AI Agents Are Rewriting the Service Economy

Trusted vouching may be more predictive of opportunity than traditional credentials.

Cosign is built around the idea that a credible endorsement from the right person can matter more than a resume or public profile. This reframes networking as a reputation graph, where relationships and signals of trust become the core infrastructure for discovering talent and opportunity. It also suggests a shift away from static credentials toward more dynamic, social proof-based matching.

The Cosign Awards are designed to surface hidden high performers through community nomination and voting.

Rather than relying on top-down curation, the awards aggregate nominations from people who know the nominees best. That helps recognize underappreciated operators, founders, and products that may have outsized impact but limited visibility. The mechanism also reinforces the platform’s broader goal of making reputation more legible and searchable.

Cosign aims to unify fragmented professional discovery data into a source of truth for the startup ecosystem.

The speakers describe how relevant information about people, companies, and relationships is spread across disconnected systems and private networks. By combining profiles, lists, and curation, Cosign wants to make discovery more efficient and reliable. This matters because better data infrastructure can reduce friction in hiring, fundraising, and collaboration across the ecosystem.

The platform is intended to expand access to opportunities for people outside elite networks.

A recurring point is that talented people often miss opportunities because they are not in the right circles, cities, or institutions. Cosign tries to increase the 'surface area for luck' by making it easier for overlooked talent to become discoverable. This is especially relevant for early-career operators and builders who may not have strong conventional signaling.

AI agents could become the future interface for professional networking and matching.

The episode looks ahead to a model where a personal agent and a company’s agent handle initial discovery, fit assessment, and even negotiation. This is where concepts like agent-to-agent communication and MCP become important, since they would let AI systems query the network and act on behalf of users. The goal is to surface only opportunities that represent a genuine 'coincidence of intent' before escalating to humans.

Venture economics have shifted: early ownership matters less than it used to because later-stage dilution is much lower.

Tisch argues that Series B/C/D dilution has fallen significantly, so seed investors can still achieve strong outcomes without forcing extreme ownership. This changes how firms should think about portfolio construction, pro rata, and return paths in venture capital.

Concentrated funds are overstating their control over deal flow and outcomes.

He считает the claim that concentrated managers win most of the deals they want is unrealistic because great founders have options and investors cannot dictate terms. The implication is that venture is a competitive market, not a product where investors can impose a target ownership model; this is especially relevant for seed funds, portfolio sizing, and sourcing strategy.

The 'lean startup' playbook is too limiting for modern venture-backed companies.

Tisch believes the best companies often need meaningful capital and can benefit from being well-funded rather than overly capital-efficient. This matters because it challenges the assumption that discipline always means spending less; in fast-moving categories like AI or software infrastructure, undercapitalization can slow company creation and execution.

Box Group’s model prioritizes flexibility and founder fit over rigid ownership targets.

The firm invests from two $275M funds and can write checks from 1% to 10%, depending on the opportunity. That flexibility helps them adapt to founder preferences, market conditions, and round structure, rather than forcing board seats, control, or concentrated ownership where it does not fit.

Early-stage venture is best understood as a service business, not just a capital allocation game.

Tisch says the investor’s job is to understand what each founder wants and sell the right kind of support. This reframes venture relationships around trust, responsiveness, and usefulness rather than purely financial metrics; it also explains why Box Group avoids excessive internal negativity and looks for at least one partner with real conviction.

The skill that matters most in venture is seeing possibility, not just identifying risk.

He argues that it is easy to explain why a startup will fail, but the real edge is squinting and seeing why it might work. This mindset is central to early-stage investing, where uncertainty is high and technologies like AI, developer tools, and infrastructure can scale quickly if the founder and timing are right.

Eve is being positioned as an end-to-end AI platform for plaintiff law, not just a tool to reduce headcount.

The product is described as handling intake, qualification, case management, and settlement support across the full plaintiff-law workflow. That matters because it changes AI from a cost-cutting add-on into a system that can reshape how legal services are delivered. This end-to-end approach is what makes the business model feel more like software than traditional legal labor.

The main bottleneck in plaintiff law is labor capacity, not the lack of demand.

Firms often have more leads and potential cases than they can economically process, but they cannot afford enough staff to handle them. AI helps solve this by automating repetitive tasks and allowing firms to take on more clients without proportional hiring. This is significant because it turns previously uneconomic cases into viable ones and expands access to representation.

AI can dramatically increase attorney leverage, with some firms reportedly scaling from about 100 clients per attorney to 500.

That kind of productivity gain changes the economics of a law firm in a fundamental way. Instead of AI merely lowering expenses, it can increase revenue by letting each lawyer manage a much larger caseload. The example illustrates why legal AI may create entirely new growth curves for small firms.

The legal profession may evolve toward a software-like operating model.

As AI takes over more operational work, lawyers may increasingly act like supervisors of systems rather than pure case handlers. The episode suggests attorneys could become more like data scientists or engineers, continuously refining workflows and improving outcomes through software. This shift is important because it changes both the skills law firms value and how they scale.

Routine legal negotiation may increasingly be handled by AI agents, leaving humans to focus on higher-value advocacy and judgment.

The discussion points toward a future where AI agents manage repetitive back-and-forth, while lawyers intervene in strategic or contested issues. Concepts like agent-to-agent negotiation and automated dispute resolution could reduce the need for human effort in lower-complexity interactions. That could make some cases cheaper to resolve, but it may also raise questions about fairness, aggressiveness, and the role of courts.