🗒 Summary
Humberto and Monica joined us to share how Influur - a ~90-person startup - went from zero AI adoption 18 months ago to 298 AI workflows built, 10,000+ hours saved per month, and a company where the legal team builds legal AI agents and the CEO competes in hackathons. And the People team leading by example was the thing that made it possible.
📡 The Journey
👉 The aha moment came from a job interview
A product manager candidate in San Francisco showed up to their interview with a working Lovable prototype of the product Influur was planning to build. The team had barely heard of Lovable. That was the moment they knew they needed to change how the whole company thought about AI.
👉 They ran a first hackathon - and it was bad
They invited both technical and non-technical teams. 90% of the outputs weren't good. But the point wasn't the output - it was identifying skill gaps and giving people a first experience. A few prototypes worked, and that was enough to build on.
👉 The technical team pulled ahead - and left everyone else behind
After restructuring the product and tech teams and merging them, they were shipping multiple times a day using AI tools the rest of the company had never heard of. The non-technical teams were still at square one and afraid AI was coming for their jobs.
👉 The People team went first - on purpose
Before going to the leadership offsite in December 2025 to pitch the company-wide transformation, Monica and the People team had already been quietly building and automating their own processes. The pitch to resistant leaders was simple: if HR can do it, you can do it. That opened the room.
👉 They built an org chart that includes agents
The current org chart at Influur shows people, AI agents, and fractional workers side by side. Monica's entry shows her directly leading 5 people - and all the agents she's built. It's a visible, structural signal that agents are part of the team.
⚡ What They Built: The AI Challenge
A two-week mandatory programme rolled out in phases across all non-technical teams:
Week 1: Daily learning challenges. Each day covers a new tool or skill - prompt engineering, building automated workflows, using agents, Lovable, Claude. Self-directed, not classroom-taught.
Week 2: Two fully protected days. No Slack, no email, no calls. Just building. The output: a working AI agent relevant to your function.
Demo Friday: Everyone presents a 3-minute pitch. Leaders build and present too - the CEO competing alongside the team was a genuine morale moment. One person won an iPad. Those who only presented a GPT (not a true agent) repeated the challenge in the next phase.
This is now baked into onboarding. Every new joiner completes the challenge in their first two weeks and shares what they built in a 3-minute Slack video.
⚡ The AI Tracker
Built in Lovable. Used every day. Contains:
- 298 AI workflows and agents built across the company
- 10,000+ hours saved per month (self-reported, tracked per submission)
- 91 people who have built at least one AI agent or workflow
- A leaderboard showing top builders by hours saved - updated monthly
- An org chart view showing people and agents side by side
- A learning library with structured challenges for every tool (Claude, Lovable, n8n) - updated as tooling evolves
- A Slack bot called Karen that lets people submit new AI use cases directly from Slack into the tracker without leaving the app
✅ What You Should Do: A Practical Step Plan
1. Start with a diagnostic hackathon - even if the output is terrible
A first hackathon where 90% of the work isn't good is still a success if it tells you where your skill gaps are and gets people over the fear of trying. Don't optimise for output. Optimise for participation and signal.
👉 We have a hackathon template in our AI playbook linked at the bottom of this page
2. Get the People team building first
If the People team has already built something - even something rough - by the time you go to leadership with the transformation pitch, you've got a live case study from the least technical team in the business. That's your most powerful argument.
3. Win leadership buy-in before you run anything company-wide
This is a cultural transformation, not a tool rollout. Without leadership buy-in you can't shift the culture. Frame it around commercial outcomes and net revenue per FTE, not AI for its own sake.
4. Make the challenge mandatory - then design it so people want to do it
Mandatory removes the opt-out, but the design has to do the motivating. Protected building time, a visible leaderboard, CEO participation, and a real prize create the environment where people push through the hard parts at 3am.
5. Let the technical team judge, not compete
Your technical team is already beyond the challenge level. Make them judges. It gives them status, gives non-technical teams credible feedback, and creates a visible bridge between the two halves of the company.
6. Track everything - hours saved, agents built, who's building
Self-reported data is fine if it's tied to visible outputs and correlated against performance. If someone claims they're saving 20 hours a week but their work output hasn't changed, that's a management conversation - not a tracker problem.
7. Bake it into onboarding from day one
If the challenge only runs twice a year it's a programme. If every new joiner completes it in week two and shares what they built on Slack, it's culture. That's the difference.
8. Give people tools and autonomy before you give them policies
Influur started with a $50 virtual card, two recommended tools, and one rule: build something. Policies came later, once people had enough context to use them sensibly. Leading with policy kills creativity before it starts.
9. Correlate AI adoption to commercial metrics for your board
Net revenue per FTE and ARR per FTE going up while headcount stays flat is the business case. If you can show the graph alongside the agent count, the board conversation writes itself.
💬 Q&A Highlights
How do you prevent gaming the leaderboard?
Culture does the work here more than controls. When 78 people are genuinely building every day and sharing in a dedicated Slack channel, the social norm is real participation. Hours are self-reported but tied to visible, reviewable outputs - if someone inflates their hours and their actual performance doesn't match, that surfaces quickly in a small team.
Are there measurable business results?
Yes. Headcount has stayed flat or decreased while output has increased. Influur tracks net revenue per FTE for their agency business and ARR per FTE for their AI product. Both curves go up from 2025 to 2026 alongside the rise in agents built. The correlation is visible and presented directly to the board.
How do you allocate time when everyone is already stretched?
It was an explicit company OKR from day one of 2026: become an AI-first company. That made it a business priority, not a side project. Once people see the value for themselves, the time allocation question stops being asked - people build because it makes their own job easier.
Do you have an AI usage policy?
Not a formal one at the start - deliberately. Teams got recommended tools, a budget cap, and a brief. Right now, non-technical teams have access to Claude and Lovable. Additional tools require a business case to Humberto. If approved, they get a virtual card to test. If it proves its worth, it gets rolled out more broadly. The technical team has stricter data access controls given proprietary data.
What's Karen?
Karen is the internal Slack AI agent built by the People team. She handles personalised internal comms, runs the entire performance review process through Slack (no need to leave the app), and manages AI use case tracking submissions. Four of Influur's five internal agents were built by non-technical people.