Product Management frameworks for Generative AI applications

Generative AI caught the world by storm in late 2022 when ChatGPT was released to the general public. Never before deep neural networks were applied in real time low latency applications used in such a mass scale. With every such hype, also comes warning of doom with Father of AI Geoffrey Hilton resigning from Google owing to his concerns about responsible development of AI by big tech and ensuring we consider incorporating “right” values into AI development before embarking further as explained by noted authors and scientists Max Tegmark and Nick Bostrom in their books Life 3.0 and Superintelligence respectively.

Despite these doomsday call outs, generative AI capabilities as available to the public today have the capability to provide valuable product and business opportunities to improve effectiveness and efficiencies with potential impact of $2.6–4.4 Trillion annual as per a Mckinsey report . Additionally product managers in tech companies are constantly encouraged to think about generative AI applications as new product offerings or improvement upon existing products to ride the hype cycle around generative AI and capture the consumer surplus as much as possible from this mind-blowing technology. While exciting as it sounds, this is a threat to business and product owing to this backward thinking of force fitting solutions to market without thinking of customer problems first. Here are some classic examples from the past where force fitting technology on customers didn’t pan out well for the business and products from even the most renowned innovators we know in the current age -

1. Failure of Next computers by Steve Jobs — Netizens may not take it lightly but Steve Jobs, the renowned product guru behind the successes of blockbuster products like iPod, Pixar, iPad which continue to fill Apple’s vaults to this day had a classic case of product failure when he didn’t think of customers needs first. When Steve Jobs was ousted from Apple in 1985, he started Next Computers and built one of the most advanced computers of his time which Tim-Berners Lee later used to create the world wide web. But at the time of the launch, the product was a massive failure. Reason — the product was priced at $6,500 and was targeted at educational enterprises which ran on tight budgets and probably did not need ultra-advanced computers to train students who were starting to learn the basics.
2. Google Glass — Smart glasses are still not a reality today. In 2015, however, when Google launched Google Glass, there was a hype around the world around its extravagant capabilities to scan the environment, detect objects, use AR/VR in real time, search an object on Google using the visuals in the glasses etc. It was a definitely a product out of a technology fairy tale but it failed to pick steam among customers not only due to its high price ($1,500) but also due to privacy issues which were probably overlooked in the initial phases of product development and research.
3. Metaverse by Meta — In 2022, Mark Zuckerberg rebranded Facebook company into Meta by pinning down on its founder’s vision of Metaverse and how it will lead the social media behemoth into the next frontier of technological innovation. With much hype, Horizon Worlds, Meta’s version of Metaverse failed to pick steam among young adult customers who found the game graphics and capabilities to be of yesteryears when they were already attuned to ultra-fine graphics and games offered on Playstation, XBox and Nintendo. While Meta continues its quest, as per media reports, Mark has already significantly cut down on Metaverse investments internally and doubled down on its core business of social media and ads leveraging the power of classic AI.

The above examples prove that sometimes even the top technology leaders make the folly of failed product launches when they don’t think backwards from customer needs. With generative AI hype cycle at its oeak and increasing pressure from executive leadership to leverage the new technology as a golden egg Goose, this can quickly turn into a bloodbath and easily fail careers, products and businesses if not done correctly. The right and the safe approach to think about application of this new technology is to think about customer problems and whether they fit into the following problem types for Generative AI use cases for internal / external customers and stakeholders.

Overall, I think there are following two main problem types which can be addressed by Generative AI -

A) Operational efficiency improvements

• User needs to do individually set a lot of variables in product page to create one unit or useful work or entity (eg setting campaigns on a DSP)
• Search heavy feature requirements
• Repetitive customer support responses
• Internal wiki search

B) Communication effectiveness

• Quality of customer support responses
• Quality of GTM materials from product marketing teams
• Partnership communications
• Sales communications

Once PMs have identified impactful problems which fit into one or both of the problem types and examples, they have a strong use case of leveraging generative AI to transform their business and products. Obviously impact must be weight against efforts and measured against existing product roadmap priorities, but the above approach must reduce variability and risks associated and help PMs tread with confidence into the new era where change is meeting us on our face.

Hope this helps and am looking forward to connecting with fellow PMs working on generative AI and discuss interesting or new approaches of using this technology to solve customer problems.