Union

Name: Union
Brand: Union

AI Native

The enterprise Flyte platform for AI development and orchestration.

Est 2021|Valuation $40M|Raised $53.1M|15 Ref Customers

LinkedIn 9K|Web traffic 9K|FTEs 43|HQ Bellevue, WA

MarketsWorkflow Orchestration

Published Oct 15, 2026

01Common complaints

Consider alternatives when: 1. The organization is cost-sensitive and at an early stage: alternatives such as Dagster or Prefect provide lower-cost entry points and free tiers that may be appropriate during validation. (https://dagster.io/pricing) 2. The organization is already standardized on Airflow with a large existing DAG portfolio and seeks incremental operational improvements: an Airflow-focused managed offering may reduce migration cost. (https://www.astronomer.io/pricing/) 3. The organization requires abundant independent reviews and third-party validation: Union’s public review footprint is currently limited based on the sources reviewed. (https://www.g2.com/products/union-ai-union/reviews) 4. The workflows are not ML/AI-specific (pure ELT/BI scheduling) and do not require dynamic execution, caching, or heavy fan-out—Union’s capabilities may exceed requirements.

Deal disqualifiers (poor fit)

Organizations that primarily need a simple cron-like scheduler for a small number of SQL/ELT jobs
Teams that prefer GUI-only orchestration and do not want to adopt workflow-as-code
Organizations with budgets that cannot support an initial platform subscription fee prior to compute usage

Known issues evidenced in public reviews

Performance / slow loading under heavy usage
Perceived limitations in customization options

Limitations and gaps in publicly available information

Sparse independent reviews, making it difficult to assess real-world support quality and reliability outside vendor claims
Benchmark methodology not published for performance claims presented by the vendor
Compliance breadth beyond SOC 2 is not clearly documented in public materials

Roadmap / near-term indications

Community and vendor announcements reference evolving offerings such as serverless and expanded GPU support. Buyers should confirm GA status, SLAs, and support scope when evaluating features referenced in preview or community posts.

02Why they buy

Differentiators supported by evidence

Built on Flyte (open-source, Kubernetes-native) with a commercial layer; this provides an open-source escape path to Flyte core.
SOC 2 Type II certification is claimed for Union Cloud, which is relevant to enterprise procurement.
Serverless preview and trial credits provided a route for initial experimentation during preview periods.

Where competitors may have clear advantages

Breadth of community and integrations: Airflow, Dagster, and Prefect have larger public communities and integration ecosystems.
Pricing accessibility: some competitors provide lower-cost entry tiers or free tiers.
Independent validation: competing products tend to have more third-party reviews and published benchmarks.

03Field reports

Review platforms (G2, Capterra, etc.)

G2: review volume for the product is limited, which constrains generalization. The available feedback highlighted platform reliability and collaboration features, while also noting performance issues (slow loading) and limitations around customization.

Capterra: no clearly relevant listing for the orchestration product was identified; some unrelated product listings with similar names were identified and are not relevant to this offering.

Community and open-source presence

Union contributes to Flyte documentation and community resources; Flyte is an open-source project and Union is positioned as a primary maintainer.
Flyte community forums and discussion spaces include both vendor announcements and independent community conversation.

Reported reasons for selection (in vendor and third-party sources)

Likely selection drivers cited in vendor messaging and third-party reporting include:

Requirement for ML-specific orchestration beyond static DAG scheduling
Kubernetes-native execution and infrastructure-aware scaling
Preference for a managed offering rather than a fully self-managed Flyte deployment

Reported reasons for churn or non-adoption

Public data on churn is limited. Available review commentary and pricing materials suggest potential concerns including:

Performance and UI responsiveness under high concurrency
Customization limitations in the UI or product configuration
Team-tier entry price relative to competitors with lower starter tiers

Vendor- and press-reported outcomes (not independently verified)

Press coverage includes an anecdote quoting an ML infrastructure engineer at a large payments company attributing increased offline training job throughput and more frequent model releases to Flyte usage. This anecdote refers to Flyte rather than the paid commercial offering.

Implementation details and time-to-value

Public, detailed case studies with step-by-step implementation timelines were not found in the reviewed sources. Vendor documentation suggests typical onboarding via CLI configuration and OAuth flows but does not provide a publicly available end-to-end implementation playbook.

Critical evaluation of metrics

Performance and efficiency claims on vendor comparison pages (for example, claims of order-of-magnitude speedups) generally lack published methodology or benchmark configurations in the public materials. Buyers should request supporting details such as workload definitions, baseline configurations, cost model assumptions, and reproducibility steps when evaluating vendor performance claims.

04Best fit

Company sizes

Best fits tend to be mid-market to enterprise organizations with dedicated data and ML platform functions. The offering is less optimized for very small, hobbyist use unless the serverless pay-as-you-go option meets those needs.

Industries / verticals

Plausible verticals based on tool adoption patterns and vendor messaging include:

Financial services and payments
Consumer internet
Logistics and mobility
Retail analytics
Healthcare and biotech (where reproducibility and auditability are important)

Roles and personas

Decision makers: Head of ML Platform, Director of Data/AI Engineering, VP Engineering
Users: ML engineers, data engineers, applied scientists, platform SRE/DevOps

Qualification criteria (good fit)

Organizations that are a good fit typically:

Run multiple multi-step training and inference pipelines and require retries, caching, observability, versioning, and governance
Use or plan to use Kubernetes-native execution and want to standardize workflow patterns
Require a managed platform or a path to BYOC/hybrid deployment models

05The pitch

What the product does (objective core value proposition)

Union is a workflow orchestration platform for data, ML, and agentic AI systems. It is built on the open-source Flyte orchestration project and is offered as a commercial distribution that provides features for authoring, executing, monitoring, and scaling complex workflows across Kubernetes and cloud environments. In operational terms: Union is intended to enable organizations to run and govern multi-step pipelines (ETL, training, evaluation, batch inference, and LLM/agent workflows) with functionality for scheduling, retries, caching, observability, and multi-environment promotion.

