Skip to content
Ziplabs

Work

Opportunities worth building toward.

Our interests span AI-native business applications, agentic business workflows and the capabilities that make intelligent software useful and dependable in enterprise environments. These are selected areas of exploration, not boundaries around what we may build.

Exploring opportunities across enterprise AI

  • AI-native business applications

    New enterprise software whose product strategy starts from AI capabilities, from horizontal business applications to industry-specific solutions that rethink how work gets done.

  • Agentic business workflows

    Human-agent workflows that orchestrate a business process across enterprise applications, agents and people.

  • Organizational context

    Making organizational knowledge usable as enterprise memory, including decisions and why they were made, with provenance, temporal context, permissions and missing information kept visible.

  • Agent actions and accountability

    Enabling agents to act within explicit decision rights and delegated authority, with oversight and a reviewable record of what changed.

  • Investigations and evidence synthesis

    Connecting evidence across systems to reconstruct events, examine competing explanations, identify uncertainty and leave the consequential decision with a person.

Selected work

A closer look at three opportunities we’re exploring in greater depth, including the enterprise problems and potential product directions behind them.

Organizational context

An agent working across a company needs the relationships, decisions and history that explain the record in front of it, and a way to see where each part came from.

The opportunity

Enterprise information lives in applications built for people. Records, messages, decisions and the relationships among them sit in different places, and they change. Some of what the organization knows was never written down, or survives only in fragments. An agent that answers from one retrieved page can miss who decided what, when and on what basis, and still sound sure of a result it does not have.

Why it matters

Work an agent takes on is only as good as the context it can use. Without the organization’s own history, an agent repeats questions that were already settled, treats an old record as current or proceeds without knowing which decision still stands. Where the record is missing, the useful result is to recognize the gap rather than fill it in. Making context usable, with its source, its age and its gaps visible, is part of building software people can rely on.

Direction

Exploring how agents can draw on an organization’s records, relationships and decision history across systems, with provenance, currency, permission boundaries and unknowns made explicit.

What the situation depends onIllustrative

Records in time: a January review, a March renewal decision and a June change of budget owner. There is no record of whether a concern was raised. The agent is not to fill that gap.

Agent actions and accountability

As agents begin to act across enterprise systems, organizations need to grant useful autonomy and still know what was done, on whose authority and with what result.

The opportunity

Most enterprises have established ways of deciding who may do what, built largely around people and services with well-understood roles. Agents stretch those practices. An agent may act for a person across several systems, in a task whose steps no one specified in advance. With too little autonomy it is not worth deploying. With too much it can do more than the task requires. A vague grant is a separate problem: afterward, no one can say with confidence what the agent was entitled to do.

Why it matters

As agents move from answering questions to taking actions, an organization has to know who delegated the work and what the delegation covered. What an agent was allowed to do, what it tried and what changed can each differ, and accountability breaks down when one is inferred from another. Keeping them distinct lets a reviewer see the granted authority, the attempted action and the actual outcome separately. That record is what makes agent governance possible.

Direction

Autonomy granted in proportion to what a task requires and what an agent has shown it can do reliably, with the authority behind each action clear and its effects open to review.

One task, three recordsIllustrative

An agent resolves one customer's billing complaint on behalf of a support lead. It refunds a duplicate charge, which it was permitted to do. It attempts to waive next month's fee without permission, and nothing changes. It sends a confirmation email it was permitted to send, but the email is not delivered.

Investigations and evidence synthesis

When complex questions span enterprise systems, relevant evidence rarely arrives in one place. The challenge is connecting those pieces into a coherent account of what happened, what remains uncertain and what deserves attention next.

The opportunity

Investigations across enterprise operations often depend on information scattered across different systems. Security incidents are one demanding example: alerts, logs, identity records and support tickets each tell part of the story. Teams still spend significant effort connecting those pieces, testing explanations and deciding what warrants action, often under time pressure.

Why it matters

Investigations unfold iteratively. Each finding can change the next question, making this a useful setting for agentic reasoning alongside established automation. A convincing conclusion without sufficient evidence can be costly. Making the evidence and uncertainty visible helps teams reach better-grounded decisions while keeping people accountable for consequential actions.

Direction

Exploring agent-assisted investigations that connect evidence across systems, surface gaps and support faster, better-grounded human decisions.

An incident, pieced togetherIllustrative

Four pieces of evidence arrive out of order from an alert, an identity record, a ticket and a log. They are drawn into one account that marks two findings as supported by the evidence and one question as still unknown. The response decision stays with the people accountable for it.

Working on something new?

We welcome conversations with people working on an enterprise problem an agent could take on, and with potential partners interested in building together.

Discuss an opportunity