- Industry
- AI, Software
fal is a generative media ecosystem for developers and creators to access the world's leading generative media models. With one API, fal customers can access 1,000+ models, with industry-leading inference speed. The fal Operations team’s job is to make sure that their fast-growing business runs as quickly as their product does.
For Erik Tchinnosian, his job changes with the needs of the company: he works on business strategy, unit economics, vendor management, and building out processes that help the company scale. Repetitive data queries and administrative work take away from the time he needs to spend investigating today’s issues and preparing for what’s coming next.
With Gumloop, he’s been able to build agents that can take care of that administrative overhead, and share those agents with his colleagues who need the same help.
“Everyone does things that they don’t want to do, and I think Gumloop is a great way to eliminate those things and be able to focus on the things that are strategic.” — Erik Tchinnosian, Business Operations, fal
Following questions without writing queries
“I used to have to do a ton of queries manually,” Erik says. Now, with his data analyst agent, Erik can ask questions (and follow-up questions) about user data, using natural language and without writing any manual queries.
Erik built the agent, and now everyone else on his team can use it too. The data analysis agent lives in Slack, and anyone with access can message the agent to get instant answers to questions like “Who were the top users of this particular model over the last three days?”
Automating CRM updates after sales calls
Immediately following a sales call, a salesperson can ask an agent to help update their CRM based on their last conversation. The agent retrieves the call transcript, the history of the opportunity as documented in their CRM, and fal’s sales playbooks that explain what each field in the CRM means and how they should be populated.
Based on this information, the agent then proposes updates. The agent only makes write changes to the CRM after a salesperson reviews and accepts the proposed updates. Erik deliberately chose to include the approval step so that the salespeople can remain in control of what goes into their system of record, without them having to do the manual work of maintaining many detailed fields.
Investigating billing disputes
Erik also uses Gumloop agents to handle more specialized operational tasks, helping Erik with his responsibility of understanding fal’s unit economics. For customer invoice questions, an agent gathers invoice and usage data, prepares documents that can be shared with customers, and identifies next steps for his fal colleagues. Other agents support usage and billing reconciliation.
Let the people with the context build the tools
Erik’s approach to adoption is that people who want to solve a problem should be able to try building an agent for it.
“You’re the person with the most context on your problem,” Erik says. When the people closest to the task can build the agents they need, team members can delegate tedious tasks while keeping human judgment where it’s needed. That way, they have more time to work on and solve complex, emerging problems.
“Anyone who asks me to help them solve a problem in their company, or ‘How would you solve this?’—I’m always telling them to use Gumloop, because it’s the easiest way to create an agent for yourself.” — Erik Tchinnosian, Business Operations, fal
Want to use AI agents to streamline repetitive work around your team’s core systems? Talk to Gumloop about getting started with enterprise AI.