Software partners earn little from referrals. AI may change that

Software companies that sell through partners may need a new plan. A typical referral deal pays a partner only 10-15% of first-year revenue. An Entrepreneur contributor argues that AI could create better-paid work for those partners, and that software makers should build their programs around it.

What changed

Cloud software, often called SaaS, made programs easier to buy and set up. Much of the work around them became repeatable, so partners drifted toward lower-margin services. The author says 10-15% is rarely enough to cover sales, setup and future growth. Partners without a profitable business around a platform will invest elsewhere.

AI, the author argues, brings back high-value work built on the platform and shaped by each customer. It is sold as expertise rather than access. Some pieces, such as connectors and managed operations, will become standard. The judgment calls will not.

Where partners fit

Those calls include which workflow should change, which data can be used, where human review belongs, and who is accountable when an answer is wrong. The examples given: a bank needing an AI-assisted compliance workflow with audit evidence, a hospital system needing a privacy-aware intake process, and a law firm needing limits on what an AI agent can recommend.

The author says most jobs will involve workflow design, data rules, integration, testing, governance and adoption, not training a new AI model. A Futurum Research survey for the second half of 2026 found 71% of active partners named custom app building as a primary growth driver. The author warns that not every custom build will be profitable.

What software makers are told to do

The advice is to ask what partners can build profitably, not how to get more referrals. Add up referral and renewal income, plus profit from partner-delivered setup and AI services, then subtract the cost of winning and serving the account. If the margin is healthy, the program has a base. If not, a higher referral percentage repeats the problem. The program should also give partners:

• A test environment that resembles real production, not just a demo login
• Training on integration, testing, governance and escalation
• Certification against real use cases
• Early access to harder projects for the strongest partners

How to measure success

Referred sales still matter but are only part of the picture. The author suggests tracking services revenue and margin where visible, time to production, active use, support burden, retention and the customer outcome the solution was meant to improve. Incentives can be tied to milestones partners can influence, such as a verified deployment or a retained account after a set period.

Limits to keep in mind

This is an opinion piece, so the view is the author's own. Vendors still need clear data rights, guardrails and support limits. The author also admits some AI work will become a commodity. The argument is that partners hold a stronger role while the work depends on customer data and practical judgment.