Workday in 2026 - a16z vs Bersin
Two sharp pieces on Workday's future hit my feed this week. Joe Schmidt at a16z published Workday's Last Workday, searching for a viable enterprise replacement. Josh Bersin published The Reinvention of Workday, optimistic about Aneel Bhusri's return as CEO and the strategy shift it signals.
I have lived with Workday for a long time. Rather uniquely, I actually learned how HR departments operate largely as a Workday consultant for dozens of customers, then moved into HR itself and starting running the operational teams in addition to the tech. I also ran a Workday consulting practice, and got deep into the services “cartel” as Schmidt describes it (making me a Workday Sicario maybe?). Now at Wave1, we help other high-growth companies avoid the mistakes I made in this time. About half our clients are running Workday, the others opting for challengers such as HiBob, Deel and Rippling.
Both pieces ask the wrong question. It isn't whether Workday survives. It's what role Workday plays in a stack that no longer revolves entirely around it.
My working view
Workday becomes the reliable backend for perhaps half the high-growth medium enterprises currently running it (Large enterprise I can't speak to with the same confidence). Point solutions and custom-built tooling augment it. Slack and the other tools people already work in become the front end of employee experience. Workday is a foundation to build on top of, not the tool customers spend their day inside.
That framing dictates almost everything else I think about the product, the pricing, the services ecosystem, and the next round of decisions HR leaders are making.
Where the critics are right, and where they overreach
Schmidt is correct that the moat isn't the product. It's the deep technical and human wiring. Anyone in our space knows that moving off Workday touches every part of the business, like replacing the engine of a car while it's moving at speed.
"Cartel" is perhaps an unfair description of the Workday Ecosystem and Schmidt misses something important. Most consultants of my generation are as much HR transformation consultants as Workday consultants. Swap the platform for Oracle or SuccessFactors and half the job, often more, is the same - process design. I'm a good example: I can't code and I'm not the sharpest technical mind, but I was a decent Workday consultant because I liked streamlining HR processes and gave people confidence as they navigated what was often scary change. Replacing Workday with something else doesn't automatically give clients the capability to design their future-state processes end-to-end. They often need help.
The challenger momentum is real. HiBob, Deel and Rippling built products that are intuitive, well-suited to medium enterprise, meaningfully cheaper and meaningfully more flexible. Larger customers are increasingly willing to accept that perhaps 30% of their requirements need temporary workarounds (human or non-human) in exchange for what is usually a nicer UX, lower TCO, and faster AI adoption within the product. Deel replacing Workday at Klarna is a good example.
The Schmidt point I disagree with most is "deploy in one month." Even with deployment agents, the floor for a global large-enterprise HCM deployment is closer to three months. That's already a huge improvement on where most enterprises sit today, but the one-month framing skips the most important part of deployment. A new system is the catalyst that forces leaders across departments to sit down and co-design the future state. An agent can't replace the conversation between Finance and HR about how to run headcount management, or between HR and IT about how identity and access actually work. I don't see HR agents and Finance agents having that conversation, nor teams ceding decision authority to AI representatives. The point of agents is to remove the busy work, not skip the strategic, sometimes emotive discussions that shape what work looks like in the company.
Schmidt also overplays the workbench-for-HR-teams point. The best HR admins are operations people, like operations people in any business function. The product instincts usually sit one or two teams over, in People Tech and Analytics. And even there, if the team wants to build something new, they'll do it in Claude Code, Cursor or Lovable and connect it back. We don't need a vibe-code platform for every department. IT definitely doesn't!
Where Bersin is right, and where the rebrand falls flat
Three points in the optimistic case deserve more weight than the headlines give them.
The configuration framework is genuinely valuable, not just legacy baggage. The workflows, security and data structures built into Workday over twenty years are exactly the kind of boundaries agents need to act safely on sensitive HR data. We are one data breach away from big pause moments on AI transformation in many companies. Every week a new vendor launches their own HRIS and discovers it's harder than they imagined (See Lattice sunsetting theirs).
The leadership reset is significant. Aneel back as CEO. Adam Godson (Paradox) and Joel Hellermark (Sana) as GMs with P&L accountability rather than vague advisory roles. Bersin reports the company narrowed fifty agent projects down to fifteen in a single afternoon. These are the kind of leaders willing to cut through the bloat accumulated over the years and get back on track.
Sana is the most interesting Workday acquisition in years, and the one most likely to actually change how the product feels daily. The cautionary note here is Peakon, which was an awesome product before the acquisition, when innovation seemed to pause as everything got redirected toward suite integration. It's still handy, but seems to have never reached the potential it once showed. For customers who aren't migrating off Workday anytime soon, Sana has more potential than anything else in the portfolio to make day-to-day processes feel intuitive in a way they never have. At Bolt we tried every trick in the book to move activity from Workday to Slack and it's still not easy. Sana might finally fix the fact that Workday's Slack app isn't as good as a basic slackbot made with Make or Zapier.
