A16z: The next target for AI-native software is the HR system that American corporations "can't replace."

Null

Author: Joe Schmidt IV

Translation: Deep潮 TechFlow

Deep潮 Guide: a16z partner Joe Schmidt IV writes directly about Workday, the HR software giant with nearly $10 billion in annual revenue: its moat looks deep, but its underlying architecture remains stuck in 2005, and AI patches can’t save it. The article dissects why Workday’s four-layer defense system (integrated binding, proprietary configuration, consulting ecosystem, multi-year contracts) is loosening simultaneously, and presents six key features and implementation paths for an AI-native HCM system. This is a typical publicly released a16z investment thesis, clearly signaling to entrepreneurs: build the next-generation Workday.

Workday is probably the most important, yet least liked, product in enterprise software. Over 10,000 companies use it, tens of millions of employees are stuck inside it daily, with annual revenue approaching $10 billion and a market cap around $30 billion.

These numbers have nothing to do with how good the product is. HR administrators’ daily routines look like this: running reports across three pages, because the system is too complex to do payroll cycles in Excel; staring at Zoom while business partners click through promotion processes step by step; waiting for IT to explain which integration broke this week.

Customer renewal rates are close to 100%, which most would interpret as a sign of a good product. But Workday’s situation is different—high renewal rates mean customers want to leave but can’t.

HCM (Human Capital Management) is the last domain in large enterprise software without an AI-native challenger. That situation is about to change. A transformation bigger than the platform migration that created Workday is underway, and Workday’s end is near.

Cloud migration created Workday

Workday itself is a product of platform transformation. In 2005, when the dominant market players PeopleSoft’s founders Dave Duffield and Aneel Bhusri were maliciously acquired by Oracle, they bet that migrating from client/server architecture to multi-tenant cloud would reset the HRIS (Human Resource Information System) category. They judged that Oracle and SAP couldn’t catch up architecturally, and within ten years, all client/server HRIS would become legacy maintenance businesses. At the same time, they saw large enterprises shifting from large upfront capital expenditures to predictable annual operating expenses, moving from license purchases and self-hosted data centers to subscriptions.

Workday won on two fronts simultaneously: it was the only mainstream HRIS designed from day one with a multi-tenant cloud architecture, and its pricing model perfectly aligned with the enterprise shift to subscription purchasing. Within a decade, it became the default choice.

Why is Workday so hard to challenge?

Over the past twenty years, every serious attempt to attack the enterprise-level Workday market has failed, for the same reason: Workday’s moat isn’t in the product itself, but in everything surrounding it.

Deep technology and human binding. Workday is at the center of hundreds of integrations—payroll, benefits, ATS (Applicant Tracking System), expense reimbursement, identity verification, finance, state taxes—none of which can be cleanly migrated. Each instance contains thousands of hours of muscle memory: an admin running the same performance cycle for four years straight; a payroll manager memorizing a seventeen-step payroll release process. Adding a new cost center can ripple through reports, integrations, job architecture, pay grades, and benchmarking systems, and only if all systems and people are updated in sequence can it work.

Proprietary configuration layer. Workday’s configuration approach differs from most enterprise systems. Integrations are built with Workday Studio—a proprietary tool with a 6 to 18-month deployment cycle, requiring certified consultants. Reports run on Workday’s own BIRT implementation, calculated fields use Workday’s expression syntax, tenant configurations are managed through platform-specific business processes and security frameworks. These skills have no market outside the Workday ecosystem: no open-source community, no Stack Overflow, developers and their employers are locked into their own resumes.

Consulting cartel. Over 10,500 certified consultants worldwide are spread across firms like Accenture, Deloitte, Kainos, PwC, KPMG, and over 150 small service providers implementing Workday. Projects last 6 to 18 months, costing $300k to over $1 million, with implementation fees often equal to 100% of annual software costs. This service economy might be more valuable than the product itself—lobbying for Workday, absorbing customer complaints, and creating vested interests in maintaining Workday globally.

