Workday Is Onboarding AI Agents the Way It Onboards People

Sequoia's Sonya Huang and Pat Grady introduce Carl Eschenbach as Workday's CEO, a former Sequoia partner, and "perhaps the most beloved executive in technology." They frame the company as the system of record for people and for the numbers those people produce. As human workers are complemented by AI agents, they argue, that same ledger is on a path to cover the agents and the work they do as well.

Grady teases him as a venture capitalist currently moonlighting as a CEO. Eschenbach laughs and then flips the joke: he still has responsibilities at Sequoia, and some days he wonders which job he is doing more. If he is honest, he says, it is the CEO seat. Remaining inside the partnership still matters to him. He calls Sequoia the most iconic venture firm of all time and says being around partners like Huang and Grady is a blessing.

The hosts preview three ways Workday intends to get paid for AI — seat-based pricing, role-based agents, and consumption-based API access — and a harder management problem: how to implement the technology inside a large company without throwing away culture and values. That tension runs through the hour. Workday has spent 20 years as the system of record for people and money. The question now is whether the next workforce, human plus digital, can live on the same platform.

Grady puts the skeptical question on the table. Workday is the ultimate system of record for people in a world where AI is replacing people — so what happens to that system? Eschenbach's first correction is that Workday records two assets, not one: people and money. The HR platform and the financials platform look like two products from the outside; architecturally, he says, they are the same. He hears the disruption question constantly. His biased answer is that the company is in a unique position, not a trapped one.

After 20 years, Workday sits on a highly curated database of more than 70 million users, and last year it processed 30 percent of job requisitions in the United States. AI, he tells them, is all about data, but not only data: it is the context of that data, and being inside the actual business-process workflow. That combination is what lets role-based agents produce results customers can see. With 11,000 customers and two decades as a core system of record, the company has also become a system of truth. People trust it.

In every tectonic shift, Eschenbach argues, enterprise customers look back to the vendor that already holds their critical systems and ask for help with the transition — the same way they did 20 years ago moving HR and finance from on-premises to the cloud. Domain-specific agents with highly curated data, he says, are much more beneficial in the enterprise than general-purpose LLM models. Asked whether incumbents are in trouble because of startups, he refuses the or: it is an and. Plenty of Sequoia-backed companies, he notes, add value on top of the system of record. There is room for both.

Grady asks the product question underneath the moat. If an AI developer shows up wanting Workday's treasure trove of records and context, how much of that data is actually usable, and how much is locked behind permissions, architecture, or 20-year-old company scar tissue? A lot of companies already connect today, Eschenbach says — identity systems, Active Directory — especially when a new employee is onboarded. They do it through a general-purpose open API. That path is about to close.

As agents flood the enterprise, someone has to onboard them in a secure, responsible, ethical, and compliant way, with identity, access controls, and governance over the data they can see. That, he says, is the risk. When Workday talks to CIOs, the first sentence is usually that they love agents and love the technology, and the second is that they are watching agent sprawl. His analogy is shadow IT. Departments once bought cloud services around central IT; security teams spent years clawing back control. The same script is playing now, except the new shadow workforce actually does the work.

Workday is building an AI gateway so agents cannot hit those generic APIs anymore. They will have to come through the gateway to get in. About a month and a half before the taping, the company announced an Agent System of Record: a unified ledger for human workers and digital workers. Humans already come through Workday, get assigned to an organization, receive benefits, get tracked, and get measured. Agents arriving tomorrow raise the same questions — who onboards them, which org they belong to, what they can access, who controls them, who measures them — and today the honest answer is almost no one.

Once humans and agents share a platform, workforce planning can cover the whole mix, because the workforce is no longer only human. Huang restates it: the system of record is still people and money, except "people" now includes agents. Eschenbach pushes back on peers who forecast a 20, 30, or 40 percent cut in human headcount. He does not believe that is the path. Grady then maps Aneel Bhusri's original vision forward. ERP took stock of machines when manufacturing was the economy; Workday took stock of people in a knowledge economy; an AI economy needs a system that takes stock of agents, sitting on the same financial backbone, because people and agents are fluid.

