Use ChatGPT Wisely to Improve Productivity

The AI will mostly beat you to an answer. But is it valid? Ask better questions.

Analyst perspective

Use generative AI as a muse – and nothing more.

For several years, we've been studying productivity and asking whether tools and automation have rendered it a pointless pursuit.

The rapid adoption of generative AI, inspired mostly by ChatGPT, brings the conversation to a head because, obviously, you're most productive if you can get the AI to do your work for you.

But what if the work output is wrong? What if the quality is low, or it's completely untrustworthy? Worse, perhaps, what if the work output is partially untrustworthy and you didn't know that?

Use ChatGPT wisely, and you'll accelerate toward your most effective and productive self. The AI is your muse, there to help you reflect external inputs off your own experience, character, and discernment as you explore your own conclusions. The AI is the fool, the court jester, and you wear the crown.

Abdicate your intellectual throne, blindly trust the AI, and you revert to the role of passive observer. You no longer act as protagonist as you devolve into the bit part of "hapless dupe."

This is the first part of my series on human productivity. Is it going to immediately become some quaint retro nostalgia piece as we stop producing on our own? I doubt it, and I hope not.

For now, productivity still matters. AI is useful but it's nothing more than a muse (so far).

Barry Cousins
Distinguished Analyst and Research Fellow
Info-Tech Research Group

Executive summary

Your Challenge

• People are busier than ever, mostly in matrix management structures: nobody knows or cares about your workload vs. your capacity.
• Generative AI is exploding in popularity after ChatGPT captured a lot of attention. New entrants are rapidly gaining notoriety, making a lot of people fear being left behind by evolving technology.
• ChatGPT will do your research and writing for you, and it promises to free up time for other work. However, there is a lot of uncertainty about how much to rely on it.

Common Obstacles

• The market hasn't figured out the moral and ethical implications of using generative AI in a business setting. The legal implications are better understood and easily simplified: Don't use AI-generated output as your own.
• It's difficult to know what to trust and believe. ChatGPT sometimes gives provably inaccurate answers.
• The software is a political actor. We cannot be certain about how its opinions will change over time.

Info-Tech's Approach

• Don't jump to the conclusion that it is OK to use ChatGPT as a member of your team.
• Explore, verify, and confront the new forms of bias that come from generative AI.
• Use ChatGPT as nothing more than a muse. It does a decent job of generating ideas quickly, if you don't mind being confined to the universe of common wisdom.
• Challenge and validate any assumptions being made from the software.

Info-Tech Insight

Generative AI is here to stay, even as we prove it is fundamentally untrustworthy. It's foolish to trust ChatGPT to do your writing, though a good skeptic will see it as a thought-provoking muse.

ChatGPT rocketed to prominence as a virally growing technology

Business and IT leaders are debating whether to add it as a trusted technology and perhaps even a member of the team.

ChatGPT went viral in late 2022 as little more than a cheap parlor trick. People were asking it things like, "Compare the Gibson Les Paul to the Fender Stratocaster" and calling their friends together to marvel at the response. Before long, image-generating AI was playing a valued role in content creation as well, opening up possibilities like never before.

Before investing in consumer AI and abandoning your brain's job, confront the underlying questions:

• What's your goal with getting content automatically generated?
• Do you need reasonable output, quickly, like an advanced form of Wikipedia?
• Are you looking to be "right"? According to whom?
• Do you need to prove that you're objective? Factual? Correct?
• Are you trying to be helpful? To whom? How?
• Do you want your creative output to be respected?

This research will:

1. Review what we know so far about generative AI and ChatGPT
2. Identify new forms of cognitive bias being discovered as a result
3. Recommend the best way to use ChatGPT to improve productivity in researchers, creators, and writers

So far, we know that ChatGPT is fast and articulate

If you're simply going through the motions of intelligent discourse, ChatGPT might seem to be astonishingly effective.

1. ChatGPT writes quickly and effectively
   • ChatGPT is a broad consumer offering that responds to written instructions and attempts to provide thorough written answers.
   • The responses are clearly written to a reasonably low level of educational achievement. The language is unambiguous on the surface.
   • The writing style is confident and assertive, making it business-ready for most writing expectations.
2. It has vast knowledge
   • The software is useful for fast recall of commonly available information.
   • As an LLM (Large Language Model), its knowledge is based on massive public data sets with common wisdom and sentiment assumed to be a reasonable proxy for truth.
   • On one hand, we're told that the underlying data set is 'as of' 2021. However, we also see ample evidence that it's updating and/or searching for more current information in real time.
3. It's freemium
   • For now, users can experiment with ChatGPT free of charge. Availability and responsiveness are not assured in any way, and the service is often overloaded or unavailable because of excess demand and frequent updates.
   • A $20/month option has been added, for those who want a more reliable, higher-priority experience.