How to Build an AI Workforce with Lindy AI

TL;DR

Lindy AI lets users build complex agents through plain-English conversations, then use Computer Use to operate websites and applications, including tools without APIs. Practical workflows include customer support, personalized LinkedIn outreach, sales nurturing, recruiting, grocery ordering, and social media management, with human approval added before sensitive actions such as issuing refunds.

Transcript

I just saw the future of AI agents and I'm not  even overselling it. I had the founder of Lindy AI show us new stuff around how to use plain English  to do vibe marketing, to build out AI agents that make you money and make you more productive.  A lot of people have promised this. A lot of people have promised AI agent platforms that do  this, but ... Read More

Key Insights

• Agent Builder is a conversational tool for creating complex AI agents from plain-English instructions. Instead of requiring users to engineer every workflow manually, an agent interprets the requested goal, builds the operating logic, and allows the user to refine its behavior through further conversation.
• Computer Use enables agents to interact directly with websites and applications in ways similar to a human operator. This capability expands automation beyond tools with conventional integrations, making it possible to work with systems that lack APIs or impose expensive and lengthy API-access processes.
• Customer support automation is more useful when it can complete operational tasks, not merely retrieve policy information. An agent can receive a ticket, inspect an order in Shopify, prepare a refund through Stripe, and request human approval before completing a financially sensitive action.
• AI agents provide consistency and scalable execution once a workflow has been configured successfully. The discussion argues that this can simplify seasonal e-commerce operations, although it also acknowledges that agents are imperfect and may require human involvement when decisions carry meaningful risk.
• An AI sales development representative is not an instant source of unlimited pipeline. Effective outbound sales still requires a proven motion, substantial iteration, and a full-time person who manages the agents, with the goal of making that person more productive rather than removing supervision.
• Multichannel outreach can place email, consent-based text messages, phone calls, and LinkedIn contact inside one workflow. The transcript specifically notes that text and phone outreach require opt-in, while LinkedIn messaging is described as the most effective channel for the discussed operation at that time.
• A lost-deal nurturing agent can monitor new customer wins and product releases for reasons to restart conversations. It uses stored details about a prospect's industry, company profile, use case, and previous objection to generate a timely and relevant follow-up message.
• Automation creates value by enabling work that people know they should do but routinely avoid. Personalized LinkedIn outreach and regular posting are examples where agents can reduce the effort and discomfort that otherwise prevent a potentially valuable activity from happening consistently.

Questions & Answers

Q: How do you build an AI agent with Lindy AI?

You build an agent by describing the desired behavior to Agent Builder in plain English. For example, the demonstration instructs the agent to accept a LinkedIn profile, tailor a message to that person's profile, and use Computer Use to send a direct message about AI agent services. The workflow can begin manually and later receive triggers from Slack, forms, or spreadsheet updates.

Q: What is Lindy AI Agent Builder?

Agent Builder is a feature that creates complex AI agents through conversation. A user explains what the agent should do, and the system turns those instructions into a workflow that can be refined through additional dialogue. It is characterized as an AI agent creating another AI agent, reducing the technical expertise and setup time previously associated with agent development.

Q: How does Lindy AI Computer Use work?

Computer Use allows an agent to operate websites and applications in a manner similar to a human using a computer. It can navigate services such as LinkedIn, Shopify, or Stripe to complete workflow steps. This is especially useful when a system has no API, when its integration is difficult to obtain, or when direct interface interaction works better for the task.

Q: How can AI agents automate customer support?

A customer support agent can answer questions using retrieved business information and also perform follow-up operations. It can check an order's status in Shopify or prepare a refund in Stripe. For a sensitive action, the workflow can notify a human with the proposed step, wait for a yes-or-no decision, and continue only after receiving approval.

Q: Can an AI sales agent replace a human sales representative?

The discussion recommends using AI sales agents to supplement a human rather than expecting complete replacement. Outbound sales remains difficult, even with human representatives, and a new motion can require months of heavy iteration. The proposed model keeps one person working full time to manage the agents, refine their behavior, and become substantially more effective through automation.

Q: How can an AI agent nurture lost sales opportunities?

A nurturing agent can store each lost opportunity's use case and reason for not buying, then watch for relevant events. One event is a new customer win in a similar industry or company category. Another is a product release that resolves the prospect's earlier objection. The agent can then send a personalized follow-up and invite the prospect to book another call.

Q: How can Lindy AI automate LinkedIn outreach?

A Lindy workflow can receive a LinkedIn profile, examine the person's information, draft a tailored direct message, and use Computer Use to send it through LinkedIn. The initial demonstration uses a manual message as the trigger, but the same workflow could start when someone sends a Slack message, submits a form, or is added to a spreadsheet.

Q: Why use Computer Use instead of an API integration?

Computer Use can reach applications that do not provide an API or make API access costly and slow. The transcript cites healthcare record systems as an example, stating that some integrations can cost $100,000 and require a year-long review. Direct interface operation can also work better than an available API for certain tasks, such as sending messages through LinkedIn.

Summary & Key Takeaways

• Lindy AI introduces Agent Builder and Computer Use as two complementary capabilities. Agent Builder creates workflows through a natural-language conversation, reducing the technical difficulty of configuring complex agents. Computer Use allows those agents to navigate websites and applications much like a person, including systems that lack convenient or affordable API access.

• Customer support is presented as a strong application because agents can move beyond answering policy questions. They can check Shopify order information, open Stripe to process an approved refund, and notify a human before taking sensitive action. The workflow combines automated handling, consistent execution, and human oversight at important decision points.

• Sales applications include personalized LinkedIn messages, coordinated outreach across several channels, and long-term nurturing of previously lost opportunities. The proposed nurturing agent monitors comparable customer wins and relevant product releases, then contacts prospects when a past objection becomes addressable. Human sales staff remain important for supervising agents and refining the outbound motion.