The other Sana upside nobody is discussing is developer experience. Extend never really seemed to reach the levels it could have. Compare it to Salesforce, where whole companies were built on top of the platform. (My first HRIS deployment was Fairsail/Sage People, built on Salesforce, for the HR tech nerds!) If a Sana-based agent developer is genuinely flexible and intuitive, that's finally the platform play Workday has been gesturing at for a decade.
That said, the rebrand from "system of record" to "platform for agents" doesn't really land for me. Maybe I'll be proven wrong but I've never heard a single person say they need a system of record for their agents. Yes, we need security and controls, but why do we need to track agents the same way we track workers? It looks suspiciously like a solution searching for a problem. Customers want easy integration and AI inside the tools they already use. Those are the problems Workday should focus on.
Two more Bersin claims don't add up. The first is that Workday's configuration "encodes every customer's unique policies, approvals, compliance rules, and org structure." Across the customers I work with, data quality is often poor, processes aren't optimised, and meaningful shadow work happens outside the system to fill the gaps. Workday isn't Notion. Workday is black and white: you see process and data, but rarely context. There's no real space to document the why. So when you talk to people, they often don't know it. Nothing was written down. When those people leave, the institutional knowledge leaves with them. The brain of the company sits in the heads of the people running the systems, not in the configuration. You can only fix this with strong documentation embedded in your ways of working. This isn't easy when trying to operate in an agile way.
The second is the framing that customers building on top of Workday are deluded and constructing a "shadow ERP that costs millions." The realistic number for most teams attempting it is in the hundreds of thousands, not millions, because what you actually need is a small team of engineers and PMs who build incrementally and across HQ functions. The honest question is whether the all-in cost is really so different from standing Workday up in the first place, or keeping it optimised without a large internal team. The gap is smaller than positioned and this becomes more interesting as barriers to building software are removed.
What both pieces miss
The bigger shift for me is software-as-a-service to services-as-software (Call it SaaS to SaS!).
YC is now interested in funding "full-stack" AI companies. Instead of selling legal AI agents to law firms, they want you to build an AI law firm. The same logic creates a gap for AI HR outsourcing. Deel is the closest example. It looks and is valued like a SaaS company, but Deel is actually a services company: employer of record, global payroll, HR operations at scale. The reason this is a serious question for HCM incumbents now and wasn't before is margins. BPO has much lower margins than software, so Silicon Valley ignored it. Replace the people with agents and the unit economics change. If People Ops is mostly agentic and we're moving to consumption-based commercials anyway, why not also do the work for the customer?
The question is who is best positioned to build it. A technology company like Workday? A services company like Accenture? Or neither, because the offering would need to start from scratch with AI rather than build on top of existing products and services?
Deployment is genuinely about to change
Worth flagging because both pieces touch it but neither captures the exciting parts for me.
I recently kicked off my first Ashby ATS deployment. From week one, Ashby connects to your prior ATS API and you build processes on top of that live dataset, rather than cleaning data and migrating at cutover as we've done in ATS migrations for years. That pattern is coming to HRIS deployments next. Smaller players are increasingly willing to migrate the data for you, so versions of this might reach large enterprise sooner than expected. The progression makes sense now that forward-deployed engineers are mainstream. "We'll deploy or partially deploy for you" becomes more normal.
In the same direction: the most forward-thinking organisations I work with are already building their own micro-apps and shifting self-service to Slack and other platforms where the work happens. Working towards Workday as a reliable system of record sitting behind a self-built front end, or several custom front-ends. For this to work, Workday needs better connectivity. If Workday fixes it themselves, there isn't much opportunity for new players. If they don't, plenty.
The early version is already playing out in people analytics. High-growth companies are building their own data warehouses, combining datasets, adding security layers, then vibe-coding tailored dashboards. The C-suite wants to plug Claude into the people data lake and vibe-code their own dashboards. This is a massive unlock but takes time to do properly.
What worries me most
The commercial model.
API requests, integration events and document storage are now metered. The categories that draw down credits once you exceed your allowance are exactly the ones that make Workday useful in an AI-era stack: orchestrations, customer-built Extend apps, and any external application or agent calling into your tenant. n8n, Zapier, Make, Workato augment Workday to fill specific gaps and connect it to other systems. Every push or pull draws down credits. The allowance is a soft "fair use" level rather than a hard limit, but it comes from the same bucket as the tokens powering Workday's native AI agents.
The practical effect: the more you integrate Workday with the rest of your stack, the less budget you have for the new agents that are supposed to make your life easier and increase the ROI of the system.
That isn't a neutral pricing change or procurement innovation. It's a moat. It pushes customers deeper into Workday's own ecosystem at exactly the moment the market is trying to do the opposite and take a more modular approach to their HR Tech stack. It forces a binary choice, full-stack Workday or no Workday, rather than letting customers flex to their own organisational needs. The Deels of the world are happy to fill the gaps wherever you need them and build from there.
That's the thing to watch. Not whether Workday survives. Whether it lets its customers build the stack they actually want around it. Is Workday optimising for retention across a broad base, or for wallet share from a smaller, more locked-in one? 🤔