Every major wave of enterprise software has been an architecture shift, not a user-experience upgrade.
Sol believes AI represents another one of those platform shifts.
SaaS didn’t beat client-server because browsers were more convenient. It won because it changed how enterprise software was built, deployed, and maintained.
Agentic software doesn’t start with chat. It starts much deeper than that. Today, much of what is sold as “agentic” is a conversational interface layered on top of an existing system. The interface matters, and every enterprise application should have one. But to us, that’s just the first layer.
We believe an agentic system is much broader: software that understands intent, senses business change and maintains itself to match, operates within the governance you’ve defined, and safely participates in an ecosystem of AI agents.
That’s the architecture we’re building at Sol.
Layer one: the interface you talk to
This is where most of what’s sold as “agentic” stops: a chat interface bolted onto a system that hasn’t changed underneath. Ask in plain language, get an answer. Maybe it completes a simple task instead of making you hunt for the right form.
That layer is today’s baseline, and every vendor should have one, including us. Employees and managers should be able to ask for what they need and have it happen, with no training required.
But answering a question isn’t the same as understanding a situation and working on behalf of a user. The difference is context, and context isn’t something you bolt on with a chat window. It changes what a system of record has to be. Traditional systems record outcomes: who was hired, what they’re paid, and how it was approved. They were never built to hold the reasoning behind those records, so no interface sitting on top can retrieve it.
Sol is built to hold that reasoning. Our agents capture the rationale behind decisions, carry memory across sessions, learn from each interaction, and pull in context from beyond the HR system, to better understand the broader picture of how work is getting done. A basic chat interface helps you finish a task. An agent with that depth of context can guide your next decision, at the moment you’re making it.
Imagine a new manager who is navigating a transfer for one of their people. Instead of a form full of fields they’re not sure how to answer, Sol walks them through it: picking the right transfer date to stay in sync with payroll, handling the compensation changes, clarifying whether a backfill is required, bringing the right business partners into the process at the right time, and sending everything for approval on their behalf.
Layer two: the agent that maintains the system
There’s a second interface most employees miss: the one your HR team uses to change how the system works. This is the piece that’s genuinely different in this generation.
At Workday, I spent nearly a decade building enterprise platforms, including leading the platform product team and launching Workday Extend. My job sat at the crux of the SaaS value proposition: how do you scale a single product across thousands of enterprises while giving each enough flexibility to run their business their way?
The answer was configuration. Vendors could run a single version of the software centrally, upgrades stopped being a project you scheduled a year out, and customers no longer needed to manage their own infrastructure. It was a massive leap forward. But configuration came with a hidden cost.
SaaS made it easy to upgrade the vendor’s product. It did nothing to make it easy to upgrade your implementation of that product. It got only half the equation right. Every time your business changes (and it changes constantly), your configuration drifts further from what you actually need. The only way to close the gap is to call someone, pay them, and wait.
Think about the decisions you made when you first set up your system: how roles and jobs were structured, how approvals were routed, which fields were required. Some were made years ago, maybe before you got there, and whether they were right at the time doesn’t matter now, because the business has moved and you’re still living with them. Changing one process often means a project nobody has the time, budget, or appetite to take on.
This next platform shift isn’t only about making software easier to use. It’s about making software capable of maintaining itself.
In the SaaS era, implementation and configuration lived outside the product. Every meaningful change required someone to translate a business decision into system changes: update workflows, rewrite rules, adjust security, reorganize data structures, test everything, and deploy it safely. The software never understood why the business had changed. It only reflected whatever people painstakingly configured it to do.
An agentic system changes that relationship. It translates business intent directly into system behavior. It understands what needs to change, determines the implications, proposes the right changes, and knows when to bring people into the loop for review, approval, or decisions that require human judgment. The mechanical work of keeping the system aligned no longer falls on administrators. It becomes part of the product itself.
The defining feature of the SaaS era was a single version of the product. The defining feature of the agentic era is a product that keeps itself aligned with your business.
Layer three: the workflow engine at the core
Today’s HR systems expect you to predict the future.
Every new process has to be mapped in advance. Every exception needs its own branch. Every approval path has to be anticipated, configured, tested, and maintained. When your business changes, as it inevitably will, you go back and redesign the workflow.
That’s one of the hidden costs of today’s HR systems. They don’t just store your processes, they force you to encode every possible path before anyone can use them.
We’ve designed Sol differently. Instead of asking administrators to define every step in a workflow, we ask them to define the rules, constraints, approvals, and outcomes that matter. Who can approve what. Which policies are non-negotiable. Where human review is required. Agents reliably stay within those boundaries, automatically.
This does two things. First, it makes workflows dramatically easier to change: as your business evolves, you update the policy, not every branch of every workflow. Second, it lets the system take on more responsibility over time. AI models are more capable: they can reason about the best path to an outcome while staying within the governance you’ve already defined.
Just as importantly, we’ve designed for humans to stay in the loop as a design principle. The system should understand which decisions are routine, which require approval, and which depend on human judgment. Good automation isn’t about removing people from the process. It’s about involving them where they add the most value.
One of our design partners described a situation that stuck with me. An employee was traveling internationally when their child became seriously ill. They reached out to HR in a panic. There were policies that applied and workflows to be followed, but that wasn’t what they needed most in that moment. They needed a person. The best system isn’t one that automates every interaction. It’s one that recognizes when empathy, judgment, or discretion matter more than efficiency and knows that humans must be in the loop at these critical times.
The real benefit isn’t less workflow maintenance today: it’s a system that can evolve with AI. As the technology improves, your organization benefits automatically, while the rules, governance, and human judgment you’ve defined stay intact.
Layer four: a foundation open to the agent ecosystem
The last layer is also the one we feel most strongly about: openness.
The first generation of enterprise software assumed the application was the destination. If you wanted to work with your HR data, you logged into the HR system. Integrations existed, but they were expensive, brittle, and usually built one at a time. The safest approach was to keep the walls high and the APIs narrow.
That model doesn’t work in an AI world.
Companies are increasingly relying on AI to help employees, managers, and leaders get work done. Those experiences won’t all live inside a single application, nor should they. The HR system shouldn’t be competing with every new AI tool. It should be the trusted foundation they build on.
That means designing for interoperability from the beginning. Sol was designed from day one to work with AI agents. This means AI tools can securely connect to Sol and get work done. Sol is built CLI and API-first with support for emerging standards that let agents securely interact with enterprise systems. A security model that extends naturally to AI, where every request is authenticated, every permission is respected, every action is auditable, and organizations have visibility into exactly what data was accessed, by whom, and for what purpose.
The goal isn’t to keep AI away from your workforce data. It’s to let you embrace it with confidence. The HR system shouldn’t be a wall around your data. It should be the secure foundation that allows every trusted agent to use it responsibly.
The question worth asking
If there’s one thing I’d leave you with, it’s this question:
When your business changes, what has to happen before your HR system changes?
Think carefully about your answer.
If every business change first has to become an implementation project, you’re still relying on software that depends on people to keep it aligned with your organization.
If instead the system understands the change you’re trying to make, operates within the governance you’ve defined, and knows when to bring people into the loop, you’re looking at something fundamentally different.
The next generation of enterprise software won’t be defined by who has the most AI features. It’ll be defined by systems that can continuously adapt alongside the organizations they support and enable companies to run at the pace of human ambition.
That’s the future we’re building with Sol.