How do you take a mixed workforce to market? Eschenbach backs up a week. He had just reminded Sequoia steward Roelof Botha that the old product-led-growth orthodoxy was oversold. For a stretch the industry claimed you would not need 37-year sales veterans, a top-down enterprise motion, or people like him. With AI, he says, the times are coming back. It is a sale to heads of departments, heads of functions, or the CIO. PLG is useful, "but it's not the panacea that everyone thinks." Distribution is becoming a human motion again.

Two years earlier, when copilots arrived, many peers told customers they now had a copilot and would raise prices 20 percent. Eschenbach refused. Copilots, he argued, would become part of the foundation of every product: built in, not bolted on, and covered by the annual subscription. The company got beat up for leaving money on the table. His answer was that customers should pay when Workday can show true value with agents and a clear return, not because a chatbot showed up in the UI.

Today's AI list price has three lanes. First, a per-seat uplift when an agent or feature is valuable to the whole employee population. Second, a price on the agent itself — he sketches a payroll agent at $50,000 a year against a $200,000 administrator doing pure administration, a comparison a CFO can finish quickly. Third, consumption: how often outside systems hit Workday for data. He points to Bret Taylor's Sierra, which charges on consumption and outcomes. Grady asks whether fewer payroll people means fewer Workday seats. Eschenbach is not worried. AI should let that administrator do higher-value work rather than disappear, flipping the equation so technology works for people instead of the other way around.

The narrative still needs work. Walk into a CEO, CFO, or CIO conversation and it immediately becomes ROI: this will save a boatload of money. Sometimes it will. If that is the only story, employees hear that their jobs are being replaced. "We need to talk about AI from a growth value proposition, not just an ROI value proposition." Savings should be reinvested to drive growth. That, he says, is how employees, agents, and AI peacefully coexist inside a company — and how the technology actually produces company-level gains instead of a morale crisis.

Grady calls Eschenbach one of the most productive people he has met and asks what he actually uses. Inside Workday the stack includes Slack AI, Zoom AI, and Gemini. Every 90 days, analyst reports for earnings go into Gemini and come back as summaries, saving hours of reading; missed Zoom calls get the same treatment. The company is also launching Everyday AI, which he describes as a first-of-its-kind program at this scale: all 20,000 employees must go through AI training, with a full curriculum and a test. The logic is personal productivity first, company gain second, and less fear of the tools. He was scheduled to launch it about two weeks after the taping.

On products, the standout is a recruiter agent from HiredScore, acquired about a year earlier. Talent acquisition is one of the most expensive functions in HR. Give the agent to a recruiting team, he says, and they can see a 50 percent jump in recruiter productivity: hundreds of resumes collapse to a top two or three in seconds. Time to hire drops 20, 30, even 40 percent. After people are in the building, Talent Optimization and a talent mobility agent match an employee's skills, background, and experience to a manager's project. Customers have seen attrition fall by as much as 40 percent, because people can see new internal opportunities instead of leaving. He also claims customers can need 50 percent fewer additional recruiters — a number he calls pretty powerful.

Grady asks whether AI can finally solve matching. Historically the problem had to be collapsed into title, geography, and pay band, none of which predict whether someone will do a good job. Eschenbach says it is all about skills, not pedigree. Accenture, one of Workday's largest partners, now hires 30 to 35 percent of new employees without a college degree, matching skills to job outcomes with AI. Because Workday is the system of record, a company of 10,000 already has a skills database; the first look for a new role can be inside. Huang then asks the human question. Her Sequoia interview with Carl was memorable because of chemistry — the feeling of whether you want to work with someone every day. Can an agent do that?

