AI’s impact on Product: holding PMF, deploying agents for PMs, and mastering Google’s prompting strategies
👋 Hi and welcome to Brighteye Bulletin - your fortnightly take on all things in Edtech, HRTech and Talent Empowerment, both from a Startup & VC perspectives.
Articles and Resources of the week
Is PMF harder to keep in the AI era?
For years, the narrative has centered on how to find PMF; that elusive moment when customers pull your product out of your hands. But CJ Gustafson argues that today, the real challenge isn’t just finding it, it’s holding onto it. And VCs, you might have noticed that by now.
Gustafson points out that the old playbook assumed you could ride PMF for five solid years before needing to reinvent. Around year 3, you’d start launching a new product or geography to jump the S-curve and extend your growth runway. You know the playbook.
But the AI era is reshaping that timeline.
AI accelerates both success and obsolescence.
AI startups today can rocket to $50M-$200M ARR in record time, but they can also lose PMF just as fast. Some are cycling through PMF gains and losses many times within 2 years, compressing what used to be a 5-year period into 5 weeks. Teams need to be prepared for continuous reinvention.
Hitting early velocity no longer guarantees long-term success.
The once-reliable ‘triple-triple-double-double’ growth model doesn’t guarantee you’re safe anymore. Rapid growth might signal a genuine breakout, or it might mask an unsustainable surge driven by hype or timing, something only revealed once the enterprise renewal cycles begin to test retention and durability.
The AI agent playbook every PM needs
In the latest guest post on Lenny’s Newsletter, Tal Raviv delivers one of the sharpest guides yet on how AI agents can reshape product management.
Raviv argues that while PMs juggle endless “everything is my job” responsibilities, too much time gets lost on repetitive but necessary tasks (e.g., syncs & coordination). Enter AI agents: not just chatbots, but systems that listen, plan and act autonomously.
Key takeaways:
AI agents are not chatbots; they are action-takers.
The most promising systems today proactively plan, tap live data, take real actions inside tools (like updating CRMs or sending Slack messages with Zapier integrations), and self-correct over time.
The biggest wins right now are in targeted automations, not full replacements.
Think judgment-light but time-heavy tasks: surfacing meeting context, monitoring competitor signals, summarizing routine feedback. Tools like Zapier, Lindy AI, Gumloop, and Relay App are quietly becoming indispensable.
Start small, reduce risk, and iterate.
The smartest teams aren’t automating entire processes overnight. They begin by isolating the most painful step in a workflow, designing an agent to handle only that, and tightly constraining permissions to keep risk low. For example, set agents to draft outputs, not take irreversible actions.
Effective automation still requires deep human clarity.
Agents succeed only when teams can clearly define the task, the relevant data sources, and the criteria for success. Poorly scoped or vague requests will fail.
Be cautious about outsourcing too much intuition.
While agents can accelerate data gathering and pattern detection, they should not replace first-hand exposure to raw customer signals. Raviv emphasizes that over-reliance can degrade a team’s real market sensitivity.
Google dropped its ultimate prompt engineering playbook
Google has released what could become the go-to guide on prompt engineering, and it’s essential reading for anyone working with LLMs. This guide is packed with actionable strategies to generate reliable AI outputs.
Prompt engineering is quickly becoming a core competitive edge. Whether you’re using tools like Replit, Lovable, or integrating LLMs into customer workflows, mastering these techniques is crucial.
This guide covers:
LLM configuration
Understand how parameters like length, temperature, Top P and Top K shape model behavior and output consistency.
Prompting strategies
Master system and role prompting, few-shot learning, chain-of-thought and tree-of-thought reasoning to get more sophisticated results.
Best practices
Learn to craft clear instructions, apply schema and formatting techniques, and use structured templates to increase the quality and predictability of outputs.
Investment news
Berlin-based Solda, which is building autonomous AI agents capable of carrying out the whole sales process, has secured $4M in seed funding. The round was led by Accel, with participation from Altair Capital.
Oslo-based Pistachio has raised $7M in Series A funding led by Walter Ventures, with participation from Idékapital, Angel Invest, MP Pensjon, and J12 Ventures. The startup offers a fully automated platform intended to provide cybersecurity training for employees.
London-based Dex, a platform offering an AI voice talent agent, has secured $3.1M in pre-seed funding. Concept Ventures and a16z speedrun led the round, supported by BAs including Charlie Songhurst (Meta), Nilan Peiris (CPO, Wise) Eric French (COO, Deliveroo) & Stephen Whitworth (CEO, Incident.io).
London-based Capably has closed a $4M seed funding led by Boost Capital Partners, with participation from Concept Ventures, Sure Valley Ventures, Araya Ventures, Haatch, Koro Ventures, Wayra, and Ascension. The company allows for the deployment of AI agents across organisation to delegate complex and routine tasks.
Jobs
Ornikar is looking for a Head of Brand (Paris).
Uphill is looking for a Account Executive, a Senior Product Designer and a Senior Product Manager (Lisbon).
Genially is hiring a Customer Success Specialist (remote).
Hackthebox is hiring a Senior Software Architect (Athens), a Senior Customer Success Associate (Singapore) and many others.
Thank you for reading! If you enjoyed it, please share with friends and colleagues interested in European Edtech, HRTech and Workforce Empowerment. We also explore the AI space - stay tuned for more updates!
Onward & upward!