This is a supplement from Part 1: The Trajectory of AI LLM Tools in Software Development, where I laid out a case for how these emerging platforms will be our future partners in development. This is a brief summary of ten areas of focus for software leaders committed to understanding and leveraging AI LLMs for software production.
Learn. By using a variety of AI LLMs across different development use cases leaders can gain a tangible feeling for how they might help and how quickly they are evolving. If you aren’t already deeply in a pilot, you’re late.
Accept that a new partner is coming. The best LLM tools are already phenomenal assistants for tactical development needs. Engaging in the spirit of a potential future partnership – albeit a subordinate one, for now – is a first step. I dig into this partnership concept in Part 1. Undoubtedly most people will find this notion too great of a leap, too quickly.
Exploit the opportunity: Don’t wait. Inaction by leaders will emperil the long-term sustainability and special position in the development value chain our human talent holds. By learning about the specific potential of these tools and reorienting the organization, they will help their people maintain relevance and improve their position vis a vis the coming “coopetition.” This is more than a classic case of competitive differentiation, it’s an imperative for the viability of our talent.
Clarify the boundaries of where and when people matter: Engage staff members to best imagine how human talent continues to create differentiated value as their direct contribution domain shrinks. Continue to push the team to the frontiers of strategic development creativity while delegating labor-saving activity to AI LLM tools. You can’t establish a productive partnership – nor a competitive advantage – without knowing the capabilities and growth rate of the participants.
Recalibrate adoption cycles: Evaluation, re-evaluation, and adoption cycles for these tools requires rethinking the traditional pace of the org. Adapt to the new fast. Tools that were insufficient in the fall may clear the bar spectacularly by spring.
Recruit and reshape talent: Talent filtering, training, and performance management must be changed for organization to meet the need and continue to adapt. Candidates must know how to leverage the latest AI LLM tools and understand how to effectively partner with them. Set expectations for competency and push the bar ever higher. Ensure your recruiting team has transparent and up-to-date policies regarding use of these tools during the interview cycle. Expect your job candidates to use these tools, even during interview coding tests. Push candidates to tell stories about how they use these tools to understand their capabilities. Your developers who are most capable with the tools will eclipse others in terms of productivity. This exposes a risk around managing velocity.
Prepare for increasing velocity. A ramped-up team that leverages these tools will generate code more quickly. Higher velocity is a good problem to have, but it’s still a problem. Your backlog will drain faster, bugs will come quicker, and your end of cycle tests will become a bottleneck if you haven’t pulled quality upstream and automated, automated, automated.
Policies, ethics, governance, and human oversight: An AI assistant can help spot biases and ethical risks, but there is abundant evidence of human bias in AI LLMs, producing bizarre distortions of reality. We can expect grad school case studies in future MBA ethics courses thanks to the recent release of the comedically flawed Gemini, by Google. Who – or what – can be trusted to oversee what comes out of your combined product development efforts? Will these high-potential tools be regulated to a place of banal servitude or unleashed to some dystopian extreme?
Intellectual Property: Select tools carefully so that you don’t put corporate IP at risk. Are your developers pasting core IP (code) into AI bots today? There will be mountains of litigation before we have clarity here.
Communicate: The truth sounds a bit scary. This can be an opportunity or a crisis depending on how you shape the message. Lead from the front by preparing the organization. Things are changing, quickly. The world will always need software. Developer roles will shift to new opportunities in the value chain.
There's no stopping this train. Get on board early!