Remember the old way of adding a Custom Colors section to PowerPoint? It often involved a complicated process of hacking XML files, unzipping presentations, and manually editing code.

Fast forward to today, and AI 🤖 has completely changed the game.

I’ve been exploring a workflow using ChatGPT-5 that makes this once-tedious task incredibly simple.

Here’s how it works:

↳ You provide a list of your Hex color codes to generate the necessary XML code.
↳ You can then upload your presentation and have the AI insert the custom color theme directly into the file for you.

What used to take deep technical know-how now takes just two simple prompts. In the final file, your new palette appears perfectly in a dedicated “Custom Colors” section within the Shape Fill menu. This is a fantastic example of how AI is streamlining our workflows.

See the complete visual guide on Figma Slides and give it a try: https://hub.ontopraxis.ai/custom-colors-ppt

Jack Dobson: Well, as mentioned, this is DevCon, one of one. To kick us off, why now? Why have we started DevCon at this point, and what’s the journey that led us to today?

Shyam Sankar: I'm so excited to kick off our first DevCon here. It’s really the culmination of a lot of work and focus on the product side over the last two years to make this a developer platform. If you think about Palantir, we hold sacred the primacy of winning — thinking backwards from first principles to manifest outcomes that matter, not just inheriting market assumptions.

There’s a legitimacy crisis in the world today — doors falling off airplanes, sclerotic institutions, entropy in the system. Palantir’s journey has been about counteracting that.

Over the last two to three years, we’ve seen that to honor this primacy of winning, we need to extend our culture and tool chain to developers like you, our customers, to build these solutions. For example, 700 uniformed service members from the 101st Airborne built a search-and-rescue operating picture for Hurricane Helene in 24 hours. These aren’t trained computer scientists, but builders who saved lives with code — something unimaginable a decade ago. Now, we’re externalizing our tool chain, tradecraft, and culture, making you all forward-deployed engineers.

Jack Dobson: Incredible. We’ve seen the products develop over time, but it’s not just a product story. There’s a mindset shift in approaching problem-solving and building solutions. Have you seen our customers develop that mindset and intuition over time?

Shyam Sankar: Yes, ideas we earned inductively in the field that were once heterodox are now truisms. Ten years ago, it was radical to say that separate systems for insight and operations were a bug, not a feature. People were comfortable in their data science silos, disconnected from the factory floor, screwing up the OODA loop. Now, it’s accepted that data isn’t the new oil — it’s the new snake oil if you can’t act on it. Customers are now receptive to absorbing the trade craft and approach behind our software, making the timing perfect for DevCon.

Jack Dobson: On that note, we’ve seen this evolution across customers in commercial and government sectors. Is there a step-by-step process to get to that mindset, or do you just have to break through?

Shyam Sankar: There’s no process — only the content. You have to work backwards from the problems and roll up your sleeves. The AI revolution is experiential; those experimenting and iterating from prototype to production are learning the fastest and getting the best results. Hand-wringing, commissioning studies, or debating use cases leaves you in the dust. Look at Europe versus the U.S. — the performance gap reflects this difference in approach.

Jack Dobson: Speaking of AI, the model landscape is evening out with incremental progress, but deployment still has leaps and bounds to go. How is Palantir thinking about AI at this stage?

Shyam Sankar: Borrowing from Stephen Cohen, one of Palantir’s co-founders, there’s AI supply and AI demand. On the supply side, models are improving but converging —closed and open-source models are getting more similar, and the rate of improvement is leveling off. The real gap is on the demand side: how do you use these models to achieve economic value?

We’re in the early innings of building the necessary tool chain. It’s like the shift from analog to digital computing — we’ve abstracted zeros and ones over analog waveforms with error correction and hardware/software ecosystems. With LLMs, we’re in an indeterminate phase, wrestling with stochastic outputs in a historically deterministic tool chain. We’re ahead in providing developers the tools to get from prototype to production fast. Look at self-driving cars: the 2005 DARPA Grand Challenge was a great prototype, but it took 20 years to handle city driving. We need to focus on getting to production quickly as a community.

AI as Labor for Automation

Beyond copilots, AI’s value is in automation and enterprise autonomy, decomposing workflows into human, expert, and software roles.