His answer is firm. There will be a human in the loop, especially in hiring. He would not accept someone into the company based 100 percent on skills matching. The productivity dividend, he argues, should fund a skills transformation: networking, giving feedback, collaborating, leading with empathy — crafts the workforce lost during Covid and cannot rebuild on Zoom alone. Grady adds a more abstract version. Google commoditized knowledge; OpenAI is commoditizing intelligence; what remains scarce is agency, the decision about what question to ask and what to achieve. Agents can reason through steps. Wanting something is still human. Eschenbach's closer: technology only enables change. People still have to implement it, embrace it, teach it, and teach new skills.

Cloud computing, he reminds them, was supposed to shrink CIO organizations; those teams are larger than ever. Developers were supposed to vanish too. He expects fewer junior developers, but architects and senior engineers will still do the final work and decide what ships. Huang jokes that nobody wants Carl vibe-coding the production codebase. Then Grady turns to the timing. Eschenbach became CEO in January 2023, two months after ChatGPT, walking into a 17- or 18-year-old company with many billions in revenue and tens of thousands of people. From the outside it looks like two transformations at once: cleaning up how the business runs, and the AI shift.

Eschenbach refuses personal credit. The team drove the change with him; the hard part was winning their support. What will never change are the six values Aneel Bhusri and Dave Duffield wrote on day one, before they built the product — Duffield having earlier founded PeopleSoft. People first, customers, integrity, innovation, fun, and profitability sixth, because if the first five are right, the sixth follows. He still talks about those values in front of employees and customers. Culture, though, does change as a company scales: more operational rigor, more efficiency, a bigger-company way of running. Values stay still. Culture has to evolve.

The moving parts are go-to-market, technology velocity, and the org chart. He has re-segmented the market, gone deeper into verticals, and rebuilt the partner motion. He does not want an ecosystem around Workday; he wants an economy, where partners deploy, sell, get rewarded, and innovate on the platform — build once, sell to 11,000 customers. On technology, Bhusri had already been putting machine learning in the platform for a decade; this wave is an acceleration. Eschenbach wants the company to move with urgency at scale and disrupt itself before anyone else tries. The dataset, he thinks, is the reason they have to run.

The hardest thing in his career, he says, was a restructuring earlier this year: about 8 percent of the workforce, roughly 1,650 people. Doing that at a values-based company, on his watch, was brutal. Getting leadership, Bhusri, and the board aligned was a heavy lift. They tried to do it with empathy, care, and generous packages so people leaving felt as well treated as that kind of moment allows. The operating expense it freed now funds the go-to-market and product bets. Two and a half years in, he feels the pressure and the pride of stewarding a company that already manages more employees than anyone else.

Grady asks whether the moat changes in the AI era, or whether the next 20 years are simply the AI version of being a core system of record. Eschenbach says the moat remains the data. "Data is the new UI/UX for AI." Garbage in, garbage out; highly curated data with context produces great agent outcomes. That moat lets Workday build the next one: role-based agents. Many things called agents today are super-copilots or task agents — RPA on steroids, automating a repetitive step. Role-based agents, because Workday knows human skills and the context of the data, would pick up skills rather than just knock out a task. That, plus trust, is how he wants customers to navigate the unknown.

Lightning round. The Sequoia partner he misses most is Emily, which Grady calls the only answer that offends no one. Eschenbach talks about arriving at the firm as an operator with no ego, writing investment memos that came back covered in red ink, and cold-calling as if he were 30 years younger. Grady says what people miss is the beginner's mindset. Everyone assumed a great operator would automatically succeed in venture. Instead he learned the job from scratch, asking after partner meetings what a DA or an MAU was. That apprenticeship, not the operating résumé, is what impressed the partnership.

Asked for his Mount Rushmore of CEOs, he names EMC's Joe Tucci, a leader people would run through walls for, and Microsoft's Satya Nadella, who came up through technology and then adapted. Longevity and pliability impress him more than a fixed style: the CEOs who personally transform as the era changes. At home he uses Gemini and ChatGPT for questions; his kids send him AI jokes he does not always follow. For founders growing into the CEO job, one piece of advice, simple enough to put on a note: focus on the success of others, not on your own. He thanks them for having him back inside the Sequoia building, and that is where the hour ends.