AppGen And Low-Code Platforms

Forrester's own framing of what happened to this market is that AppGen is eating low-code. The category name now carries both terms, with the newer one first, which is a fairly precise description of a takeover in progress.

What the rename concedes is worth sitting with. Low-code spent a decade selling abstraction through visual tooling: drag components onto a canvas, configure rather than code, and let people who are not developers build things. AppGen sells abstraction through language: describe what you want and receive an application.

Both are answers to the same question, which is how an organisation gets more software built than its engineering team can produce. The interesting part is what the shift reveals about which half of the original low-code proposition was actually load-bearing.

What the category covers

Forrester defines AppGen and low-code platforms as software development tools that use AI and machine learning to automate creating, editing, and releasing applications, letting developers and non-developers work through prompts, visual and declarative techniques, and other AI assistive tooling to generate software. Many also generate backend workflows and AI agents.

The use cases the research identifies as core are application and workflow generation, core business applications, customer-facing applications, digital process automation, and low-code tooling for both professional and citizen developers. Extended use cases include AI agent development, application monitoring and management, task automation, integration-only deployments, and front-end-only web applications.

Read that list and the scope problem becomes obvious. This one category covers a business analyst assembling an approvals workflow, a professional development team building a customer portal, and an enterprise generating agents that act inside its processes. Those are three different buyers with three different risk tolerances, and forty vendors serving some subset of them.

The generation is not the product

Here is the thing that decides purchases in this category, and it is not the quality of the generated output.

Any competent platform can now produce a working application from a description. The models are good enough that generation quality is converging, and it will keep converging. What does not converge is everything that has to be true for the generated thing to survive contact with an enterprise.

Forrester makes this argument directly and specifically. It contrasts enterprise AppGen platforms with the well-known standalone generation tools, naming Lovable, Replit, Claude Code, Cursor, and Windsurf, and characterising what they produce as a mixed bag. Not because the generation is bad. Because they are not backed by an enterprise platform, which leaves gaps in autoscaling, built-in security, and the management of the applications, workflows, and agents that get created.

The objective of an AppGen platform, in Forrester's framing, is to provide enterprise-grade scale, security, and management abstracted away from the creator, so the person building can just build.

That is the whole category in one sentence. The creator experience is the marketing. The platform underneath is the product.

It also explains the continuity with low-code. The reason enterprise low-code vendors survived a decade of cheaper alternatives was never the visual editor, which was always replicable. It was governance, lifecycle management, identity integration, deployment control, and the ability to answer questions about what is running and who built it. AppGen inherits exactly that moat.

The boundary with agentic development platforms

Worth being precise, because Forrester runs a separate category next door and the two get conflated constantly.

Agentic development platforms are aimed at professional engineering teams building and maintaining software of every kind, with agents doing the work across the development lifecycle. Forrester explicitly excludes application generation platforms from that definition.

AppGen platforms are aimed at putting creation in the hands of both professional and citizen developers, with the platform absorbing the enterprise concerns.

The distinction is roughly who carries the operational burden. In an ADP, the engineering team still owns the software and the agents accelerate them. In AppGen, the platform owns the operational envelope and the creator owns only the thing they described.

If your problem is that your engineering team cannot ship fast enough, you are probably looking at the wrong category here. If your problem is that four hundred people in the business have requests your engineering team will never get to, this is the right one.

Inside The AppGen And Low-Code Platforms Landscape, Q2 2026

Published in Q2 2026 and authored by Ken Parmelee with Chris Gardner, Kylie Cadogan, and Kara Hartig, the report maps forty vendors across the market and is aimed at CIOs and application owners.

It is a Landscape, so nobody is scored and nobody is ranked. Vendors included are described as notable vendors, and several are already citing that phrasing in marketing. Inclusion means Forrester considers a vendor part of the market, which is meaningful and is not a rating.

Publicly confirmed inclusions so far include Newgen Software, Netcall with Liberty Create, and Kissflow. Expect the confirmed list to expand as the remaining thirty seven work through their announcement cycles.

The framing Forrester published alongside the report is more useful than the vendor list, and unusually pointed for a market map.

The market is evolving faster than most organisations can absorb it. AppGen and low-code platforms are expanding development capacity to everyone in the organisation while simultaneously introducing fragmentation, complexity, and uneven platform maturity. As vendors converge applications, workflows, and agents into single environments, the gap between tools that build quickly and tools that scale safely is widening.

And the line that should worry anyone about to buy: most companies are not set up to manage distributed, high-speed development across teams with very different skill levels.

