Artificial intelligence may be the most discussed technology in business today, but what’s it actually like to manage AI agents as colleagues? That was the challenge addressed by Krish Ramineni, Co-Founder and Chief Executive of Fireflies.ai at the London Tech Expo this week.
Ramineni, who jointly founded the decade-old note-taking startup which received a $1 billion valuation last year described how the company has built a network of AI “agents” designed to function like colleagues, operating across departments, from customer support and sales to recruiting, engineering and executive operations.
Working continuously in the background, these systems perform tasks that once required large teams of staff. At the centre of that team is Fred, the company’s flagship AI assistant.
“Fred does everything from taking my notes to prepping me for meetings to helping me go through all of my emails every morning,” Ramineni told a packed audience at the London Excel Centre. “It’s like having a teammate who never sleeps, always works 24/7, and can take care of the mundane stuff so humans can focus on creative work.”
But Fred is far from alone in Fireflies.ai’s growing virtual workforce.
Ramineni said his most mature AI colleague helps manage customer support. Trained on historical tickets, product documentation, sales calls and webinars, it now handles roughly 80% of incoming queries. “It’s not just answering questions,” Ramineni said. “It can debug accounts, reset settings, even process smaller refunds. We only had to hire three extra people to manage a fifteenfold increase in users.”
Another AI bot helps out in the sales department.
“Sales isn’t structured like customer support,” Ramineni said. “So we started with the back-office tasks — updating the CRM, taking meeting notes. Then we gave the AI more agency: qualifying leads, answering questions, even running automated demos. The conversion rate went up 30% because leads got immediate answers instead of waiting weeks.”
Even the company’s recruitment department now employs a voice-based AI agent to conduct initial interviews, asking questions, listening for follow-ups, and compiling scoring rubrics. “By the time our HR team starts their day, they have a ranked shortlist with summaries from dozens of interviews,” he said. “In one month, the AI interviewed more candidates than our entire team did the previous year.”
Engineering, too, is being automated. “When a customer files a bug, the AI checks the code base and proposes a fix,” he said. “Then another AI reviews the code before it goes to staging. We’ve fixed thousands of bugs this way, and it runs 24/7.”
Each of these agents draws from what Ramineni describes as a company-wide “knowledge graph”, aggregating emails, documents, meeting transcripts and messaging data. “If the AI doesn’t have context, it fails. It’s not dumb, it’s just missing information,” he said.
Ramineni said he has even created his own so-called ‘Krish-bot,’ which answers questions to staff based on previous answers he has given.
“One of the coolest examples of this is when I am offline, there is a Krish-bot that goes into my [Slack] channel and is able to answer questions based on things that I have said in the past. It’s not a perfect replacement. I hope my people don’t want to replace me just yet, but it is a good way for me to be able to always be there to help my team at any moment in time, while you’re sleeping. Your AI should be working 24/7; your competitors are going to be deploying all of these AI employees. So your AI employees will need to compete with their AI employees.”
There’s plenty of other editorial on our sister site, Electronic Specifier! Or you can always join in the conversation by visiting our LinkedIn page.
