Vertical Software AI Playbook - From System of Record to System of Action
We have spoken with hundreds of vertical software founders over the last twelve months. Three groups have emerged.
The laggards have not seriously thought about AI. They still treat it as pixie dust, discuss its limitations and hallucinations, and ship a chat box built on top of a dashboard.
The performers have figured out how to use AI inside their own company. Engineering ships faster. Customer success answers tickets faster. Sales gets cleaner notes. Real productivity. Real margin. No real change to the product.
The out-performers are doing something different. They have realized this is the moment when a vertical SaaS company can stop being a system of record and become a system of action. The product no longer just stores data, prints reports, and routes approvals. It runs the workflow. It makes the decision. It writes back into the operating process and triggers the next step. Orchestration, analytics, and execution in one product.
The gap between performer and out-performer is where the next decade of value in vertical software gets created. Most founders are looking at it from the wrong side. This piece is what we tell the ones we want to back.
Why this moment
AI usage is everywhere. Enterprise value capture is not. McKinsey’s 2025 State of AI survey found that 88 percent of organizations regularly use AI in at least one function, but only 39 percent report any EBIT impact at the enterprise level. Menlo Ventures estimates that enterprises spent $37 billion on generative AI in 2025, up from $11.5 billion the year before, and that more than three-quarters of that spend was bought from vendors rather than built internally. Customers want packaged AI workflows that get to production fast and deliver something they can measure.
Inside vertical software, the gap matters even more. ICONIQ’s 2026 AI snapshot reports that 70 percent of builders are focused on vertical AI applications and that 49 percent say their primary differentiation comes from the application layer rather than the model. The model is becoming a commodity. The advantage is now in who controls the workflow that the model runs inside.
The out-performer playbook
The out-performer playbook follows a three-rung ladder. Every founder we meet sits somewhere on it, and most are honest about which rung once we draw it for them.
Rung 1: Feature
At the feature layer, AI improves the product experience without changing who owns the workflow. This is where chat boxes, summarized notes, drafted emails, and generic copilots sit. It can lift adoption, shorten time to value, and reduce clerical friction. It rarely creates moat. The same OpenAI or Anthropic model is available to your competitor and to a horizontal startup that just raised a round. If your AI strategy ends here, expect it to become table stakes within eighteen months. You will be expected to ship it. You will not be allowed to charge extra for it.
Rung 2: Workflow
At the workflow layer, the product completes work the user used to complete. This is the rung where vertical software has a structural advantage no horizontal AI vendor can replicate. The incumbent platform already understands the fields, records, edge cases, actors, timing, and compliance logic of the job. Clio’s Manage AI is built directly into Clio Manage and converts scheduling, planning, communication, and billing into completed actions, with review and approval still in the user’s hands. Procore Agent Builder lets construction firms build agents that draft RFIs, manage submittals, and generate daily logs from inside the platform. Feature AI gives a better answer. Workflow AI gets the job done.
This rung is where the conversation about pricing actually opens up. If your product completes a task the customer used to pay a person to complete, your pricing model is no longer constrained by seats.
Rung 3: System of action
The top rung is the one most founders underestimate. At the system-of-action layer, the product owns the operating process end to end. Humans, AI-assisted humans, and autonomous agents all work inside it. Veeva is the cleanest example in regulated software: data, content, and agents on a single platform, with application-specific safeguards, validation packages, audit trails, and direct execution across clinical, regulatory, safety, quality, and commercial workflows. The product has stopped being something the customer logs into to record what happened. It has become the place where work is initiated, coordinated, and concluded.
When you reach this rung, you stop selling access. You start selling outcomes. The pricing question changes from “how many seats” to “what was the result.” The right to expand grows because you are now adjacent to revenue, not adjacent to data.
The moats are moving
If model-level differentiation is fading, the natural next question is what still creates defensibility. The honest answer is that the moats most vertical SaaS companies have leaned on for the last decade are weakening, and a different set is strengthening in their place.
The old moats were schema ownership, interface control, and time spent in the application. AI weakens all three. Schema can be replicated by a foundation model with a few well-formed prompts. Interface control matters less when the user is delegating tasks to an agent rather than clicking through menus. Time in app becomes a vanity metric when the work happens in the background.
The new moats are workflow context, operational data, and trust architecture. Workflow context is the lived depth of a domain: the decision paths, the edge cases, the data trails, the taxonomies that encode how a job actually gets done. Generic AI will not learn this in a quarter. Operational data is the high-consequence subset of customer data that sits near money movement, compliance signoff, scheduling, billing, customer communication, and regulated content. Trust architecture is the audit trails, permission models, validation packages, and human-in-the-loop checkpoints that make automation legally and operationally usable in a real business.
Trust architecture is the one most underestimated. Founders treat it as compliance work. In a regulated or high-stakes vertical, it is product. It is the reason the customer is willing to let your software touch their workflow at all.
Pricing follows the ladder
Where you sit on the ladder dictates how much room you have to monetize.
At the feature rung, the answer is almost always seats. Customers will not pay separately for AI features that look like everyone else’s AI features. Bundle them, use them to drive adoption and reduce churn, and stop pretending you can charge for them.
At the workflow rung, usage-based pricing becomes credible. The customer is consuming a real unit of work: an RFI generated, a draft completed, a record reconciled. ICONIQ’s 2026 AI snapshot shows usage-based pricing has grown to 35 percent of AI revenue models and is still climbing. The trick is to define a usage unit that is meaningful to the customer, not to the LLM bill.
At the system-of-action rung, outcome pricing is on the table for the first time. ICONIQ reports that 18 percent of companies are now experimenting with outcome-based pricing, most often tied to cost savings or revenue generated. Outcome pricing only works when you control enough of the workflow to credibly affect the result.
The five questions every vertical software founder should be able to answer
If you are honest about where your product sits on the ladder today, these are the five questions that determine whether you have a chance to climb it.
1. Are we saving the user clicks, or are we owning a job they used to pay someone to do?
2. Does our AI write back into the workflow, or does it only draft text?
3. What operational data and permissions do we control that no horizontal vendor can replicate?
4. Where in our product can we credibly price on outcomes because we actually influence them?
5. What approval, audit, and compliance architecture makes our customers comfortable letting us automate further?
We don’t expect every founder we back to be able to answer all of these questions today, the truth is, most cannot. That is the diagnostic that we can bring to the equation to help chart a course to moving up the ladder, driving more value for customers, and capturing economics along the way.
Why we wrote this
We are Wingman Growth Partners. We back lower middle market vertical software companies with $5 million to $30 million of revenue across the United States and Canada. Our team trained at some of the largest investment firms in the world before deciding the most interesting work in software was happening at a stage those firms cannot reach. We took the network, the operating know-how, and the institutional capabilities we built there and focus them on a deliberately small set of companies. We back founders who are turning their system of record into a system of action, who treat workflow context as a moat, and who are not waiting to be repriced by AI-native competitors.
If that is the company you are building, we want to meet.