Reimagining our approach to Product Engineering with Generative AI
Innovation around generative AI (GAI) is happening at an unprecedented speed. The big challenge for companies is how to move quickly to unlock the potential of GAI when no clear playbook exists.
It’s a challenge we’re spending a lot of time exploring this year. While we’ve tackled big technological transformations over the past 20 years – like reimagining our member experience as we embraced the shift to mobile – the capabilities of GAI are making previously impossible product experiences possible and are advancing at a pace unlike anything we’ve ever seen in our industry. We’re working urgently to harness the powers of GAI, while adapting to its evolving nature with agility.
To navigate through this entirely new transformation challenge, we’re reimagining our approach to engineering new features and products – including how to set up cross-functional product teams for maximum velocity, their approach to learning and iteration, and the supporting technology platforms that power them. All while keeping our culture of being “member-first” and “human-centric” as our true north. As a result, we’re delivering early innovation and introducing highly complex new features at a faster pace than ever before. And we’re doing it in a responsible way that adds value for our members and customers in their core workflows – including our latest generative AI tools in Recruiter 2024, AI-powered coaching in LinkedIn Learning, Accelerate in Campaign Manager, and AI-assisted search in Sales Navigator.
We’re creating a new playbook for building products with generative AI by addressing three core questions.
How do we set up our product teams to maximize development velocity?
The rapid pace of GAI advancements is prompting us to find ways to move with even greater urgency to innovate. Changing the size and focus of our development teams is helping us increase the speed of decision making and improve alignment. The development of Collaborative Articles, our first generative AI feature that launched in March, was an early testing ground for this new approach. We assembled a narrowly focused, cross-functional team with a shared objective: Create a new experience we couldn’t have created without GAI and work quickly to get it into the hands of members to test. This shared objective streamlined the individual workstreams and deliverables we were working towards. The team also had a direct line to leadership to receive daily feedback on their direction and next steps, which created faster alignment around critical decisions.
We’re also empowering our teams to make decisions with less need for stakeholder approvals. That’s in large part due to our culture of being member-first and human-centric, which remains our constant through-line across all of our development work. We continuously challenge ourselves to go back to “why” we are building a new feature. Staying focused on the “why” gives us the confidence in making decisions while eschewing standard development process activities – like design reviews – so that we can ship features faster.
How can GAI experiences can make our members and customers more successful?
Empowering our teams to embrace a learning mindset is a key success factor in working with GAI. We’re creating brand new product experiences powered by nascent technology. It’s impossible to know with certainty whether new features will help our customers and members in the way we anticipate. So in order to learn, our priority is to put something we think is valuable in the hands of our members quickly to test it – and then create a tight feedback loop where we can iterate and adjust.
That approach might seem easier than it truly is. It’s very common in engineering cultures to want to aim for perfection. But amazing things are possible when you are able to remove the weight of delivering something perfect in order to be successful, and instead encourage - and even reward - the mentality that things may break along the way and we’ll correct them as we go. We’ve communicated that openly and clearly with our development teams, and it’s fostering innovation and flexibility that supports us in testing and learning new things about GAI.
The good news is that we're able to get things right more often than not by bringing our knowledge about the world of work, our customers’ businesses and insights from our platform into the products our members and customers use. You can’t get that from GPT services right out of the box; it’s something we infuse into experiences and features.
What platform capabilities should we build to enable success and scale?
Engineers face a lot of unknowns in working with GAI and unknowns generally lead to a heavy investment of hours and resources dedicated to learning, testing and iterating. It's extremely time consuming work. Developing a consistent, scalable platform for GAI is helping everyone work more efficiently today.
With easier access to the right tools and knowledge, everyone can innovate with urgency and consistency. One example is how we evaluate GAI prompt quality - going from very manual, spreadsheet-based workflows to tooling that enables structured human evaluation of prompts. Investments like these are paving the way for our new GAI developer platform at LinkedIn, which includes the necessary tools and best practices for working with GAI. Different teams have common access to technologies, can share resources and pass on learnings about usage limits, processing natural language, prompting patterns and applying AI responsibly. Teams don’t have to rediscover things on their own.
It’s also important to us to get everyone in the organization comfortable with GAI. Ideation is a key part of that. That’s why we’ve created a safe environment for learning and iterating with prompts – our Generative AI Playground – that everyone at LinkedIn can access, not just the engineering team. Collectively, these steps are helping us more efficiently develop and deploy GAI features with the urgency and scale required.
Embracing this moment of technology transformation
It’s undeniable that AI is changing work for all of us. And being an engineer in this moment of technological revolution is thrilling, as we sit at the center of building the AI-powered solutions that help our members and customers grow their careers and businesses. To capture the full potential of this transformative technology, we’ll continue revising our product development playbook to adapt alongside GAI.
Great write up Erran Berger! Check us out at Crafter to see how we're reimagining how product and engineering teams work https://usecrafter.com
🏆11x LinkedIn Top Voice | Data analyst | Data scientist | Artificial Intelligence| Certified in Data Science by State University of New York at Potsdam | IBM Certified Data Scientist.
8moAmazing article
Sr Director of Engineering @ Reddit | Board Member | Speaker | Previous - LinkedIn, HBO, Microsoft
1yLove the article, great read! It'll be fun to collaborate in a work session or panel discussion with fellow tech leaders, to get into how they're advancing their Gen AI platforms and tools while also trying to move with the same urgency on GAI features. Personally quite curious about the balance between in-house development and leveraging open source or integrating off-the-shelf workflows.
B.S. in Computer Engineering Student of 2024 at Sacred Heart University
1yThis was an interesting read. Going to be interesting on how you apply AI to product engineering
Director, Ironclad | RevOps | LinkedIn alum | UChicago Ph.D.
1yErran Berger when is GenAI coming to the Member experience?