Report: Top AI vendors for enterprise agent building for non-technical users

Executive summary

Enterprises looking to enable non-technical teams to build AI agents without heavy engineering work commonly choose one of three approaches: embed agent creation into a broad productivity platform, use a low-code application platform with AI features, or adopt an integration-first automation platform with visual agent builders. The research shows Microsoft (Copilot Studio / Power Virtual Agents), OutSystems (AI Agent Builder), and Tray.ai (Merlin Agent Builder) are frequently recommended choices — each with clear strengths and specific trade-offs.

The promise vs. reality

Supporters of Microsoft argue that Copilot Studio delivers a true no-code experience where business users can "describe their needs in natural language" and get production-ready agents, coupled with enterprise governance, compliance, and deep connectors across Microsoft services (learn.microsoft.com).

"Users can create agents by simply describing their desired functionality in natural language... Copilot Studio will generate the corresponding dialog automatically." (learn.microsoft.com)

Advocates for OutSystems point to decades of low-code experience, enterprise certifications, and built-in guardrails that let business teams assemble AI-enabled workflows and agents using visual components and templates (outsystems.com).

"OutSystems provides a visual development environment that simplifies the creation of applications, enabling both technical and non-technical users to participate in the development process." (OutSystems case materials)

Tray.ai's Merlin Agent Builder promises an integration-first, template-driven approach with 700+ connectors and built-in RAG pipelines so non-technical users can assemble agents that act across systems without writing glue code (tray.ai).

"Merlin Agent Builder offers a no-code workspace where users can build agents visually, either from scratch or by starting with pre-configured accelerators." (Tray.ai press release)

Strengths and where the vendors really shine

Microsoft (best fit when your enterprise is Microsoft-centric)
- Strengths: natural-language agent authoring, deep integration with Microsoft 365/Dynamics/Power Platform, enterprise-grade governance (Copilot Control System), extensive connectors, and strong compliance posture (encryption in transit/at rest). Several public case studies show measurable ROI and scale. (learn.microsoft.com)
- Real-world wins: Dow, Pets at Home, ABN AMRO and others report large-scale agent deployments and operational gains. (blogs.microsoft.com)
OutSystems (best fit for organizations standardizing on low-code delivery)
- Strengths: mature low-code UX for business users, strong governance and compliance certifications (SOC/ISO), prebuilt templates and AI-assisted tools (OutSystems Mentor), and proven success in modernizing legacy systems. (outsystems.com)
- Real-world wins: banking, insurance, and large enterprises leveraging OutSystems to speed delivery and maintain controls (case studies: Ahli United Bank, CITIC Telecom CPC). (casestudies.com)
Tray.ai (best fit where deep cross-system automation and integration are primary needs)
- Strengths: extremely connector-rich platform (700+), focus on RAG pipelines and IDP for unstructured data, visual agent composer, on-prem/hybrid connectors, and enterprise governance add-ons (Agent Gateway / Merlin Guardian). Early customers report fast migration of legacy integrations and strong efficiency gains. (tray.ai)
- Real-world wins: Cvent migration of 500+ integrations, AVID Property Group invoice automation, and ITSM agent deployments with marked ticket reductions. (businesswire.com)

Common limitations and where promise breaks down

Microsoft
- Critics report lingering technical complexity for advanced integrations, potential vendor lock-in, confusing licensing, and real-world incidents of hallucinations or security issues like prompt injection exploits if governance is weak. "Copilot can hallucinate" or require IT help for complex data hookups are recurring themes. (techcommunity.microsoft.com)
OutSystems
- Critics note vendor lock-in, potential performance/scalability bottlenecks for very large or heavily customized apps, a non-trivial learning curve for complex scenarios, and AI model flexibility that currently favors Azure OpenAI / Bedrock over arbitrary model selection. (mindfulchase.com)
Tray.ai
- As a newer, fast-evolving entrant, Tray.ai raises questions about maturity at extreme scale, potential governance gaps if organizations don't adopt Agent Gateway controls, and some reported platform limits (e.g., concurrency constraints) that could affect very high-volume deployments. Vendor lock-in and export/migration mechanics are also areas to examine closely. (tray.ai docs)

Practical selection criteria for non-technical users (short checklist)

Ecosystem alignment: Do you already run Microsoft 365/Teams/Dynamics? If yes, Copilot Studio reduces friction. (learn.microsoft.com)
Governance & compliance: Verify SOC/ISO attestation, audit logs, DLP and identity controls (Azure/Entra, Dataverse, Purview). (learn.microsoft.com)
Integration needs: If you need broad cross‑system actions (CRM, ERP, ITSM), favor an iPaaS-style builder (Tray.ai) or ensure OutSystems connectors cover your systems. (tray.ai connecters)
No-code maturity: Test the visual builder by having a non-technical user build a simple agent in a 1–2 day workshop — if they need continual developer help, the solution isn't truly no-code. (outsystems AI docs)
Security testing: Run threat scenarios (prompt injection, data exfiltration) in a staging environment, and validate monitoring, tokenization, and access controls. (tray.ai docs)
Exit strategy: Confirm data export, model portability, and migration paths to avoid costly lock-in. (windowsforum.com discussion)

Quick pilot plan (2-week)

Week 0: Select a single, high‑impact use case (ITSM ticket triage, HR onboarding, or a knowledge bot). Define KPIs (time-to-resolution, ticket deflection, CSAT).
Week 1: Provision environments, connect one or two critical data sources, and have a business user build the agent using only platform UI and templates.
Week 2: Test with 50–200 real queries, run security checks, measure KPIs, and evaluate how much developer involvement was required.

Bottom line

If your organization is already invested in Microsoft tools and governance is a top priority, Microsoft Copilot Studio / Power Virtual Agents is the pragmatic first choice for enabling non-technical users to build agents at scale — provided you plan for governance and cost controls.
If you want an enterprise-proven low-code application platform where business teams can own app and agent delivery with strong compliance, OutSystems is a solid choice — watch out for licensing, scalability design, and model flexibility.
If your primary need is cross-system automation, complex integration, and quick migration of legacy connectors into agentic workflows, Tray.ai’s Merlin Agent Builder is compelling — but treat it as a newer, more specialized option and validate concurrency, governance, and export capabilities.