Jack Dobson: We announced AIP in April last year (2023), approaching its first full year. How has our approach to use cases evolved, especially as we move beyond chat?

Shyam Sankar: The last few months have solidified my conviction that the value lies in automation and enterprise autonomy. Chat is a dead end — charismatic but limited. A copilot might boost productivity by 50%, but the real economic value is in treating AI as a type of labor. Decompose problems into what humans, subject matter experts, workflows, and traditional software can do, then string them together. The day-two problems— telemetry, agent behavior, refining edge cases —become critical. Automation is limited by edge cases, not the happy path. Our tool chain helps developers capture feedback and handle those edge cases fast, promoting users to managers of AI agents.

Jack Dobson: Here at DevCon, people will be building over the next couple of days. Historically, some of the strongest workflows I’ve seen at Palantir were built by customers— experts empowered by the tooling. What examples have you seen of customers evolving from users to agent governors? Are there leaders or laggers across commercial and government sectors?

Shyam Sankar: The use cases I’m most excited about are where humans aren’t perfect and time is the enemy. On the government side, machine-assisted disclosure of intelligence sharing stands out. Sharing intelligence with foreign partners used to take three days — humans reading guides and raw data. In three days, the missile’s already landed.

Working Backwards from Outcomes

Shyam emphasizes solving real problems over following rigid processes, a mindset shift critical for impactful development.

Now, LLMs rewrite intelligence, cite the guide, and flag where humans should QC (Quality Control), cutting it to three minutes. Humans were imperfect even with three days, so this high-pressure use case drives pragmatic adoption. Beyond copilots, which boost the median worker, AI levers up the best humans — core drivers of productivity — institutionalizing their impact across workflows.

Jack Dobson: As people leave DevCon and return to their organizations, there’s chaos in the creativity required to deliver this. How do you see organizations changing as they adopt AI — will they seek protocols, or should they brace for more fluctuation?

Shyam Sankar: Brace for creative chaos. Organizational boundaries from the pre-AI era won’t hold. The winners will reimagine how they’re organized, pushing technical expertise into operations. Take Hurricane Helene: 700 service members tailored software like an Ironman suit to meet the moment — not formal developers, but empowered builders. AI liberates developers to move into the business, focusing on high-impact outcomes, not boilerplate code.

Creative Chaos as Opportunity

Organizations must embrace disruption, reimagining structures and pushing technical expertise into operations to succeed in an AI-driven world.

Jack Dobson: Looking inward, how do you see Palantir’s culture changing with forward-deployed engineers and customers as builders?

Shyam Sankar: I tell our forward-deployed engineers: developers are now your customers, just like embedding in Detroit or Djibouti. Metabolize pain and excrete product. I want DevCon attendees to be perpetually dissatisfied — your complaints signal what the platform needs. Internally, AI requires reinventing tradecraft from first principles.

Clinging to old workflows gets marginal gains; maximalist reimagination creates opportunities — like three-month hires leading big projects over three-year veterans.

Jack Dobson: We’re rolling into 2025 soon. What are you most excited about next year?

Shyam Sankar: Automation — Project Autopilot internally. It mirrors our government-side Target Workbench, automating high-stakes targeting processes. We’re bringing that trade craft into a tool chain for developers to build automations, govern edge cases, and create feedback loops, instantiating armies of AI agents.

Jack Dobson: With our last few minutes, what should builders and technical leaders at DevCon bear in mind over the next two days?

Shyam Sankar: Focus on what makes you different and special — what creates alpha in your institutions. AWS deprecated infrastructure; our tool chain can deprecate backend development. That’s hard, undifferentiated work. The advantage lies in applications that run your business. Our declarative approach lets you compete on speed and focus on what sets you apart. Let your ambition run unbridled.

Disclaimer: The views expressed by the speakers are their own and do not represent ontopraxis. All insights are attributed to the speakers; any inaccuracies are my responsibility.

The podcast captures key highlights:

↳ enterprises embracing tailored AI models with a focus on governance, 

↳ the persistent challenge of moving AI past proof-of-concept due to data silos, and

↳ the intriguing value-shift pyramid—where infrastructure dominates today, but applications will lead tomorrow. 

