Subscribe
Sign in
What AI agents do you have running in production?
There are so many new AI agent platforms ( @Wordware @Lindy @CrewAI @zapier and so on) that I'm finding myself curious how everyone is using them.
What AI agents are you using in production? What do they do? Are they working and reliable? What would make them better? Are they replacing roles? Augmenting existing ones?
And if you're not yet using agents in production, why not? What should these companies build that would make you adopt it quicker?
9
Zapier
Heya —happy to chime in!
I'm leading the New Products Support team at Zapier, so I'm able to first-hand experience how our customers are using Zapier Agents on a daily basis. The use-cases (all in production!) that I have seen already are:
Sales
Lead qualification and enrichment: incoming leads are enriched with public data (from the web) and uses tools like Clearbit/Clay, etc. and scored accordingly.
Lead funnel automation: nurturing leads through various stages. Often tied into the first use-case of qualification, but not exclusively so. Alerting team members in Slack to ICP/VIP leads is super powerful as well.
Sales call-prep: pull info from a CRM and additional data sources to prep for your next meeting.
Support
AI deflection: equip an Agent with your voice and help documentation, and an Agent is able to trigger on a new Zendesk ticket, and automatically respond (or add an internal comment for a human-in-the-loop experience).
Categorization/labeling: being able to classify and route tickets helps streamline internal operations. This can include sentiment analysis based on your own criteria.
HR
Job application: an Agent can categorize, summarize and recommend job applications.
Finding jobs: an Agent can look daily at job boards and alert you whenever a relevant job is posted. This can be paired with visiting multiple job boards, summarizing these and sending an email.
Operations
Impact tracking: using your own framework, trigger an Agent and analyze business impact and store this in Zapier Tables to keep track of how as an IC or Mgr you're contributing to your company's success.
Reporting: run a weekly report on business data, highlight key points (either good and/or bad) and get recommendations on what to focus on next week.
Task management: adding tasks with an Agent, get a daily reminder of the goals for the day, daily prep (number 1 focus), etc.
Etc.
I think the power of Zapier Agents, specifically, lies in the fact that it connects with all Zapier integrations (more than 7,000 of them now!). Being able to create, read, update and/or delete data in an agentic workflow across all business apps is a killer combo—and I'm personally super bullish on the concept of Agents.
What about reliability?
Any agent will rely on underlying LLMs. The better they get, the better agents will become. There's a lot of extra magic that happens under the hood to improve all this, but I'm personally impressed with what AI (and thus agents) can do, given the right instructions are given to the agent.
Hope this helps!
@leo_selie Great examples. I'm excited to start playing with Zapier agents.
I can smell problem discovery process here hah!
Im using the following AI Agents in production:
Agent that fetches YT videos (particular queries) , transcribes them and gives me summary
Agent that fetches Crypto data from 3rd party APIs
Agent that takes the data from above, and using RAG with economic analysis documents gives me a market state summary
Agent that fetches posts from people i follow on Linkedin. That is an actual bot though, running on PC not a tool which calls API and the Agent is using. But i can still call it an agent.
What is missing is easy multi-Agents. Programmatically. I find the current no-code tools with UIs, really "boring" and monolithic and i dont use any. Code is King. Or Queen.
@cryptosymposium what do you mean by multi-agents? Isn't the primary use case of the no-code "agent builders" to be able to stitch together a bunch of workflows?
@steveb yeah yeah it is. And that is what they doing, but as with every new tech, i think in terms of UI, it targets people only familiar with tech. Also, multi-Agents is something that can be more sophisticated. I think right now, all the platforms make Agents in a deterministic way. i.e
1. Connect tool X
2. Do this
3. Then connect tool Y
4. Then do this
However, the interaction between the agents and the reasoning without human intervention is still not properly developed. Cause it is one thing to automate things, it is another thing for the things to think for themselves. Not in smarter than us ways, but in "smart enough, dont need humans" ways.
Hope you get my point.
@cryptosymposium
I’ve been trying out @autogen, @Cognosys, and Superagent for different tasks. They work well, but I feel like they could be better at remembering context and working smoothly with other tools. Would love to hear how others are using AI agents in real life!
My AI agents vary based on use case but I have production-ready ones for automation and decision-making.
We are experimenting with AI agents with a select few of our customers. Since we are an embeddable workflow builder designed for our client's customers (external use), it introduces different challenges with user experience and security. But we're pretty excited about the potential. We'll likely do a PH launch once we've refined the approach.