AI Gives SaaS Partners a New Revenue Layer Beyond Referral Fees

Seventy-one percent of active SaaS partners named custom application development as a primary growth driver, even as their sentiment toward packaged enterprise software weakened, according to Futurum Research's 2H 2026 partner survey.

That survey of partners, not end customers, lands in the middle of a squeeze described in an Entrepreneur contributor's analysis. A typical SaaS referral deal hands a partner 10-15% of first-year revenue. On a $100,000 first-year subscription, that is $10,000 to $15,000 — rarely enough to cover sales, implementation and the cost of staying in the account.

The math thinned for a reason that has nothing to do with vendors repricing commissions. Software got easier to buy and deploy, and much of the surrounding work became repeatable. Partners moved toward lower-margin services. Without a more profitable business to build around the platform, the piece argues, they will invest elsewhere.

What does AI give partners back?

The argument runs that AI returns something SaaS gradually took away: high-value work built on the platform, shaped by the customer's context and sold as expertise rather than access.

Not every AI engagement will be unique, and the contributor does not claim otherwise. Connectors, retrieval patterns, evaluation frameworks and managed operations will become more standard. What does not travel so easily from one customer to another are the decisions that make an AI system useful: which workflow should change, which data can be used, where human review belongs, what counts as a good answer, and who is accountable when it is wrong.

Those decisions map onto regulated settings. A bank may need an AI-assisted compliance workflow with escalation paths and audit evidence. A hospital system may need an intake process that respects privacy, system permissions and clinical review. A law firm may need clause triage with firm-specific limits on what an agent can recommend.

The building blocks are familiar. Making a workflow safe, useful and adopted is not. That work depends on industry judgment built from years of being the person the client calls when something breaks.

Takeaway: The defensible work is judgment about a customer's context, not the code that connects two applications.

Where does the model break?

Not every bespoke build turns out profitable, and the survey's 71% does not say otherwise. What it suggests is that customers are looking beyond generic software for tailored engineering, and that partners see more room in services than in a shrinking resale margin.

The contributor also draws a deliberate line: this is not a case for every partner training a foundation model, nor for every SaaS company turning itself into a services business. Most customer deployments will involve workflow design, permitted data, integration, evaluation, governance and adoption — not fine-tuning.

What should vendors pay for?

The framing question shifts from "How do we get partners to refer more deals?" to "What can partners build profitably with us?"

Run the full math first: referral and renewal income, plus gross profit from partner-delivered implementation and AI services, minus the cost of acquiring and serving the account. If a partner can create a healthy contribution margin while making the product more useful to the customer, the program has a foundation. If it cannot, a higher referral percentage is only a more expensive version of the same problem.

Then make that work easier to do well:

• a sandbox that resembles production, not just a demo login
• training on integration, evaluation, governance and escalation, not only product features
• certification against real use cases
• a seat at the hardest solutions before the vendor has fixed every implementation detail in-house

Measure the result accordingly. Referred ARR still matters, but it is only part of the picture. Track:

• services revenue and margin, where they can be observed
• time to production, active use and product adoption
• support burden and retention
• the customer outcome the solution was designed to improve

Outcome-based incentives can work too, if they attach to milestones a partner can actually influence: a verified deployment, adoption by an agreed user group, a successful evaluation, or a retained account after a defined period.

What has to be true for it to work?

Discipline still applies. Vendors need clear data rights, guardrails, support boundaries and a settled view of who is accountable when a workflow fails. Partners need room to earn services revenue without creating a maintenance burden the vendor cannot support. The piece treats these as operating details rather than reasons to keep the old model.

One caveat for readers: the survey is labeled for the second half of 2026, and the piece does not disclose its sample size, margin of error or how "active partners" is defined. The 71% figure is a directional signal about where partners expect growth, not a measurement of partner profitability.

Takeaway: Vendors that track only referred ARR will miss the margin that partner-delivered AI services create.

What does this mean for SaaS founders?

The point is not that AI work will never commoditize. Some of it will, and it should. The claim is that AI gives partners a larger, more defensible role while the work still depends on customer data, workflow knowledge and practical judgment. A program built around that role is harder to reduce to a commission table, gives partners more reason to invest, gives customers more value from the platform, and gives the vendor something more durable than a percentage to compete on.