For years, agencies have had a familiar answer when a client asks for functionality that stretches beyond the website: “There’s software for that.”
Need a customer portal? There’s a SaaS product. A searchable member directory? There’s a plugin. A better way to manage volunteers, appointments, properties, applications, or leads? Someone has probably built something close.
“Close” is the operative word here.
The software might accommodate 80 percent of what the client wants. The other 20 percent gets handled through spreadsheets, Zapier workflows, manual processes, compromises, or an uncomfortable amount of explaining why the technology can’t quite follow the way the business actually operates.
Vibe coding gives agencies another answer: “We can build that for you.”
That could prove to be one of the more consequential opportunities AI creates for agencies. It expands the definition of a web project and gives agencies a practical way to create software tailored to their clients, then charge for the value that software provides.
Clients have spent years adding more technology around their websites.
CRM. Scheduling. Payments. Forms. Portals. Marketing automation. Membership software. Dashboards. Industry-specific SaaS. Internal spreadsheets doing surprisingly important jobs.
Each tool serves a purpose, but the resulting marketing and technology stack often reflects the limitations of available software as much as the business itself. Vibe coding loosens that constraint.
Instead of searching for software with the closest available feature set, an agency can describe the experience a client actually wants and use AI to generate the application.
The economics of doing so are changing quickly.
McKinsey’s research into AI-driven software development found that its highest-performing companies were seeing 16 to 30 percent improvements in productivity, customer experience, and time to market, alongside even larger improvements in software quality. McKinsey expects AI-assisted development to continue moving toward a model in which small teams direct AI agents to produce increasingly complete applications.
Gartner, meanwhile, forecasts the low-code development technology market will reach $58.2 billion by 2029, citing agentic AI and citizen development among the forces driving adoption.
Imagine an agency working with a nonprofit that runs several volunteer programs.
The organization wants volunteers to register online, which isn’t particularly difficult. But its program structure is unusual. Staff need to filter volunteers by location, availability, training, language, certification, program eligibility, background-check status, and a few fields specific to the organization.
There are perfectly good volunteer management platforms available. One might even cover 80 percent of those requirements, and, historically, that might have been enough! Building bespoke volunteer management software for one nonprofit would have been difficult to justify. Not to mention well beyond a non-profit organization’s budget.
With vibe coding, the calculus changes. That same agency could create a portal designed specifically around the nonprofit’s programs. Volunteers log in and maintain their profiles. Staff use the exact filters they need. Administrators manage records through an interface built around their terminology and workflow.
The agency hasn’t recreated Salesforce. It's built a relatively focused application that fits the client exceptionally well.
Once you start looking at client requirements through that lens, potential applications are everywhere.
A property management client could have an owner portal where customers access documents, review maintenance activity, and check property performance.
A commercial landscaping company could give customers a portal showing upcoming visits, completed work, recommendations, and open service requests across their properties.
A professional association could offer a member directory with filters specific to its industry, along with private resources available according to membership level.
A contractor could use a quoting application that calculates estimates according to its own services, materials, property variables, and pricing rules.
A franchise business could have a dashboard that gives individual locations access to their performance data while giving corporate employees a network-wide view.
Even the agency could benefit. An interactive SEO audit, website grader, ROI calculator, or assessment tool can become part of its own marketing funnel.
These aren’t necessarily $500,000 software projects, but that’s precisely why they’re interesting.
There’s a commercial difference between finding software for a client and providing it.
When an agency recommends another company’s SaaS product, the agency may make money from implementation, integration, consulting, or, best case scenario, a referral agreement. The software revenue largely belongs to the software company.
Vibe coding creates more room for the agency to own that relationship.
A custom application could carry an initial strategy and build fee, followed by recurring fees for access, management, support, and ongoing improvements. An agency could include it in a larger monthly engagement or package applications alongside websites, SEO, advertising, and other services. Specialized agencies have an additional advantage.
