Enterprise AI · Venture building
Ziplabs explores where AI can change how enterprises work. It builds the product and the system around it, then develops the strongest opportunities into new businesses.
What we do
The gaps are rarely only technical. A product has to fit how people work, the system has to hold up in daily use and the economics have to justify the change. We work on all three together, from customer discovery through product development to the business case.
01
Customer discovery around an expensive problem, and a reason AI now changes what is possible.
02
Who the product serves, what it replaces and why an enterprise would adopt it.
03
Where the path stays fixed, where an agent decides and what it may touch.
04
Building the software, with the evaluations, tracing and recovery that hold up in daily use.
05
Evidence from users and buyers that the product is worth adopting and paying for.
06
Go-to-market, pricing, partnerships and the team the business requires.
Perspective
The model matters. So does the software built around it and the business it has to serve. These views shape what we pursue and how we build.
The opportunity is work that is expensive today and people willing to change how they do it. The answer may be a new AI-native product, an end-to-end workflow redesign or a narrower change in one part of the work. The choice of model, and how much autonomy it gets, follows from that.
Related terms: Workflow redesign · Adoption
Where the path is known, a fixed workflow is usually easier to predict and test. An agent loop is worth its cost when the next step depends on what the last one found. The architecture makes that choice. Orchestration coordinates what runs, which tools are called and where the work passes to a person. A second agent is justified only when it adds something substantive, such as work that can run in parallel.
Related terms: Agentic workflows · Orchestration · Multi-agent
The software around the model keeps task state, selects what each call sees and carries out tool calls. Context is chosen for relevance and stale material is cleared as the task runs. Decision memory that outlives a task, including what was decided and why, needs rules of its own.
Related terms: Agent harness · Context engineering · Decision memory
MCP connects agents to tools and data, while A2A supports communication between agents. Access scopes can limit what a connection may reach. The business still defines decision rights, including what an agent may do on someone’s behalf, when a person must intervene and who remains accountable for the outcome.
Related terms: MCP · A2A · Decision rights
Evaluations run tasks that resemble the real work and check the resulting records rather than the agent’s report of success. Those results guide the next change to the software. Traces show failures, latency and cost. Idempotent operations keep a retry from sending a payment or a message twice.
Related terms: Evaluations · Observability · Reliability
Token prices are the visible cost. Outcome-based economics start from what it costs to deliver a result the customer accepts, with discarded attempts counted and the customer’s review effort measured beside it. The choice of model shifts both, along with how quickly a result arrives.
Related terms: Unit economics · Model selection · Latency
AI usage can grow while the business stays the same. Value realization depends on the workflow changing around what the system does well, people adopting it at scale and outcomes that justify what it costs to run. Over time those changes become part of an AI-enabled operating model, which a competitor cannot copy by adopting the same model.
Related terms: Operating models · Scaled adoption · Defensibility
Current work
Selected explorations in organizational context, agent actions and evidence-driven decisions, alongside a broader interest in AI-native business applications and agentic workflows.
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.
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.
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.
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.
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.
Direction
Exploring agent-assisted investigations that connect evidence across systems, surface gaps and support faster, better-grounded human decisions.
Insights
Essays on enterprise AI, agents and the businesses around them.
· Updated · 12 min read
Most useful systems combine fixed steps with steps where a model decides what to do next. The design work is choosing which is which, and building the harness that makes both dependable.
Read the essay· 10 min read
When a model does part of the work, customers buy results and each result carries its own delivery cost. Adoption depends on trust in the output, and the lasting advantage sits in the workflow around the model.
· 6 min read
Early in a venture, a measure is worth tracking when it would change the next commitment of time or money. That rules out some familiar counts and brings commercial signals such as willingness to pay forward.
His background spans hands-on engineering in AI and machine learning, distributed systems and databases through product and operating leadership as CPO and CEO. At Ziplabs, he brings product strategy, technical architecture and commercialization together to develop enterprise AI software and businesses.
About ZiplabsContact
We welcome conversations with enterprise leaders, founders, investors and potential partners working on problems where AI could change how the work gets done.