Speed becomes the constraint

Three structural trends run through Forrester's read on this market, and they compound.

Software creation is moving deeper into the business. Development stopped being centralised in engineering years ago, and business users now build applications, workflows, and agents directly, embedding creation closer to where demand originates. That expands capacity and dissolves ownership boundaries, pushing accountability outward faster than governance adapts to catch it.

Distributed development replaces bottlenecks with fragmentation. Teams move faster without engineering queues, but autonomy produces divergence in tooling, standards, and architecture. What removes delay also removes consistency, and the coordination overhead grows with adoption rather than shrinking.

Which leaves speed itself as the constraint on delivery. The bottleneck is no longer how long it takes to build something. It is whether the organisation can absorb, integrate, secure, and maintain everything being built.

This is a genuinely different problem from the one low-code was sold to solve. Low-code addressed a capacity shortage. AppGen creates a coordination problem, and coordination problems do not respond to more capacity.

The fourth wave of this

Forrester is candid that the industry has seen this pattern before, and names the precedents: no-code was going to eliminate developers, RPA was going to automate everything, citizen developers were going to build all the enterprise apps. Each brought real value and real confusion.

The specific lesson carried forward is about what unmanaged democratisation produces. Shadow IT, unmanaged APIs, and RPA sprawl all generated the same outcomes: security exposure, zombie software running with unknown resource consumption, and artefacts that could not be extended or improved and therefore had to be rebuilt.

That is a fair description of what happens if you give four hundred people a generation tool with no platform underneath. The applications work. Nobody knows how many exist, which ones touch customer data, who owns the one that broke, or whether the person who built the thing running a critical process still works there.

Forrester's prescription is not a return to central IT approval for everything, which would destroy the value of the model. It is that the organisation should provide the platform people build on, so the guardrails are structural rather than procedural.

That is the correct answer and it has an uncomfortable implication for sequencing. The platform decision needs to precede the enthusiasm, and in most organisations the enthusiasm arrived first.

What Forrester says these platforms cannot do

The research is specific about capability limits, which is unusual and useful.

What these platforms can do today: generate applications that can be extended and enhanced afterwards, generate backend workflows from patterns, automate parts of the development lifecycle including testing, documentation, and deployment, and create agents that handle straightforward tasks or workflows.

What they cannot do: understand your business context, autonomously ensure compliance, security, or governance, or replace the judgement and nuance held by people who understand how the business is supposed to run.

The middle item on that list is the one to take into vendor conversations. A platform can enforce compliance controls that somebody configured. It cannot determine what your compliance obligations are or notice that a generated application has quietly created a new one. That gap does not close with a better model, because it is not a model problem.

Forrester's summary position is that AppGen does not end software development but does drastically reduce the need for coding, and that the organisations who benefit will be the ones with a clear view of what it can and cannot do rather than the ones who buy the narrative.

What changed since the last scored evaluation

For context on how much has shifted, the last Forrester Wave in the adjacent low-code space was The Forrester Wave: Low-Code Development Platforms For Professional Developers, Q2 2021, a twenty seven criterion evaluation of fourteen providers: AgilePoint, Appian, GeneXus, HCL Software, Mendix, Microsoft, Oracle, OutSystems, Pegasystems, Salesforce, ServiceNow, Thinkwise Software, Unqork, and WaveMaker.

Fourteen vendors scored against twenty seven criteria in 2021. Forty vendors mapped and none scored in 2026. That is not a market that consolidated. It is a market that was reopened by a technology shift, and the absence of a current Wave means there is no analyst ranking to lean on while it sorts itself out.

What to test

Bring a real request, not a demo scenario. Ask the vendor to build something that has to authenticate against your identity provider, read from a system of record you actually run, and respect a data residency rule. Generation quality is not the variable. Integration and constraint handling are.

Ask what happens on day four hundred. Who owns an application whose creator left? How does the platform surface an app that has not been opened in six months? Can you find every application that touches a given data source? These are the questions that decide whether you have a portfolio or a sprawl.

Test the guardrails by trying to violate them. A creator should be structurally unable to deploy something that breaches policy, rather than instructed not to. If the control is a training module, it is not a control.

Separate the citizen and professional developer stories deliberately. Several of these platforms are genuinely strong for one population and thin for the other, and a demo will not show you which, because the demo is run by someone expert in both.

And decide who owns the platform before you decide which platform. This category redistributes software creation across the organisation, and Forrester's own read is that most companies are not currently set up to manage that. The tooling choice is the easier half of the problem.

Forrester does not endorse any vendor named here, and inclusion in a Landscape report is not a rating or a recommendation to buy.