Multi-year contracts lock in customers. Workday locks clients into multi-year contracts. Even if they want to switch tomorrow, they are structurally forced to wait until the contract expires.

These four layers stack up, making Workday’s total revenue retention one of the highest in enterprise software, and for most large enterprises, it’s one of the hardest products to dislodge.

Why now is the window

Some say: “It’s been twenty years, challengers come and go, Workday is unshakeable.” That’s true. New entrants either target startups (Rippling and Gusto’s path) or niche markets that are hard to crack (Deel’s cross-border employment).

Workday isn’t sitting still either. In 2025, it launched over 25 AI features under the Illuminate brand, rolled out a dozen or so intelligent agents, acquired Sana Labs and Pipedream, introduced a usage-based Flex Credits pricing model, and signed clients like Accenture, Nike, and Merck. AI ARR surpassed $400 million, growing triple-digit year over year.

But Flex Credits and “AI ARR” are more about procurement innovation than product innovation. Signing Flex Credits and running core HR processes with agents in production are two different things. The reason Flex Credits exist is pragmatic: every CIO and CFO’s 2026 KPIs include “AI investment,” which requires actual spending to prove; every traditional software vendor’s earnings call asks about “AI revenue” first. Flex Credits is a mutual procurement structure—clients commit to a credit pool, which is accounted for as AI expenditure in budgets; Workday records this as AI ARR; which agent runs which process can wait until Illuminate produces something useful. Both sides meet their KPIs, no one needs to commit to specific deployments, and they celebrate the deal.

Don’t just take our word for it. A major Workday service partner recently wrote, “Most organizations don’t even know these capabilities exist, let alone how to activate them.” Clients are already resisting paying incremental token fees on renewed subscriptions. Long-term Workday admins describe Illuminate as: the same manual work, now wrapped in a chat interface. Each Illuminate feature is an overlay on the same form approval engine—AI ARR can grow threefold, but the underlying product remains unchanged.

But this time is different. Three variables are changing simultaneously, exposing Workday’s Achilles’ heel.

First, enterprise IT is finally re-evaluating core systems. Large companies are assessing AI readiness for ITSM, ERP, and HCM systems they thought could lock in for a decade. The pace of change in AI tech stacks has turned old architectures into liabilities. A company aiming to lead in AI can’t rely on a 2005-designed HRIS. When CHROs and CIOs ask “What does an AI-native version of this look like?” Workday’s answer—selling consumption credits on the same engine—is the breakthrough.

Second, the tools needed for rebuilding are in place. The same moves are happening at a higher layer of enterprise tech stacks: companies like Tessera are already doing AI-native SAP migrations at Fortune 500 scale—ERP complexity is an order of magnitude higher than HCM, with a single ECC to S/4HANA upgrade costing $700 million and taking three years. HCM faces similar issues but on a smaller scale. Plus, vendor-owned pre-deployment teams (not consulting firms like Accenture) mean implementation is no longer a fortress.

Third, Workday can’t seal its gaps from within. The company is betting on three directions: Illuminate as the desired intelligent agent product, Sana as the new “work entry point,” and the upcoming Agent System of Record as the governance layer for enterprise agents. All three are built on the same form approval engine—powerful but twenty years old, difficult to configure and modify, unable to keep up with modern HR needs. AI can’t change the underlying infrastructure here.

Workday’s real underlying asset—the trillion-dollar transaction data set—sounds substantial, but in operation, what matters is how data connects to workflows, permissions, and integrations. Each layer of this tech stack is now a liability. Workday can overlay AI, but turning it into an AI-native system without a complete overhaul is impossible—something a public company driven by installed base can’t do.

