What matters here for IT leaders is not simply that the vendor field has become more crowded; it is that AI platform selection is shifting from a tooling decision to an operating-model decision. Once platforms are expected to execute work across business processes, the evaluation criteria move beyond model performance and data-science usability toward runtime control, integration depth, process observability, and the ability to contain failure when autonomous actions go wrong.
The source is right to frame this as a portfolio market, but that creates a management challenge many enterprises underestimate: a multi-platform strategy only works if architecture, governance, and skills are designed for it. Without common guardrails for identity, data access, audit trails, agent testing, policy enforcement, and cost monitoring, a portfolio becomes fragmented experimentation at scale. Interoperability is therefore not just a feature checklist item; it is the mechanism that determines whether teams can compose agents across systems without multiplying operational risk.
CIOs should also read the rise of agentic platforms as an organisational warning. If business units can buy workflow-centric or SaaS-embedded AI faster than central IT can define standards, platform sprawl will arrive before strategy does. A practical response is to separate decisions into three layers:
- enterprise control plane requirements: security, governance, observability, and vendor portability;
- use-case fit: process intensity, data gravity, and integration needs;
- workforce impact: which roles supervise, approve, and continuously tune agent behaviour.
The competitive question is no longer who has AI pilots. It is who can industrialise governed digital work without locking themselves into the wrong platform assumptions too early.
The Forrester Wave™: AI Platforms, Q3 2026 has just published, and if you’ve read previous evaluations in this category, prepare to recalibrate. Agentic AI has redrawn the boundaries of what an AI platform is, what it must do, and what vendors compete to provide it. The 15 vendors evaluated represent one of the most heterogeneous fields we’ve ever assessed in this market. And the stakes of choosing among them have never been higher. Choose your platforms like your future depends on it. It does. As you read this evaluation, understand that:
- Agentic AI has redefined the category and the evaluation. For the previous decade, AI platforms were largely synonymous with data science workbenches: environments to prepare data, train models, and generate insights. But agents don’t stop at telling you what’s happening; they understand context, navigate workflows, and complete work. We therefore assessed platforms not just on how well they help data scientists, but on how well they model and execute enterprise processes. That’s where agentic value is created.
- New entrants have widened the field, not narrowed it. A category once dominated by data science specialists now includes vendors with roots in workflow automation, robotic process automation, enterprise SaaS, data management, and cloud infrastructure. That heterogeneity is a feature, not a bug. Each vendor clusters around a distinct sweet spot, whether data science depth, workflow specialization, industry solutions, or application development. Buyers must scrutinize both the breadth and depth of each vendor’s capabilities against their highest-value and most imminent use cases.
- Your AI platform strategy will be a portfolio, not a monolith. Every platform in this evaluation is general purpose; But forcing every use case onto a single platform means most of them land outside its affinity. Enterprises get the most value by implementing each use case on the platform whose sweet spot it best maps to. Treat sweet-spot fit and interoperability as your key decision criteria.
- A vendor’s vision, innovation, and roadmap are critical. AI is advancing at a blistering pace, and agentic AI won’t merely automate existing processes. It will reshape how enterprises are structured, as digital workers take on entire functions and human roles shift toward orchestrating and governing fleets of agents.
- The margin for error is thin and compounding. Enterprises that convert AI into completed work across core processes will pull ahead on cost, speed, and customer experience simultaneously. Those that stall in pilots and proofs of concept will watch the gap widen quarter by quarter. An AI platform decision made carelessly, or deferred indefinitely, is an existential risk.
Where Are The Frontier Model Companies?
Readers will notice that the frontier AI labs are not in this evaluation. That’s deliberate, not an oversight, because:
- The vendors in this Wave evaluation are mostly model-agnostic. With the exception of a few vendors with their own model families, such as the hyperscalers, the platforms in this evaluation let enterprises bring the models of their choice. Their value lies in what surrounds the model: data, process context, agent development, governance, and deployment. That model flexibility is a feature, not a gap, because it lets enterprises swap in better models as they emerge without re-platforming.
- Frontier AI labs are becoming platforms in their own right. They are rapidly building agents, tooling, orchestration, and enterprise services layered on top of their frontier models. That market deserves its own rigorous evaluation, and are covering it: The Frontier AI Model Platforms Landscape publishes in Q4 2026, followed by a full Forrester Wave™ evaluation in Q1 2027. Together, these two evaluations will give technology leaders a complete map of the AI platform decision space.
Dig Into The Evaluation
The Wave evaluation is a starting point, not a verdict. To get the most from it:
- See how all 15 vendors measure up. We evaluated Amazon Web Services, C3 AI, Databricks, Dataiku, DataRobot, Google, IBM, Microsoft, Oracle, Palantir, Pegasystems, Salesforce, SAS, ServiceNow, and UiPath. Forrester clients can use the interactive provider comparison experience to adjust criteria weightings and build a shortlist tailored to their priorities. Some of these vendors also offer the ability to download a copy of the report.
- Reach out to us. Forrester clients can schedule a guidance session with us to discuss the Wave results, pressure-test their AI platform strategy, match their highest-value use cases to vendor sweet spots, and plan for the agentic future.
Enjoyed this article? Sign up for our newsletter to receive regular insights and stay connected.