Workday's agent stack — what Sana, Flowise, and Pipedream actually do

A quick line on each acquisition. Workday announced the Flowise deal in mid-2025 and pulled it into the core developer story. The Sana deal, the largest of the three at about 1.1 billion dollars, was announced at Workday Rising in September 2025 and closed in early 2026. Pipedream followed shortly after as the third piece. Each one was sold publicly as a standalone capability. The interesting part is what they look like as a stack, because that is what your tenant will actually receive.

Sana — the front door

Sana is an AI-native enterprise platform for search, learning, and agents. It indexes content across the tools your organisation uses, answers questions in natural language, generates content, and can act through connected applications. Two product lines that matter for HR: Sana Learn (an AI-native learning experience) and Sana Agents (the orchestrator on top of company knowledge).

Workday's stated ambition for Sana is to become "the new front door to work". In practice, that means a conversational entry point sitting on top of Workday's HCM and Finance systems, plus connected sources like Microsoft 365, Google Workspace, and ServiceNow. The employee asks a question in plain language. The assistant pulls trusted data from Workday and any other system it has been wired to. Where allowed, it takes action.

Two practical consequences for HRIS. Sana shifts Workday's interaction model from "navigate to the right screen" to "ask for the outcome". This will have a real adoption story when it ships well, and a real change-management cost when it ships badly. The other consequence: Sana's reach makes Workday more useful as the orchestrator of cross-system work, even when the underlying data sits in another tool. Workday Learning, in particular, gets a meaningful upgrade. Sana Learn enables auto-generated courses and FAQs from your policies, personalised learning paths, and AI tutoring.

Flowise — the agent builder

Flowise is a visual, low-code builder for AI agents. You drag and drop blocks (LLMs, tools, retrieval, human approval steps, integrations) onto a canvas and assemble them into working agents. It supports the chaining, routing, parallel, and human-in-the-loop patterns we covered in earlier insights. The community version has been used widely in the developer ecosystem for several years.

For Workday, Flowise solved the most visible gap in the developer stack: the ability for customers to build their own agents, not just consume native or partner ones. Flowise Agent Builder slots into Workday Build as the agent design surface. The agents you build there can call any Workday API, sit inside an Extend app, run through Workday Orchestrate, and (over time) communicate with external agents via MCP and A2A protocols.

The SKU answer matters and Workday gave it cleanly. The Flowise Agent Builder is included with Workday Extend Professional, not a separate purchase. Customers already on Extend Pro get access automatically in the first half of 2026. Early Adopter programs ran ahead of GA for a handful of customers.

Pipedream — the connectivity layer

Pipedream is an integration platform purpose-built for AI workflows. The headline numbers: over 3,000 connectors and roughly 10,000 pre-built tools, accessible through a visual builder and natural-language commands. The use case Workday cares about is letting agents reach into the rest of your tech stack: Asana, HubSpot, Jira, Slack, ServiceNow, and the long tail of SaaS your business already runs on.

Why this matters. Workday's stated framing is that the real promise of enterprise AI is "not just insights, but action". For agents to act, they have to reach the systems where the work actually happens. Pipedream is the bridge between Workday's secure data plane and the everyday tools your teams use. The agent that pulls a goal from Workday and creates a corresponding Jira ticket is doing it through Pipedream, not through bespoke integrations your team has to build and maintain.

A useful mental model from Workday's own language. Flowise builds the brains. Pipedream provides the hands. Sana is the face. The three together describe the operating model of a Workday-anchored agent that can think, act, and be talked to.

How the pieces fit together in practice

Take a concrete example. An employee opens the Sana-powered assistant inside Workday and asks "what is my parental leave entitlement, and can you schedule the absence in Workday and notify my manager in Slack". Sana parses the request and routes it. A Flowise-built knowledge agent retrieves the local policy, summarises the entitlement, and surfaces the source. A Flowise-built action agent creates the Workday absence event in the standard business process. A Pipedream connector sends the notification to Slack. The Agent System of Record logs everything, attributes it to the right Agent System User, and makes the action auditable for compliance.

None of those steps required a custom integration project. None of them required the employee to leave Workday or learn a new tool. That is the picture Workday is selling. It will not be in production tomorrow at this level of polish, but every component is real today.

What this changes for HRIS teams

The role of HRIS shifts toward orchestration. The conversation moves from "which Workday screens does my user need" to "which agents are running in my tenant, what can they do, and how do I govern them". HRIS leaders who have not already started this shift will feel it in the next two release cycles.

Extend becomes more strategic. If your team is on Extend Essentials, you are now a tier removed from the agent capability that ships with Pro. The cost of upgrading needs to be evaluated against the value of having agent building as a first-class capability. This is not a marginal feature comparison anymore. It is a strategic decision about whether your HRIS team builds or just consumes.

The integration tax goes down, but the governance tax goes up. Pipedream and Sana between them remove a lot of the plumbing work that used to land on integration teams. They add a lot of governance work that lands on HRIS, Legal, and Security. Agents that can read across systems and act across systems are agents that need careful scoping, monitoring, and audit. The Agent System of Record exists exactly to support this, but the operating model around it lives in your governance board.

Learning gets reinvented before performance does. Sana Learn is one of the most concrete near-term changes to the Workday product. Auto-generated content, personalised paths, AI tutoring. If you own Workday Learning, the next twelve months are likely to be your busiest year in a long time.

Where the work actually lands, per piece

For Flowise, the work is mostly entitlement and pilot. Confirm your Extend tier with the account team; if you are not on Professional, scope what it would take to upgrade and whether Early Adopter slots are still open. If you are on Pro, pick two pilot agents you would build first in Flowise Agent Builder (one read-only knowledge agent and one transactional agent is a healthy split) and line them up with your delivery partner so the work starts the week the Builder lands in your tenant.

For Sana, the work is content readiness, not technology. Sana is only as good as the sources you point it at, and the rollout will move faster than most content libraries are ready for. List the three or four collections you would index first (HR policies, manager SOPs, benefits content, learning catalogue), check that access controls hold up when an agent reads them on behalf of different employee populations, and retire the duplicates and outdated documents before they become hallucinated answers. Workday Learning teams in particular should treat the next two release cycles as a Learning content audit, not a feature rollout.

For Pipedream, the work is naming the systems your agents will actually need to reach. Build a short list of the SaaS tools an HR agent has to touch in the next twelve months (Jira and Slack always make it; the more interesting ones are ServiceNow, your ATS, your benefits portal, and your ticketing tool), and walk it through with your integration lead so the connector inventory is honest before the first agent shows up asking to call out.

Across all three, decide ownership before you decide pilots. HRIS naturally owns the Workday-side operating model. Integration owns the Pipedream side. Learning owns the Sana content side. Weekly contact between the three is what keeps the stack coherent in your tenant rather than three parallel rollouts.

A note on hype

Workday is not the first vendor to promise "your platform becomes the orchestrator of all enterprise work". Most predecessors did not deliver on that promise. The reason to take this one more seriously is that the pieces are individually proven (Flowise has a real developer community, Pipedream has thousands of working customers, Sana has serious enterprise references), and the integration model is being built into the platform rather than bolted on through acquisitions sitting outside the core.

That said, none of this means you need to rebuild your roadmap this quarter. The components are real. The full stack story will take all of 2026 to land cleanly. HRIS leaders who treat this as a steady evolution rather than a sudden one will land in a better place than those who chase the latest demo.