As we’ve discussed, the renewal cycle for Fortune 500 systems is opening for the first time in twenty years, driven by the wave of enterprise AI platform re-architecting. No next-gen HR solution is designed at the level of Fortune 500 HRIS systems. This is a unique opportunity—targeting the enterprise market where Workday makes the most money, and where all previous challengers have failed.

What an AI-native Workday should look like

We see an opportunity to directly target Workday’s HCM business, building an enterprise-grade AI-native HR system for the next twenty years.

We want a product with six features:

Deployable within a month. Implementation is Workday’s biggest weakness and the core reason companies don’t switch. A true enterprise-grade Workday implementation must cover payroll taxes across all 50 US states, cross-border payroll in over 60 countries, ACA and SOX compliance controls, benefits vendor integrations, union agreements, Workday Studio, BIRT reports, Extend configurations. No one person knows everything; most projects take 12-18 months due to knowledge fragmentation among a dozen experts, requiring sequential scheduling. Coding agents flatten this fragmentation. An agent can ingest the entire tenant (business process definitions, integrations, audit logs, payroll runs), rebuild rules in natural language, verify against real-time integrations, retain edge cases, and generate configuration drafts in days. Today’s Workday can only configure within its constraints; an AI-native HRIS should customize policies based on actual company rules, with coding agents doing what consulting firms charge six figures for.

Built-in HR workspace. The best HR admins are essentially product managers—they know what cross-system reports CHROs need, what HR planning tools should look like, what onboarding should feel like. Today, these tasks can’t be done by one role. Cross-system data pulls require data warehouses and data teams; building workflows or apps needs developers or a Workday Extend contract. The workspace compresses all this into an agent-native interface: ask a cross-system question and get an answer; describe an app in natural language and generate a usable version; propose a process change and see impact previews. HR teams we’ve talked to are already trying to build these—like an onboarding process that drafts JD, assembles 30-60-90 day plans, coordinates IT accounts and devices.

Agent-first approach. Besides portals, employees should interact with HR through daily tools. An employee on a business trip to Milwaukee should be able to ask Slack who’s nearby within 50 miles and get an answer in the same chat. Managers approving leave should see the full context (balance, recent leave, upcoming requests, team coverage) directly in the approval interface, without jumping to another dashboard. More complex: creating a new division. Today, it takes weeks in Workday—new cost centers, job structures, headcount plans, benefits setup, payroll integration, approvals. In an AI-native system, an HR operations lead should describe in natural language (e.g., “500 people in Austin and Dublin, reporting to EVP X, with these roles and salary bands”), and the system automatically maps dependencies, highlights downstream changes, and generates configuration and rollout plans. Data should flow both ways: HR data should drive intelligent agent workflows across the organization, not be locked inside HRIS.

Open ecosystem. Connecting a new payroll provider takes 6-12 months of custom integration via Workday Studio; adding a benefits vendor is similar; pulling data into BI tools requires consulting. We’ve seen frontline teams bypass this—building custom integrations with Claude MCP to pull data into their tools, creating apps that route approvals via Slack, treating Workday as a read-only system. These teams want a truly open HRIS: their own intelligent agents should directly access HR data models, APIs shouldn’t be blocked by credit pools, and the integration layer should be a first-class product. The ecosystem’s strength lies in building the best agent builders on top of data—this is where work gets done fastest.

Security and permissions at the agent layer. HR data is the most sensitive (pay, performance, sick leave, PII), and agents operating on this data need system-native fine-grained access controls. Managers’ agents should see team pay, individual contributors’ shouldn’t. External recruiters’ agents should see open roles but not offboarding history. Proper permission controls for each agent are the critical threshold—without native architecture, it’s nearly impossible to safely deploy AI on live HR data.

Always-on compliance. HR data regulations (EU AI Act, GDPR, data residency) are expanding faster than any single admin can keep up. Inside Workday, compliance is maintained by a senior HR person reading newsletters and hoping nothing is missed. AI-native stacks flip this: an always-on agent monitors regulatory changes across jurisdictions, flags what needs updating, drafts configuration updates. This is hard to retrofit on a 2005 architecture, but natural in a 2026 architecture.

