On November 10, investors put fresh capital behind a less flashy, but increasingly essential, part of the AI boom: the infrastructure that helps agents remember, act safely, and stay within bounds.
Three funding announcements from Daylight, Graphon Labs, and Bay add up to $165 million. Daylight raised an $80 million Series B, Graphon Labs announced a $35 million Series A, and Bay emerged from stealth with $50 million in total funding. Together, the news points to a growing opportunity for entrepreneurs: when a new kind of software starts doing work at scale, a second wave of businesses often emerges to solve what that software still cannot do reliably on its own.
For AI agents, the missing pieces are becoming easier to spot. They need context that lasts beyond a single interaction. They need security that can keep pace with automated work. And they need governance that lets companies expand usage without surrendering control.
Three different answers to a shared infrastructure challenge
Each company is tackling a different pressure point in the move toward AI agents at work.
Daylight’s $80 million Series B was led by Growth Equity at Goldman Sachs Alternatives, bringing the company’s total funding to $120 million. Daylight combines AI agents with human security engineers to help organizations detect, investigate, and respond to cyber threats. Its model reflects a practical reality: automation can handle more routine work, but high-stakes decisions still call for expertise and oversight. The company’s reported valuation is around $400 million, though that figure has not been officially disclosed by Daylight.
Graphon Labs raised a $35 million Series A led by RRE Ventures, bringing its total funding to $43.3 million. Its focus is memory. The company describes its approach as “Persistent Intelligence”: infrastructure intended to help AI agents retain context and relationships across enterprise information, rather than starting from scratch whenever they encounter a new task.
Bay emerged from stealth with $50 million in total funding, made up of a $37 million Series A and a previously undisclosed $13 million seed round. The startup is building AI-agent governance: a way for organizations to monitor agent activity and evaluate proposed actions in context, allowing, blocking, or escalating them for human review.
These are not three versions of the same product. They address distinct needs that become more urgent as agents move from experiments into real workflows.
Memory is a product problem, not just a model feature
Anyone who has tried to work with an AI tool across a complicated project knows the frustration: useful context gets lost, details have to be repeated, and separate pieces of information do not always connect in the way a person expects.
That inconvenience can become a serious obstacle inside a business. An agent working across many documents or tasks needs more than a good answer in the moment. It needs to understand which information relates to which decision, what has changed, and what should carry forward.
That is the opportunity Graphon is pursuing. If agents are to support ongoing work, memory has to become dependable infrastructure, not a nice-to-have feature. This creates room for companies that help organize context, preserve relationships between information, and make the right details available when an agent needs them.
For founders, the useful question is not simply, “Can we build a smarter agent?” It is also, “What does the agent have to remember to do this job well, and how will that memory stay useful as the work grows?”
Security needs to keep pace with the work
Agents can move quickly, and they may be given access to tools, systems, and sensitive information to complete tasks. That creates a new kind of security challenge. A system cannot just ask whether an agent has permission to do something. It may also need to consider what the agent is trying to do, what came before, and whether the action makes sense in context.
Daylight and Bay approach different sides of that challenge. Daylight pairs agent-driven security operations with experienced security engineers. Bay focuses on governing agent behavior, examining actions as they happen and applying controls before work proceeds.
Their funding announcements underline a broader point for entrepreneurs: enabling technology can generate demand for protection, assurance, and oversight. As a new tool becomes more capable, people need confidence that it will work safely inside the real-world systems they already depend on.
That does not mean every startup should build another security product. It means that when a technology is adopted, the new risks and operational gaps it creates can become important business problems in their own right.
The second wave often starts with what is missing
The first wave of a technology cycle attracts attention because it is visible. New products launch. Users experiment. Companies race to adopt the tool that promises to change how work gets done.
The next wave can be quieter. It is made up of the tools and services that make the first wave dependable: storage, security, governance, integration, monitoring, and support. These needs may sound less exciting than the headline technology, but they can become essential once adoption moves beyond a few early users.
That is the pattern these announcements bring into focus. Agents need memory to carry knowledge across tasks. They need security to work in sensitive environments. And they need governance so organizations can decide what they are allowed to do.
When three companies raise a combined $165 million around those needs on the same day, it is a strong signal that investors see infrastructure as part of the AI opportunity, not an afterthought. It is not proof that every company in the space will succeed. Funding is a snapshot of investor conviction, not a guarantee of product-market fit. But it does show that important business questions are shifting from “What can an agent do?” to “What must be in place before we trust it to do more?”
A practical way to look for your own opportunity
If you are building a company, this moment offers a useful way to study emerging markets. Watch the first wave of products, then pay close attention to what repeatedly slows them down.
Ask:
- What do users have to repeat, check, or repair?
- What happens when the technology encounters sensitive information or an unfamiliar situation?
- Which tasks still need a person to review the result?
- What tools are businesses adding around the new product just to make it usable?
- Where does the current workflow break when usage increases?
The best answers may point to a business of their own. Sometimes the opportunity is software. Sometimes it is a service that combines technology with human expertise. Sometimes it is the missing layer that lets a company use the new tool without rebuilding everything around it.
Be specific about the problem. “AI needs infrastructure” is a trend. “Teams cannot preserve task context across their existing document systems” is a problem you can investigate with customers. “Companies need AI security” is broad. “Security leaders need to understand and control an agent’s actions across a session” gives a founder a sharper place to start.
The opportunity is in making adoption workable
The funding news is a reminder that successful technology shifts create room for more than one kind of winner. The agent itself may get the spotlight, but the systems that make it trustworthy, useful, and manageable can become businesses with their own momentum.
For entrepreneurs, that is an invitation to look closely at the friction around a fast-growing tool. What does it forget? What can it access? Who checks its work? What would make a cautious organization comfortable enough to use it for something important?
Those questions do not require you to chase every new trend. They ask you to notice the practical needs that appear when a technology meets real people, real teams, and real consequences.
AI agents are moving into more serious work. The next opportunity may belong to the builders who make that work safer, more contextual, and easier to trust.