The Org Chart Is Getting a Second Layer

On September 29, 2026, OpenAI introduced dots: persistent agents with their own cloud computers that can work across connected applications. It also previewed specialist dots with distinct identities, credentials and defined organizational responsibilities. Those specialist deployments are beginning with focused enterprise pilots; broader teams of dots remain part of the company’s stated direction. OpenAI

The announcement brings a larger question into focus. As software takes on more work without a person directing every step, how should that work be organized?

The answer involves more than choosing a model. Someone must define responsibilities, maintain context, manage dependencies, review exceptions and determine whether the work was completed correctly. The org chart begins to acquire a second layer—not because agents become employees, but because delegated work needs an operating structure.

Yet a coordination requirement does not automatically create a new independent software business.

Our hypothesis is that the strongest opportunities will emerge where valuable work crosses systems, requires judgment and lacks a clear software owner. The companies that matter will make those processes easier to deploy, more dependable and cheaper to operate—while proving that customers cannot obtain the same result from their existing stack.

The new problem is discretion, not persistence

Software has long executed scheduled tasks, responded to events and waited for approvals. Workflow infrastructure already addresses persistence, retries and recovery. Temporal’s agent reference architecture, for example, separates durable orchestration from individual actions and supports explicit pauses for human decisions. These capabilities do not need to be reinvented simply because a model is involved. Temporal

What is changing is the range of work software can attempt without every step being specified in advance. n8n’s new agent functionality illustrates the distinction: agents can decide which steps to take, while predefined workflows remain available as controlled tools. A workflow can also call an agent when a particular step requires more flexible reasoning. The functionality remains in preview. n8n Blog

Consider a hypothetical manufacturer preparing a customer quotation.

A standard request might follow a predictable sequence: check inventory, apply pricing rules, estimate delivery and obtain approval. A less standard request might require interpreting an ambiguous specification, asking a clarifying question, identifying a substitute component and deciding whether engineering needs to review it.

The opportunity is to automate more of that variable work. The difficulty is keeping the process reliable while allowing the path to change.

A revised specification might invalidate an earlier price. An alternative component might require a different delivery commitment. An approval might apply to an outdated proposal.

These are not entirely new coordination problems. The new challenge is allowing software to choose how to respond without undermining the controls that make the process trustworthy.

Permission to update a quotation does not establish that the quotation is ready to send.

Illustrative workflow / 01

A changed request can make earlier work stale

The agent may decide what to do next. The process still has to track which decisions need to be made again.

01

Customer request

A specification changes

02

Product choice

Recheck the component

03

Commercial terms

Recalculate price and delivery

04

Human approval

Review the current proposal

One phrase, several markets

“Agent orchestration” can obscure the distinction between several different businesses.

Execution infrastructure keeps a process running: preserving state, handling failures and coordinating individual actions. Its value is primarily technical reliability. Temporal’s architecture illustrates this layer. For underwriting, the relevant questions concern developer adoption, production workloads and the cost of operating them—not necessarily whether employees use a shared AI workspace. Temporal

Governance infrastructure establishes which agents exist, what they can access and how their activity is supervised. Microsoft positions Agent 365 as a control plane for observing, governing and securing agents, including those built by ecosystem partners. The commercial question here is whether a new product offers enough beyond existing identity, security and administrative tools to justify a separate purchase. Microsoft

Business coordination software helps a team complete a process. Dust’s Vanta deployment illustrates a shared layer connecting departmental knowledge and agents. Nexcade takes a more specialized approach, bringing together requests, rate sources, transport systems and internal rules for freight workflows, with human review points. Nexcade These products are closer to the work a business operator needs completed. Dust

The boundaries will overlap. But a product that manages agent credentials is not necessarily competing for the same budget as one that prepares freight quotations.

That distinction changes the investment case. We should ask which problem a company owns, who pays to solve it and what the customer would otherwise use. A broad category label is not a substitute for a defined market.

The independent opening is between systems

The argument for an independent platform often begins with model neutrality: enterprises will use multiple models, so they will need a neutral layer above them.

The first proposition does not guarantee the second.

OpenAI says it is working with Microsoft to integrate specialist dots with Agent 365’s governance and security controls. That offers a concrete alternative architecture: one provider supplies the agent, while an established enterprise platform supplies part of its organizational management. OpenAI

The strongest counterargument to an independent horizontal platform is therefore not simply that incumbents have more distribution. It is that customers may assemble an adequate solution from capabilities they already buy.

An independent vendor must beat that alternative after accounting for implementation, maintenance and the burden of managing another supplier. Supporting several models is helpful, but it is not enough.

We see a more promising opening where a process crosses several applications and departments, while no existing application owns the whole outcome.

In the manufacturer’s quotation process, customer requirements, production capacity, supplier information and commercial approvals might live in different systems. A useful product would not need to replace those systems. It would need to connect them into a dependable process—and reduce the work required to keep that process running.

The opportunity also needs a commercial owner. A workflow can cross organizational boundaries without having a budget attached to it. The product must solve a problem important enough that someone will sponsor deployment and remain accountable for the result.

This leaves room for both horizontal platforms and specialists. A horizontal platform might establish a repeatable way for teams to build and operate workflows. A specialist might arrive with more of the process already understood.

Neither needs to own the entire enterprise agent stack. But each needs a reason to remain in it.

