iPaaS / API workflow automation. You wire apps together through their APIs and move finance data between them, with AI on a step or two. These lend themselves to cross system, API friendly work: FP&A board pack assembly, AR/AP handoffs, bank feed and cash pipelines, recurring reporting. The picks differ mainly on who runs it and at what scale.
Top picks
Workato
Enterprise grade integration backbone with mature agent orchestration and deep finance connectors.
Enterprise · Business led
Boomi
Enterprise integration for complex hybrid estates, with Agent Control Tower governing agents across vendors. IT led, enterprise priced.
Enterprise · IT led
n8n
Low cost and self hostable, with native AI agent nodes and a strong EU data sovereignty story.
Low cost · Self hosted
Strong alternatives
MuleSoft
Enterprise integration built around reusable, governed APIs, Salesforce owned. Needs real integration engineering.
API led · Enterprise
Power Automate
Cost effective if you already pay for Microsoft 365, though premium connectors and scale need separate Power Automate licensing.
Microsoft estates
Make
Visual, low cost and genuinely maintainable by a finance team itself; EU based, GDPR and SOC 2.
Visual · Finance owned
Robotic Process Automation and Orchestration. A software robot works your screens and applications like a person would, handling multi step processes with approvals, queues and a full audit trail. These suit the messier end of automation initiatives, where legacy systems with no APIs still need to be used. These picks differ on depth, governance and cost.
Top picks
UiPath
The market leader: the deepest, most mature platform for automating finance work at scale, now with AI agents and document reading built in.
Enterprise · At scale
Automation Anywhere
Cloud first: describe a task in plain language and its AI copilot builds the automation for you.
Cloud · Business led
Appian
Best for running a whole finance process end to end, not just bots: quick to build, strong on approvals and audit.
Whole process · Low code
Strong alternatives
Pega
The most powerful option for complex, regulated finance work, with deep decision rules. Powerful, but needs real expertise to set up.
Complex · Regulated
Power Automate Desktop
The cheapest, simplest desktop automation if you already run Microsoft. Good for small tasks, less so for complex work at scale.
Microsoft · Desktop
SS&C Blue Prism
Built for regulated finance and banking, with strong audit, control and security. Pick it when compliance matters most.
Governed · Secure
No code agent builders. You describe a task in plain language, or record it, and the platform builds the agent for you, no coding needed. Some are general builders you can point at any finance job; others specialise in one, like documents or the close. The card tells you which.
Top picks
Copilot Studio
Microsoft's no code tool for building governed assistants on your own data. A sensible default for Microsoft organisations that want agents to answer finance questions and pull together commentary.
Microsoft · Governed
GoApprentice
It learns a finance process by watching you do it once, then rebuilds it as an agent you can review and replay. Good for the close and for tasks that only live in someone's head.
Broad task range
Beam AI
A no code platform for building finance agents from your own procedures. Good for high volume work such as reconciliations, invoice capture and receipt checks.
High volume · Scale
Strong alternatives
Convey
Describe a task or share your screen and its AI teammate learns the process and runs it. Each teammate has its own identity and scoped access permissions, limiting what it is allowed to touch. Good for repeatable close and operations work.
Screen taught · Controlled access
Agentforce
Salesforce's governed agent platform: build and run assistants on CRM and connected data, with guardrails and monitoring in the platform. Strongest where the work already lives in Salesforce, and finance coverage is thinner than sales and service.
If you run Salesforce
ServiceNow AI Agent Studio
Build governed agents in AI Agent Studio, monitored through the AI Control Tower. Strongest where ServiceNow already runs the workflow. Coverage outside IT processes is thinner, so finance agents take more configuration.
If you run ServiceNow
Data warehouse and data lake agents. You build agents that work directly on the data warehouse or data lake where your numbers already live. It can be a strong route for board reporting, driver analysis and cash forecasting, as long as your data is already in one place. Which one fits comes down to the platform you run.
Top picks
Snowflake Cortex
Build agents and ask questions of Snowflake data in plain English, inheriting your Snowflake roles and permissions. Data stays inside Snowflake’s governed environment, though inference may run in another approved Snowflake region.
If you run Snowflake
Agent Bricks
Build coordinated agents on your Databricks data and tools. What each agent can reach follows the permissions you set on the underlying data, with audit logging across the platform.
If you run Databricks
Microsoft Fabric
Build agents and ask questions of your data across Fabric and Power BI in plain English. The agents sit on your OneLake data lake, so answers come straight from your own numbers.
If you run Fabric
Strong alternatives
BigQuery + Gemini
Google's data warehouse with Gemini AI built in, so finance and data teams can analyse warehouse data just by asking in plain English.
If you run BigQuery
dbt
Defines trusted metrics once so every agent and report uses the same numbers, now part of the merged Fivetran and dbt Labs. Best as the foundation beneath the agents rather than an agent itself.
Trusted metrics
ThoughtSpot Spotter
An AI analyst that answers questions, builds dashboards and forecasts over your own data. Good for giving the whole team self serve analysis without writing queries.
AI analyst · Self serve
The first four tabs assemble on platforms the vendor maintains. This tab is the build route: the organisation writes and maintains the agent software, even where a cloud provider runs the infrastructure.
Engineering grade platforms and frameworks. Your developers build a bespoke agent in code, for the finance workflows where no ready made product fits. Each option here gives you both sides: a framework to build the agent, and a managed runtime to host and govern it. So you can self host for full control, or use the managed service to move faster. Which one you pick is mostly down to your cloud and model preference.
Top picks
Anthropic
Build agents with the Claude Agent SDK, with fine grained control over what each one can do. They can then be run managed on Claude Managed Agents, still in research preview. Cloud neutral, so it fits whatever you run.
Agent SDK · Any cloud
OpenAI
Build with the OpenAI Agents SDK, with tools, memory and native sandbox execution for controlled agent work. A practical route where the team already uses OpenAI models and infrastructure.
OpenAI native
Microsoft Azure
Build with the Microsoft Agent Framework, then deploy and govern agents through Microsoft Foundry Agent Service, with identity through Entra.
If you run Azure
Strong alternatives
Google Cloud
Build with the open Agent Development Kit, then run it managed on the Gemini Enterprise Agent Platform, formerly Vertex AI, working across Gemini and many other models.
If you run GCP
AWS
Build with the open Strands SDK, then run it in production on Bedrock AgentCore, with governance and monitoring built in.
If you run AWS
LangChain
Build with LangGraph, an open framework for agents that need careful step by step control, then run them on LangSmith Deployment or self host. Good when you want to own the whole stack.
Open source · Self host