Is AI killing vertical SaaS? What private software owners should do

What is the debate actually about?#

The debate about AI killing SaaS is about whether software agents will do work that people currently buy seats and interfaces to do. Investors, operators and commentators have argued both sides openly, and private vertical software owners now hear the question from boards, lenders and prospective acquirers.

The useful answer is narrower than the headlines. Agents change how work is done inside software; they do not remove the need for a reliable record of what was done, for payments to clear or for regulated steps to be followed. The question for an owner is which parts of the product are which.

Which parts of a vertical SaaS product are most exposed?#

The most exposed parts are the layers that sit between a user and a database without adding judgment, data or obligations. If an agent can read the screen and fill the form, the form itself becomes less valuable.

Few products sit entirely in one row. Most vertical platforms combine exposed layers with protected ones, which is why the owner's job is to know the mix rather than to pick a side in the debate.

What protects a vertical software company?#

What protects a vertical software company is the part of the product an agent would need to call rather than replace. The system of record, embedded payments, industry-specific compliance steps and connections to niche hardware or trading partners all fall in that category.

Workflow history protects too, and it is the least understood defense. Years of linked records showing how cases were handled in an industry, such as repair orders and parts decisions, permits and inspections, or loads and exceptions, are hard for a newcomer to reproduce. They can train better in-product features and, where the vendor owns them, may be licensable.

Customer trust and support depth matter as well. Small businesses in a vertical often buy from the vendor that understands their trade, and that relationship is documented in support and implementation records only the incumbent holds.

Is your data a moat or an asset to license?#

Whether data is a moat or a licensing asset depends on who owns it, how fresh it must be to matter and whether a licensee could use it to compete with you. Many owners assume they must choose one answer for everything, but the answer often differs by record family.

A common pattern is to hold live tenant data as the moat while licensing historical, company-owned records on non-exclusive terms restricted to a field of use. That combination rarely gives a competitor what it would need to replace the product.

Three moves for private software owners#

Three moves cover most of what an owner can control: know the records, build on the rights already held, and make the license-or-hold choice deliberately rather than by default.

None of these moves depends on predicting how far agents will go. Each improves the company's position whether the market treats AI as a threat or an opportunity, and each produces documents an acquirer will ask for anyway.

• 1Map your records: list each system, record family, years of history, owner and contract restrictions, including archives from retired tools.
• 2Build AI features on rights you hold: start with in-tenant features that use each customer's data for that customer, then consider shared models under clear terms.
• 3Decide license or hold by record family: hold customer data and live feeds, and evaluate non-exclusive licenses of company-owned history with field-of-use limits.

How will boards and acquirers read your position?#

Boards and acquirers will read your position through evidence, not adjectives. Claiming a data moat without an inventory, a rights map and examples of how records improve the product invites skepticism in diligence.

Expect questions about AI exposure in pricing, which features rely on outside model providers, what customer contracts allow, and whether any data licenses exist and on what terms. A short written answer to each, backed by the inventory, persuades more than a strategy slide.

Illustrative: an auto repair shop software company answers its board#

Illustrative: a fictional vertical SaaS company sells shop management software to independent auto repair shops, and its board asks whether agents will replace the product. Leadership maps the business: repair orders, estimates and parts orders in customer tenants; payments processing; parts supplier integrations; and company-owned support tickets, engineering history and implementation playbooks.

The team concludes that the system of record, payments and supplier integrations are hard to displace, while seat pricing is exposed. It ships in-tenant features that draft estimates from each shop's own history and pilots pricing based on repair orders processed.

For data, it holds tenant repair orders as the moat and evaluates a non-exclusive license of its de-identified engineering and support histories, limited to model training and evaluation. The board receives the inventory, the rights map and the decision in writing.

How SourceX supports a license-or-hold decision#

SourceX supports the decision by assessing company-owned records against the SourceX Enterprise Data Value Framework before any commitment. The first fit check collects metadata, such as systems, years of history and record families, and nothing is shared at that stage.

If an owner proceeds, the SourceX five-step transaction keeps the company approving each step from Supply to Delivery. Data is licensed rather than sold, so ownership and future options stay with the company.