↳ It also spotlights Databricks’ unique edge with its unified, multi-cloud platform.

Listen to how the AI hosts break it down - offering a fresh take on an already compelling discussion.

Banks are projected to spend $6 BILLION on GenAI in 2024, skyrocketing to $85 BILLION by 2030! 📈

💡 Why the massive investment? Generative AI holds the potential to:

• Boost productivity across critical banking functions

• Enhance customer satisfaction and employee experiences

• Strengthen risk management capabilities

This could unlock $200-$340 BILLION in value for the industry over the next 3 to 5 years! 🤑

The future of retail? It's not on its way—it's already in our carts!

Human + AI teaming offers solopreneurs a transformative approach to managing their businesses. It combines the strengths of human ingenuity with the power of artificial intelligence. Generative AI democratizes access to powerful tools and capabilities once exclusive to large enterprises. This levels the playing field for solopreneurs, enabling them to compete more effectively and achieve their business goals.

This partnership between humans and machines is not just a strategy for efficiency but a pathway to unlocking new possibilities and achieving business success.

Here’s how I leverage AI tools to complement my knowledge creation workflow instead of hiring FTEs or sourcing gig workers. This approach saves me the time and effort needed to find, onboard, train, and manage the talent while I build my practice.

Cost implications: Humans vs. AI

Comparing the costs of using AI tools with hiring full-time employees (FTEs) involves several factors, including salaries, benefits, and subscription fees for AI services. Here’s a breakdown of potential costs for the AI tools I use at Ontopraxis versus hiring FTEs for the roles of Researcher, Content Developer, and Designer:

Considerations Beyond Cost

1. Quality and Consistency: AI tools can automate repetitive tasks and improve efficiency, but they may lack the creative nuances and judgment that humans provide.

2. Scalability: AI tools offer scalability to handle increased workloads without a proportional cost increase.

3. Flexibility: AI tools can be switched or upgraded easily compared to the complexities of hiring, training, and possibly terminating FTEs.

4. Collaboration: While AI can assist in collaboration, human interaction, and teamwork often lead to creative breakthroughs that AI might not facilitate.

5. Skills Development: Leveraging AI tools requires a learning curve depending on the complexity and technicality of the tasks involved.

In conclusion, AI tools can offer a cost-effective and scalable alternative to hiring full-time employees, especially for startups or organizations looking to maximize efficiency and flexibility.

Remember, it's not about who's in the box on an org chart. It's about what processes and tools are in your operating toolbox. Embrace the flexibility of GenAI and let it be your virtual team.

Subscribe now

𝗧𝗼 𝗔𝗹𝗹 𝗕𝟮𝗕 𝗠𝗮𝗿𝗸𝗲𝘁𝗲𝗿𝘀:

In 2016, Bain & Company published an article, “The Math, the Magic, and the Customer,” emphasizing that successful marketing requires both data analytics (“math”) and emotional storytelling (“magic”). 👇
https://www.bain.com/insights/the-math-the-magic-and-the-customer/

Influenced by this, in 2019, we made a prediction during my stint with Dell Technologies that the 𝗳𝘂𝘁𝘂𝗿𝗲 𝗼𝗳 𝗺𝗮𝗿𝗸𝗲𝘁𝗶𝗻𝗴 𝘄𝗶𝗹𝗹 𝗯𝗲 𝗮 𝗰𝗼𝗺𝗯𝗶𝗻𝗮𝘁𝗶𝗼𝗻 𝗼𝗳 𝗠𝗮𝘁𝗵 🧮, 𝗠𝗮𝗴𝗶𝗰 🧞‍♂️, 𝗮𝗻𝗱 𝗠𝗮𝗰𝗵𝗶𝗻𝗲 🤖.

Fast-forward to 2024, and the explosive rise of Generative AI has empowered marketers to wield the power of this trifecta like never before. By seamlessly weaving together data-driven insights, compelling narratives, and cutting-edge AI, we're unlocking new realms of marketing success, captivating audiences, and driving unprecedented results. The future of marketing is here, and it's a symphony of Math, Magic, and Machine.