Exploring AI’s Transition: Services vs. Tools
In the ever-evolving landscape of artificial intelligence, the conversation surrounding AI tools and services is becoming increasingly significant. A recent discussion with a founder regarding Sequoia’s insightful article, “Services, the New Software,”, prompted a deeper examination of this shift. The essence of the article is clear: AI companies that focus on delivering outcomes, such as ‘autopilots’, are poised to dominate the services sector, as opposed to companies that simply offer tools, like ‘copilots’.
The best strategy is to start with outsourced, intelligence-heavy tasks as a wedge before expanding into higher-judgment work.
For instance, consider Harvey, which initially introduced its copilot as a mere tool for law firms, but is now transitioning towards offering completed contracts and filings. This exemplifies a broader trend where specialized agents are emerging, beginning as tailored solutions for specific tasks such as documentation writing, creating animated videos, and sales reporting. We are witnessing a surge in such agents tailored for various roles, from CFOs to CMOs, signaling that this trend will continue to expand across all professional domains.
One recent launch that caught my attention was Manicule, an innovative AI-driven technical documentation studio. By blending agent-based services with human expertise, they have developed a consulting-like model which is not only clever but aligns well with the insights from the Sequoia article.
The uniqueness of this model lies in its packaging—a concept that warrants further exploration. I wonder about the practical applications of agents in crafting documentation specifically tailored for agents to utilize effectively.
Another founder I spoke with is working on planning and specification for development workflows. Many in the field have emphasized the importance of employing a ‘planning mode’ before diving into coding. As I observe more developers creating their own orchestration platforms, it seems that proper planning may be the key to achieving meaningful progress.
If teams—along with AI agents—can collaborate on specifications, investing time in considerations related to trade-offs and implementation details, implementation should become a manageable task. Most of the preliminary effort is invested in planning, making code reviews significantly easier, as referring to the original specifications will often clarify the logic behind the code far better than sifting through the code itself.
Looking ahead, as models reach certain thresholds, it’s plausible that code reviews could become obsolete, though this remains to be seen.
* Sponsors who make this newsletter possible 🙂
Want to partner with us for the next quarter?
Email us at shanice@bensbites.com or k@bensbites.com
