Building a Palantir Practice

How and why you should

Palantir Image

Palantir is quickly gaining steam in the US commercial enterprise software market. The company has a well-established track record of delivering massive value in record time for the world’s largest institutions, and there’s a reason for that. Palantir represents a new type of platform that sits above the HyperScaler layer. Often referred to as Supercloud or Sky Compute, this layer enables enterprises to skip the time-intensive (and costly) process of platform engineering and go straight to business value. It also enables a cloud-agnostic approach and interoperability.

Palantir runs on all major clouds (and on-prem) and orchestrates the HyperScaler on your behalf. Palantir Foundry/AIP offers a complete solution to an entire category of engineering problems, including (but not limited to) distributed system orchestration, eventing, applications (and supporting databases), data analytics, and data science. Typically, companies have to engage in months, if not years, of undifferentiated heavy lifting to provide these capabilities themselves by deploying and orchestrating HyperScaler services and point solutions. The process is error-prone and a bottleneck to the use cases the business cares about. In many cases, it would make more sense for the business to pivot to a software company rather than amortize the investment over a customer base of one, but most (if not all) never get to that point.

This is why Palantir is winning. They enable leveraging a Sky Computing platform to focus the enterprise on solving its hardest problems within a timeframe that makes the solution relevant. Companies also want to unwind their sprawling array of SaaS products and build bespoke solutions that align with the business’s mission, not the SaaS shareholder returns. This is also why if you are not considering building a Palantir practice, you should be. But it’s not the only reason.

AI is Rewriting Software

Software has been forever changed with the introduction of AI. While many buzzwords are floating around to describe this tectonic shift (like agents or agentic), they fail to capture the true essence of the shift. AI is enabling two new capabilities that will forever reshape the software landscape:

1. Unstructured inputs to the API economy — legacy software places an extreme cognitive load on users. The end user is expected to enter precise and complete information into the correct places, interact with the screens, and transitions in precisely the correct way to get to the outcome they desire. Those days are over. Large Language Models (LLMs) are, for the first time, making it possible to interact with software systems with incomplete, nonprecise language. For example, check out this video demonstrating a personal assistant workflow using speech-to-text in Palantir Foundry.
2. Loosely coupled non-deterministic control flow — Legacy software requires software engineers to encode the possible states an application could reach before the software is compiled and distributed. This is done through what is known as control flow and state management. How the developer implements these details dictates the possible paths the application can explore and the possible states it can end up in. This limits the number of workflows an enterprise software system can support in a single release. Using neural nets as a replacement to (or in addition to) traditional control flow allows models to reason about the next best action based on the user's intent and what has happened so far in the execution. I created X-Reason as an example of incorporating this pattern in web and mobile web applications built in React.

Collectively, these capabilities enable Text-2-Action, which could rewrite the textbooks on enterprise use cases. Former Google CEO Eric Schmidt has stated, “When they're delivered at scale, it’s going to have an impact on the world at a scale that no one understands yet.” I tend to agree with Schmidt.

Winner Take Most World

The old pixels of legacy software are not flexible enough to deliver the same advantages as a complete ground-up rebuild with AI at the core. This represents a tremendous opportunity for those who can find a cost-effective way to develop bespoke offerings at a record pace. The first mover advantage for those who can take cost out of the business leveraging AI software is enormous. Palantir Foundry/AIP is fully differentiated in this regard and comes with all the AI building blocks required to make it possible.

For a complete breakdown of the Foundry/AIP platform and how it accelerates the delivery of AI software, read my previous article.

Value Delivered vs. Dollars Spent

Many detractors of Palantir often cite three perceived problems:

1. It’s a black box (no, it’s not)
2. It’s expensive (compared to what)
3. It’s vendor-locked (good)

For a complete response, read my previous article. Foundry's perceived high cost is the biggest obstacle to adoption. While Palantir can be expensive on paper, it is often cheaper when applied to activities-based costing (ABC) models. Further, the most expensive solution is the one that delivers the least value, not the one with the highest sticker price.

Enterprises should do due diligence when unpacking the cost of a solution (do a TCO analysis) and developing the value tracking mechanics. A good primer on how to do this can be found here. Without a good understanding of the value a system is delivering, enterprises will spend bad money before good and continually fail to deliver business value.

Palantir makes it easy to track the value delivered by their solutions, deploying both metrics and analytics and measuring pre/post-deployment levels of critical business KPIs. Partners like PwC also deploy their own value-tracking mechanics and use frameworks like Foundry Analytics Foundations to provide custom reporting.

Customers can cut professional service fees substantially by leveraging an updated operating model that focuses Palantir resources on solution components south of the Ontology. In contrast, their employees focus on developing components north of the Ontology and provide feedback to the Palantir team regarding what Ontology objects and features are required for upcoming sprints. An operating model similar to the following could reduce spending by 15–25% while accelerating system delivery and scale of adoption.

Proposed Operating Model for Palantir Customers

Conclusion

Now is the time to build your Palantir Practice. Businesses that fail to future-proof their margins will have them beaten into submission by those that do. Palantir customers already have a tremendous head start in this space. A failure to acquire Foundry talent and ecosystem partners now may result in those resources becoming unavailable for future initiatives. Also, current policy makes securing supply chains with AI and automation more critical than ever. Don’t wait. Start your Palantir practice today!