Measure the work, not the agents

An organization deploying many agents is not the same as every task benefiting from a multi-agent architecture.

Google Research’s January 2026 evaluation of 180 agent configurations found that performance depended on the relationship between the architecture and the task. Multi-agent coordination helped on some parallelizable work and reduced performance on some sequential tasks. Adding agents was not a universal improvement. Google Research

There is an economic trade-off as well. In its June 2025 account of building a multi-agent research system, Anthropic reported substantially higher token consumption than ordinary chat interactions and emphasized that the value of a task must justify the additional resources. Those observations concern its particular systems, not a universal cost benchmark. Anthropic

The best coordination software should sometimes choose fewer agents—or a conventional workflow.

A more useful commercial example comes from Vanta. In a case study published by Dust, Vanta describes using departmental agents covering finance, compliance and customer feedback to prepare quarterly business reviews. The resulting workflow assembles presentation materials and supporting context. Vanta reports saving roughly two hours per week for each of 200 representatives, or approximately 400 hours per week in total. Dust

That is more informative than an agent count: a defined process, specific inputs and a reported reduction in preparation time. But it remains a vendor-published customer account, without a complete accounting of implementation, maintenance and review costs. It does not establish that Dust is the only way to achieve the result. Dust

The investment question is whether that outcome can be reproduced economically across customers.

We would measure the fully loaded cost of a completed, accepted workflow: model and infrastructure usage, implementation, ongoing support, human review and rework. Time saved in one department is less valuable if a comparable burden appears elsewhere.

This is also where a software thesis can turn into a services thesis. If each deployment requires extensive process documentation, data cleanup and bespoke engineering, revenue growth may conceal a labor-intensive delivery model.

Services are not inherently a problem. The question is whether successive deployments become more repeatable—or whether every new customer starts the work again.

The investment measure / 02

Count the work the customer accepts

An agent count or gross time-saving claim leaves out what it takes to produce a usable result.

The output

A completed and accepted workflow

Account for the full cost

Model & infrastructure

Implementation

Human review & rework

Ongoing support

The advantage has to improve with use

A platform can accumulate considerable information without becoming meaningfully better.

Conversation history, execution logs and stored preferences may make a product harder to replace. They do not necessarily improve its decisions. They can also preserve mistakes, outdated assumptions and inefficient processes.

Return to the hypothetical quotation workflow. Human reviewers might repeatedly flag a particular specification change as requiring engineering approval.

A useful feedback mechanism would turn those corrections into a tested improvement: identifying similar cases earlier, routing them appropriately and reducing unnecessary escalations without missing important exceptions. The record would need to include not just the original request and response, but what was corrected and whether the eventual outcome was acceptable.

The test is whether accumulated operating experience reduces the cost and difficulty of completing the next unit of work.

That improvement need not come from training a proprietary model. It could come from better routing, clearer business rules, more reliable evaluations or a better division between fixed workflows and model-driven decisions.

There are also three distinct advantages to separate.

Customer-specific operating knowledge can create switching costs. Reusable deployment methods can improve the vendor’s economics across customers. Cross-customer learning can improve the product itself.

One does not automatically imply the others. A company may become deeply embedded in individual customers while learning little that makes its next deployment easier. Conversely, a platform may develop reusable capabilities without needing to pool sensitive customer information.

The strongest business would combine operational relevance with repeatability: it becomes more useful inside an account while becoming less expensive to deploy and support across accounts.

What would strengthen—or weaken—the thesis

Annual recurring revenue growth and net revenue retention are important starting points. We would read them alongside evidence of increasing operational dependence.

Is the customer delegating more valuable work? Are workflows completing successfully with less correction? Are subsequent deployments faster? Does expansion reflect a broader role in the business, rather than simply higher model consumption?

The counterevidence matters just as much.

We would become less positive if customers mainly used the product as an interchangeable interface to models; if implementation staffing grew in proportion to revenue; if human review remained high without a corresponding increase in task complexity; or if customers could recreate the useful workflows in their existing software with little loss.

A further test is what happens as models improve. Does the product become more capable and economical, allowing it to take on additional work? Or do stronger models eliminate much of what the product was charging for?

The most attractive companies should benefit from model progress without depending on today’s model limitations for their relevance.

Some may ultimately become part of larger platforms. ServiceNow’s completed acquisition of Moveworks in December 2025 brought together an enterprise assistant, search and reasoning capabilities with an established workflow platform. It demonstrates strategic interest in combining these functions—not that every company in the category will earn an attractive acquisition outcome. ServiceNow Newsroom

Acquisition can be an outcome. It should not be the premise.

A second layer, not another bureaucracy

The org chart metaphor has a limit. Enterprises do not need to reproduce every management layer in software. They need a clearer division of responsibility and less effort spent moving work between people and systems. Human accountability does not disappear when execution is delegated.

Our view is that this creates room for substantial businesses—but not one undifferentiated market, and not necessarily one new platform that governs everything.

The independent opportunity is strongest where a company makes difficult, cross-system work reliably executable, then improves both the customer’s outcomes and its own delivery economics over time.

The emerging need is coordination. The investable business is the one that can deliver it better than the alternatives.

The next valuable layer will not be the one that creates the most agents. It will be the one customers trust with more work—and have a durable reason to keep paying for.