Beliefs on AI

Cacheflow · June 2024

Outline

This deck outlines Cacheflow’s perspectives on AI and its impact on our company, product, and market.

We believe that a shared set of principles will guide us in creating thoughtful, efficient, and practical solutions. By aligning on these principles, we can navigate the complexities and uncertainties of AI development with a unified vision, ensuring our efforts are focused on delivering innovative and impactful AI-driven products.

Embracing these beliefs allows us to transcend the hype and disillusionment often associated with emerging technologies. Our goal is to integrate AI in ways that genuinely solve business problems, enhance user experiences, and provide greater value to our customers.

Our approach combines the strengths of traditional UI alongside the efficiencies of AI, enabling us to deliver products that empower users while leveraging the latest technological advancements. By viewing AI as an enhancer rather than a replacement, we aim to augment human capabilities and provide superior solutions to our customers.

​

It is essential for everyone to share the same mission, which can be achieved by agreeing on a common set of beliefs.

The challenge lies in the difficulty of extrapolating traditional HCI patterns into this new reality, marked by high uncertainty and a lack of standards and guidance. This makes it more difficult to agree on direction as we create products and services if we do not fundamentally share a point of view on the most uncertain elements.

We need a shared point of view

The disappointment after the initial AI hype should not deter us from embracing the new paradigm.

Trough of Disillusionment: Being disappointed at this stage should be expected. Our belief is that this is a temporary state, mainly due to the staggered nature of progress, and it should not deter us from embracing this new paradigm.

Trough of Disillusionment (Gartner) shouldn’t block us

Utilize AI only to address genuine business problems. Simply offering ‘AI features’ will not provide a competitive edge.

AI is quickly becoming commoditized and there’s not much needs for providing access to end-users “trying out AI” anymore. It’s quickly moving into productivity/tooling.

AI for business problems

We will avoid any inflated statements about AI, ensuring our communication remains genuine, and grounded in our actual knowledge and work.

To help us build trust internally and externally. There’s enough inflated statements in the market already.

No fake posturing

Our current ‘traditional UI’ empowers users with control, while AI will enhance their efficiency. We need both.

Traditional “imperative UI” (step-by-step detailed) is excellent at making users feel in control. Gen AI assistants excel at making users feel efficient but often at the expense of precision control. Cacheflow is uniquely positioned to combine both strengths.

AI is not perfect

AI will augment human efforts rather than completely replace them in mid-market+ enterprise B2B sales.

Creating business proposals requires precise communication and understanding of the buyer’s needs, which AI alone cannot fully achieve. AI can assist by providing drafts and suggestions, but human oversight and editing are crucial for accuracy and personalization. “Larger deals”: $100+ ARR. Many larger deals are closed not on the calls or booked meetings but in other contexts where business leaders meet and socialize. Harder for AI to take over and collect context.

AI as an enhancer, not a replacement

The interplay between Cacheflow and OS-installed, native AI assistants will become critical.

We believe the omnipresence of always-on AI assistants is inevitable. The ease and benefits are so substantial that OS-installed assistants will become the primary, and often only, stop for user assistance across work tasks. We should not envision a future where this new layer isn’t integral.�See Microsoft Co-pilot + screenshare Minecraft game.

OS-installed assistants are part of the solution

More powerful AI models will make Cacheflow stronger and better.

We do not believe that even much more powerful models can replace salespeople in the next few years for most enterprise B2B sales. There is too much out-of-context knowledge in business and personal relationships that translates to missing context for the models. Even with infinite power, there is not infinite understanding. However, infinitely powerful AI assistance will significantly enhance the salesperson’s ability and efficiency.

Better AI models, better Cacheflow

Planning for iteration and experimentation is critical to achieve success and build understanding.

New text-to-action models can automate more complex tasks, reducing the burden on users. Cacheflow should experiment with these (and other) models to streamline processes further and improve productivity.

Iteration and experimentation is key