Cultivate genuine innovation through experienced evaluation, strategic intent and mechanisms that reward novelty over replication.
From the internet, smartphones to 3D printing, recent decades have ushered in general-purpose technology that increases efficiency and collapses the cost of routine tasks. The latest general-purpose technology – you guessed it, generative AI (GenAI) – has the potential to also extend the frontiers of creative search and innovation.
The rise of GenAI tools has not just accelerated innovation, it can democratise it. Our recent research examining the Mozilla Firefox add-on ecosystem found that new add-ons – think grammar checkers, photo editing tools, ad blockers – surged 34 percent following ChatGPT’s launch in November 2022. This boom wasn’t confined to experienced developers. Novice creators, riding on GenAI’s capability to synthesise domain knowledge, ideate and generate code, piled into the market.
Our data shows that add-ons introduced by new developers maintained comparable quality ratings to those by experienced developers, suggesting GenAI helps bridge not just knowledge gaps but execution gaps as well.
Less rosy, however, is our finding that the new products weren’t necessarily more novel than existing ones. This begs the question of the kind of innovation GenAI enables. Our paper, which won this year’s Best Paper Award by the Technology and Innovation Management Division of the Academy of Management, suggests that when humans rely heavily on these tools without strategic intent, they become efficient in producing sophisticated variations rather than genuine breakthroughs.
The democratisation paradox
Between June 2021 and May 2023, we tracked over 31,000 software add-ons in the Mozilla Firefox ecosystem, using ChatGPT’s November 2022 launch as a natural experiment. For each add-on, we recorded its release date, update chronology, developer identity, primary category and user counts.
The data showed a 34.4 percent increase in new add-on introductions per category after November 2022. By contrast, updates to existing add-ons declined by 20 percent, potentially degrading user experience in the long run and even leading to security risks.
It’s like what happens in a gold rush: Everyone races to stake claims on fresh ground while their existing plots go unattended. But whereas the easy-to-reach surface gold (efficient recombinations) is exhausted quickly, the deeper, more valuable ore (genuinely novel innovations) remains buried without deliberate excavation strategies.
Old wine in new bottles
Perhaps our most significant finding concerns the nature of these new innovations. Despite the explosion in original product introductions, the actual novelty was rather thin. Developers were introducing new add-ons with similar functions – ad blockers, grammar checkers, etc. – and characteristics to existing ones.
This pattern reveals what’s lacking, innovatively speaking, about how GenAI operates. These systems are trained on large datasets and optimise for more efficient knowledge recombination. What they struggle with is making creative leaps that go significantly beyond what has existed before.
It’s rather like giving everyone access to the same ingredients and kitchen utensils: more meals will be served, but don’t expect Michelin-starred creativity unless someone ventures beyond existing recipes. GenAI is good at efficient recombination and helps developers combine knowledge from diverse domains more quickly, without necessarily pushing them into genuinely novel territory.
Rethinking innovation strategies
Platform owners, whether it’s Mozilla, Apple, Google or others, must tread a fine line between encouraging innovation and ensuring that new products aren’t merely incremental variations of existing offerings. This requires establishing guidelines or automated checks for AI-generated code to screen for genuine novelty. It also calls for the implementation of incentives that motivate developers to engage in truly novel innovation.
Equally critical is addressing the maintenance deficit. Platforms should incentivise regular updates through rewards, visibility boosts or algorithmic advantages for products that remain secure and compatible over time. Without such mechanisms, neglected code accumulates vulnerabilities and compatibility issues multiply, ultimately degrading the user experience despite the appearance of a thriving ecosystem.
Effectiveness vs. efficiency
For organisations, our findings offer two takeaways. First, as technical barriers collapse, differentiation increasingly depends on factors GenAI cannot easily replicate. True creativity, domain-specific insights and quality evaluation – these are the new competitive frontier.
Lauded management consultant Peter Drucker’s famous advice is relevant here: “Efficiency is doing things right; effectiveness is doing the right things.” GenAI has made us more efficient. Whether or not we are effective depends on ourselves.
Second, and perhaps most counterintuitively, traditional software development skills haven’t become obsolete. They’ve only become more valuable in different ways. Experienced developers benefit from GenAI, using it to amplify their efficiency and complement their existing capabilities. More crucially, they retain something novices lack: the judgment to evaluate whether AI-generated suggestions represent true innovation or sophisticated imitation.
We recommend a restructured approach to innovation teams. Novice developers, empowered by AI tools, can drive initial ideation and rapid prototyping. Experienced developers then critically evaluate the initial output, identifying which concepts warrant serious investment and which merely replicate existing solutions with superficial variations.
Here’s where prompt engineering becomes key. Current prompting techniques often focus on efficiency: extracting knowledge quickly and generating functional code. What’s lacking is crafting queries that push GenAI towards genuinely novel insights rather than optimal recombinations of existing patterns.
Organisations might consider developing specialised roles focused on prompt engineering for creativity. One solution might be practitioners who understand how to coax breakthrough thinking from systems architecturally leaning towards efficiency.
In short, experience and skill have become valuable in new ways. It’s less about implementation and more about discernment.
More and better
GenAI has sparked an innovation boom. But without intentional strategies to push beyond the tool’s default behaviours, we risk a bumper harvest of slim pickings. The future of AI-enabled innovation will be defined by how we harness the technology to produce not just more, but better.