How to get started

Building all six features takes time. The starting path looks like this:

Partner with a few Fortune 500 design collaborators doing AI readiness assessments on their HR tech stacks. First, use mapping and migration tools to identify unique, regional rules and edge cases in their tenants, then automate manual tasks around Workday (payroll spreadsheets, performance handoffs, ticket queues), and only switch fully when the system overhaul is scheduled.

Business architecture is critical here. Workday’s multi-year contracts lock in HRIS budgets, but Fortune 500 HR organizations have adjacent budgets that aren’t locked—HR operations, HR tech, transformation, innovation, consulting. A well-scoped implementation or automation subscription can be cleanly funded from these budgets, with real SOWs and procurement processes, avoiding direct confrontation with Workday renewal negotiations. When renewal windows open, the company is already inside the tenant, delivering measurable value to the CHRO. At that point, the question shifts from “trying an unknown vendor” to “expanding our trusted vendor into the budgeted spend.”

Another factor: Workday will leverage its product suite to suppress startup challengers. Most Fortune 500 Workday tenants run the full platform (HR, finance, payroll, adaptive planning). CIOs won’t dismantle their entire stack for a challenger that only does HR. The best strategy is to follow the Playbook used against PeopleSoft: attack the highest leverage points, integrate with existing systems in adjacent areas, and build native solutions in strategic directions. New entrants can treat existing finance and payroll instances as stable integrations from day one, replacing the most disliked Workday modules (performance, compensation planning, org restructuring, natural language reports) and gradually extending the platform’s coverage. Sales should emphasize continuity: payroll keeps running, integrations stay live, renewal cycles don’t cause operational gaps.

Every step dissolves a layer of defense: native agent workflows replace muscle memory, natural language configuration retires XpressO, pre-deployment teams bypass consulting cartels, adjacent budgets wedge open multi-year locks.

Don’t expect Workday to go down without a fight

Workday is already mobilizing. Over the past fourteen months, it laid off over 2,100 employees, and co-founder Aneel Bhusri returned as CEO with a clear mission of AI transformation. A company valued at $30 billion, with over 10,000 clients and a service ecosystem possibly larger than the product itself, won’t sit idly.

Expected tactics include: aggressively bundling Adaptive Planning, Payroll, and Finance cloud offerings to make renewal look like a package CFO won’t dismantle; offering multi-year renewal discounts during challenger evaluation periods; spreading FUD through consulting partners managing nine-figure Workday businesses. Creating contractual friction during tenant exports, and accelerating M&A when challengers gain traction. These actions don’t fix the underlying architecture issues, but any one can delay challenger projects by a quarter—challengers ignoring these costs risk running out of runway. The point isn’t that Workday won’t fight back, but that its current architecture can’t be rebuilt into a truly enterprise-grade system that Fortune 500 companies need.

Opportunities

HR software remains one of the few remaining enterprise software tracks: incumbents have flaws, architectures need rewriting, buyers want alternatives. The global HCM software market exceeds $40 billion and is growing; two years ago, Workday’s market cap peaked near $80 billion. We believe AI transformation will give rise to an even larger company.

And the stakes are bigger than the business itself. As companies move toward human-AI hybrid work models, running on a single system, HRIS will become the underlying infrastructure of company operations—who reports to whom, who has what permissions, who earns what, who is responsible for what, and compliance boundaries. Building these on a 2005 architecture caps how much AI can be deployed across the company.

Right now, somewhere, an HR admin is typing 17 salary adjustments into Workday’s performance cycle, field by field, while a business partner watches on Zoom to make sure she picks the right job code. This is happening in every Fortune 500 company today, using products that cost millions annually. Someone will build the next-gen Workday—a system designed for intelligent agents rather than form approvals. Once built, there will be no turning back.