perfect task for bosses, and in my experience of helping managers have this dialogue, bosses love it and add value in taking it on.”
Obviously, this is a two-way street. Both the team and the boss must buy into it.
But if we estimate that generative AI can save at least 25 percent of the time spent on generating artifacts (like emails and slide decks) while preserving 80 percent of the quality, we have choices. One is to shoehorn that back into a “strive for perfect artifacts, then submit to audit” model of getting to high-quality decisions.
The other—and a far better exploitation of generative AI’s strengths—is “iterate faster, and put the quality in through collaborative conversations.” These conversations should lead to convergence on both a preferred decision and the most important facts to verify in order to make that decision with confidence–a perfect use for human talent (rather than tweaking otherwise “good enough” generative AI output).
Reorganize the way you arrive at decisions
In the transition from steam power to ubiquitous electricity in factories, the big gains only came when factory floors were reorganized to take advantage of freedom from the constraints of steam engines and belts. In the transition to ubiquitous generative AI, we should take a lesson from the past and always think about new ways of working in order to make best use of the technology–including how we organize knowledge work to arrive at great decisions.
Learn more about generative AI.
About Bryan Kirschner:
Bryan is Vice President, Strategy at DataStax. For more than 20 years he has helped large organizations build and execute strategy when they are seeking new ways forward and a future materially different from their past. He specializes in removing fear, uncertainty, and doubt from strategic decision-making through empirical data and market sensing.
>>> Read full article>>>
Copyright for syndicated content belongs to the linked Source : CIO – https://www.cio.com/article/1265353/chatgpt-isnt-an-ai-strategy-but-it-should-be-a-strategic-tool.html