Build a volunteer portal for one nonprofit and you learn quite a bit about volunteer management. Build another and you start recognizing common requirements. Eventually, the agency may have a repeatable foundation it can adapt to several organizations while retaining the custom details that make each implementation useful.
The same could happen with portals for property managers, intake systems for law firms, quoting tools for contractors, dashboards for franchises, or membership applications for associations.
McKinsey sees a similar opportunity emerging for technology service providers. Its research suggests that as AI changes how software is built, service providers can create value by developing bespoke applications around a customer’s particular business context. Customers surveyed by McKinsey identified the ability to customize solutions and domain expertise among the core factors they consider when selecting a service provider.
That combination should sound familiar to agencies. They already have the client relationship. They understand the business. Now, increasingly, they can build the software too.
Vibe coding reduces the technical barrier to creating software. It doesn’t eliminate the thinking required to create good software. If anything, that makes the agency’s knowledge more valuable.
The important questions come before the prompt. Who will use this? What should they be allowed to do? What information matters? Which steps are unnecessary? How does this fit into the client’s existing customer journey? What would make someone return to it?
An AI coding tool can generate a filter, but only an agency understands which filters belong there. This is where vibe coding fits naturally with work agencies already do. Agencies translate business goals into digital experiences. They understand brands, customers, conversion paths, content, interfaces, and client expectations.
AI gives that expertise a larger canvas.
There is, however, an important difference between making something impressive in 20 minutes and selling it to a client.
TechCrunch has documented the tension already emerging around vibe coding. Developers interviewed by the publication praised the speed of AI-generated development while also describing the additional oversight required when generated code moves toward production. Gartner makes a similar point about low-code and no-code development: speed needs to be accompanied by governance if organizations want to produce secure, high-quality applications at scale.
For agencies, the question shouldn’t only be, “What can this generate?” Instead, they have to ask what happens afterward.
Can multiple team members work within the platform? Can clients receive access to their projects without receiving access to everything else? Can the experience be presented under the agency’s brand? Can permissions be managed?
Then look underneath the application.
Who handles hosting? Where does the data live? How does authentication work? What happens when different users require different access? Who is responsible for maintaining those foundational systems?
The more web apps an agency sells, the more consequential these questions become.
There’s an appealing simplicity to asking AI to generate everything. Some parts of a production application are better when they don’t need to be invented each time. Authentication is a good example.
That nonprofit volunteer portal from earlier might require volunteers, coordinators, and administrators to have different levels of access. A client portal might contain information that should only be available to a particular customer. Those aren’t details you want treated casually.
The same applies to databases, hosting, backups, payments, and other infrastructure. That’s where the choice of vibe coding platform starts to matter much more for agencies.
Duda Vibe, for example, combines conversational app generation with infrastructure and agency management capabilities from the broader Duda platform. Projects can use an automatically connected secure database and built-in user authentication, including user roles. Duda also provides hosting and the underlying platform infrastructure rather than requiring agencies to assemble those pieces separately.
Just as important, it treats the application as client work.
Agencies can invite clients to their Duda Vibe projects through Duda’s white-label platform. Duda supports client and team accounts, permissions, branded client access, and centralized management, allowing an agency to manage websites and Vibe projects as part of the same broader operation.
At scale, these considerations become especially important. Building one AI-generated application is an experiment. Building 50 applications for paying clients is a service line. The platform underneath them it all has to reflect that.
Agencies don’t need to abandon third-party software. In many cases, existing products will be the right recommendation. But the threshold for making that recommendation is changing.
When a client wants its website and marketing stack to work more closely around its business model, accepting an 80 percent fit isn’t longer the obvious answer; the remaining 20 percent deserves a closer look.
It might be a portal. A dashboard. An internal tool. A quoting system. A directory. A calculator. Or something particular enough that nobody has bothered to package it as SaaS. Until recently, that specificity made custom software expensive.
With vibe coding, it can make the project worth selling.
