The next frontier of enterprise AI is autonomous decision making. Companies that once became dominant by creating systems-of-record (SOR) focused on data ingestion, business transactions, and analytics now have their eyes on the next level of businesses’ Maslow’s hierarchy of needs: reasoning and decision making. Palantir is a prime example of this shift, as they recently repositioned with the 2023 - 2024 launch of AIP. CEO Alex Karp put it bluntly: “This is not a chatbot. It’s a decision-making platform…Everyone’s building copilots. We’re building the commander.” We’re thinking about the future commander as the AI system-of-action (SOA) – the abstraction layer that sits above and across fragmented SORs to produce real action and outcomes, not just data and analytics.

In a previous piece, we laid out why there is currently a unique opportunity for domain-specific industrial AI brains. The last decade saw a boom in domain-specific SORs  – Procore (construction), Veeva (healthcare), and ServiceTitan (contractors) – that replaced generalist platforms like Excel or Salesforce. These are now being significantly improved upon with AI capabilities that unlock unified integrations, semantic search, interoperable data, knowledge management, and RAG pipelines. 

Yet most startups never reached SOR status – the pointy wedge they initially focused on didn’t widely expand beyond a specific persona or workflow within a large organization. The result: a software sprawl of point solutions that has left many enterprises juggling dozens of vendors with little cohesion. Most enterprises currently operate across numerous record keeping systems that look like a mezzanine between a database and the human-level decision making engine one might hope for. App fatigue is real.

Now, with AI-powered reasoning, unstructured data ingestion, and ubiquitous APIs, there’s a leapfrog opportunity: unify fragmented workflows into an SOA. SOAs won’t just give you another dashboard or map, they’ll abstract away the low-level signal and instead offer to take the wheel and help steer. True SOAs will reduce operational friction, increase scope of design, automate manual tasks, and turn every user into a better decision-maker. In all likelihood - as every technology platform shift to date has proven - many incumbents will be displaced by more agile and forward thinking startups, and in the current moment it’s once again a jump ball for vertical market dominance. 

So how do you build a commander? 

Building a commander isn’t just about stitching together SORs. This is necessary but not sufficient – without automating and accelerating deep and specialized workflows, attempts at a new SOA won’t have staying power. A few key questions emerge that founders in this space must wrestle with: 

1. What deep workflow is the right entry point?

Assessing the massive surface area of an industry to find the right initial beachhead for a deep and compounding SOA workflow is difficult. Compliance, design review, document generation? - the list goes on. Industrial markets have dozens of distinct engineering, operations and knowledge work persons, all of whom have different degrees of interest in software and rely on different SORs. Product vision and discipline, and a keen sense of where the underlying technology is going, have never been more important. Ultimately the question becomes - what entry point grants the greatest ‘right-to-win’ to take the SOA mantle over time? 

2. What is the actual north star metric that senior execs care about?

Across industries, the typical SOA roadmap looks something like: 1) integrate with all existing tools, 2) ingest historical data, 3) model next-best actions from the ground up, and 4) recommend or execute optimally. To build a SOA that improves enterprise decision making, founders need to uncover precisely what the customers’ north star metric or objective function is, which is sadly often entirely non-obvious in organizations that are weighed down by decades of organizational bureaucracy and corporate gumpf. A deep and methodical customer discovery process is needed to surface any asymmetries in risk (i.e. does success of your solution get the economic buyer promoted, or conversely does failure mean he gets fired?) and deeply understand incentives. 

3. Who is the real economic buyer I’m selling to? 

The value in scope for SOA is much more strategic than traditional vertical software, and senior execs are paying attention. As a consequence, founders building SOAs likely need to start higher in the org chart than they would for a typical software sales motion. They also need to wrangle the inevitably numerous stakeholders and hone in on the real economic buyer. Historically, it’s been difficult to build vertical software that senior leaders actually engage with, so startups go after the lower-level users who live in Excel. With more executive-level value now in play, the requisite customer discovery period for SOA is more complex than traditional vertical software, because it resembles full fledged process engineering and change management as much as it does traditional product development.

4. What is the right sales motion?  

Industrial sectors like energy, construction, manufacturing, mining, real estate, and supply chain are inherently complex and high-stakes – making self-service or product-led growth strategies largely ineffective. Building and selling a SOA in these domains presents a unique challenge: not only are these organizations large and often slow-moving, but AI-powered SOAs will likely touch mission-critical operations in hazardous environments. When grid failures, well malfunctions, or equipment breakdowns can put lives at risk, the trust barrier is significantly higher. In short, a strategic enterprise sales motion isn’t optional - it’s essential.

5. How do you deploy?

‍Palantir leans heavily on services, but for truly scalable SOA software products, teams must strike a delicate balance: deeply integrated, but not bespoke. For those who do, the payoff will be immense. This of course raises questions of pricing and value capture - agentic systems may break conventional SaaS models and become outcomes based. Regardless of how SOAs are priced, some amount of services revenue may initially be necessary for industrial segments where customers are used to more hand-holding for implementation. 

Which domains are ripe for a SOA commander? 

1. Business processes where ROI on spend is theoretically knowable:
   For workflows where you can clearly quantify whether a dollar spent will result in greater than a dollar earned - like marketing, CAPEX, or labor allocation - SOAs can optimize for ROI. Today, the data to make these decisions is scattered, and the analysis required is costly and time consuming. SOAs will unify inputs and surface the optimal next move – or just take it.
2. Systems where steering is intuition-based today: Today, many huge industries rely on the best-guess of field engineers for optimizing operational processes and set points. SOAs will deliver a ‘universal controller’ - both for continuous and iterative processes - which can monitor and steer operations based on production data, engineering parameters, and equipment specs. Relatedly, in engineering domains like aero/astro, mechanical and electrical, design choices are largely human-driven today, and it’s impossible to consider the full universe of possibilities. SOAs will effectively be applied-scientist copilots – reaching across the numerous interdependent engineering disciplines – providing advanced simulation, trade-offs assessment and cost-integration to steer the optimal design and system performance.

Are you building the next industry vertical commander?

We're at the forefront of investing in this emerging category and are excited to connect with others who see the same potential. Our investments in Trunk Tools (document-unifying SOA for commercial construction), Revsure (full-funnel SOA for marketers) and Optiml (portfolio-wide SOA for building owners) are early bets on this future. 

If you’re building an SOA or searching for your wedge, let’s talk.