Target market and market positioning

Union targets: - ML platform / ML infrastructure teams that support multiple internal model developers - Data engineering and ML engineering organizations building production pipelines at scale - Enterprises that require governance, RBAC/SSO, auditability, and predictable costs The vendor positions the product as an AI-focused orchestration solution (distinct from general-purpose schedulers) and emphasizes scaling, reproducibility, and infrastructure-aware execution.

Get a comprehensive analysis of Union including market position, competitive landscape, adoption trends, and peer benchmarks.

Download full report →

Published Oct 15, 2026

01Common complaints

Consider alternatives when:

The organization is cost-sensitive and at an early stage: alternatives such as Dagster or Prefect provide lower-cost entry points and free tiers that may be appropriate during validation. (https://dagster.io/pricing)
The organization is already standardized on Airflow with a large existing DAG portfolio and seeks incremental operational improvements: an Airflow-focused managed offering may reduce migration cost. (https://www.astronomer.io/pricing/)
The organization requires abundant independent reviews and third-party validation: Union’s public review footprint is currently limited based on the sources reviewed. (https://www.g2.com/products/union-ai-union/reviews)
The workflows are not ML/AI-specific (pure ELT/BI scheduling) and do not require dynamic execution, caching, or heavy fan-out—Union’s capabilities may exceed requirements.

Deal disqualifiers (poor fit)

Organizations that primarily need a simple cron-like scheduler for a small number of SQL/ELT jobs
Teams that prefer GUI-only orchestration and do not want to adopt workflow-as-code
Organizations with budgets that cannot support an initial platform subscription fee prior to compute usage

Known issues evidenced in public reviews

Performance / slow loading under heavy usage
Perceived limitations in customization options

Limitations and gaps in publicly available information

Sparse independent reviews, making it difficult to assess real-world support quality and reliability outside vendor claims
Benchmark methodology not published for performance claims presented by the vendor
Compliance breadth beyond SOC 2 is not clearly documented in public materials

Roadmap / near-term indications

02Why they buy

Differentiators supported by evidence

Built on Flyte (open-source, Kubernetes-native) with a commercial layer; this provides an open-source escape path to Flyte core.
SOC 2 Type II certification is claimed for Union Cloud, which is relevant to enterprise procurement.
Serverless preview and trial credits provided a route for initial experimentation during preview periods.

Where competitors may have clear advantages

Breadth of community and integrations: Airflow, Dagster, and Prefect have larger public communities and integration ecosystems.
Pricing accessibility: some competitors provide lower-cost entry tiers or free tiers.
Independent validation: competing products tend to have more third-party reviews and published benchmarks.

03Field reports

Review platforms (G2, Capterra, etc.)

Capterra: no clearly relevant listing for the orchestration product was identified; some unrelated product listings with similar names were identified and are not relevant to this offering.

Community and open-source presence

Union contributes to Flyte documentation and community resources; Flyte is an open-source project and Union is positioned as a primary maintainer.
Flyte community forums and discussion spaces include both vendor announcements and independent community conversation.

Reported reasons for selection (in vendor and third-party sources)

Likely selection drivers cited in vendor messaging and third-party reporting include:

Requirement for ML-specific orchestration beyond static DAG scheduling
Kubernetes-native execution and infrastructure-aware scaling
Preference for a managed offering rather than a fully self-managed Flyte deployment

Reported reasons for churn or non-adoption

Public data on churn is limited. Available review commentary and pricing materials suggest potential concerns including:

Performance and UI responsiveness under high concurrency
Customization limitations in the UI or product configuration
Team-tier entry price relative to competitors with lower starter tiers

Vendor- and press-reported outcomes (not independently verified)

Implementation details and time-to-value

Critical evaluation of metrics

04Best fit

Company sizes

Industries / verticals

Plausible verticals based on tool adoption patterns and vendor messaging include:

Financial services and payments
Consumer internet
Logistics and mobility
Retail analytics
Healthcare and biotech (where reproducibility and auditability are important)

Roles and personas

Decision makers: Head of ML Platform, Director of Data/AI Engineering, VP Engineering
Users: ML engineers, data engineers, applied scientists, platform SRE/DevOps

Qualification criteria (good fit)

Organizations that are a good fit typically:

Run multiple multi-step training and inference pipelines and require retries, caching, observability, versioning, and governance
Use or plan to use Kubernetes-native execution and want to standardize workflow patterns
Require a managed platform or a path to BYOC/hybrid deployment models

05The pitch

What the product does (objective core value proposition)

In operational terms: Union is intended to enable organizations to run and govern multi-step pipelines (ETL, training, evaluation, batch inference, and LLM/agent workflows) with functionality for scheduling, retries, caching, observability, and multi-environment promotion.

Target market and market positioning

Union targets:

ML platform / ML infrastructure teams that support multiple internal model developers
Data engineering and ML engineering organizations building production pipelines at scale
Enterprises that require governance, RBAC/SSO, auditability, and predictable costs

The vendor positions the product as an AI-focused orchestration solution (distinct from general-purpose schedulers) and emphasizes scaling, reproducibility, and infrastructure-aware execution.

Get a comprehensive analysis of Union including market position, competitive landscape, adoption trends, and peer benchmarks.

Download full report →