Why We Reduced Fhinck from 50 to 6 (and Doubled Revenue)

"The executive who still says 'AI is for the technical team' is signing their own pink slip and just hasn't realized it yet. Most companies that invest in AI don't fail for lack of tools. They fail for lack of courage. Lack of code. Lack of walking into the back office and redesigning everything." — Paulo Castello, CEO Fhinck, November 2025

This is the firsthand account of Fhinck's transition to AI First, told by the CEO in the first person. No glamour. No pyrotechnics. What worked, what hurt, and what we learned.

The Starting Point — 2023, Before the Pivot

In January 2023, Fhinck had 11 years in the market and was recognized as the Task Mining company for Shared Service Centers (SSC). We had reference clients in finance, healthcare, retail, and manufacturing. We had 50 employees, well-structured processes, and growing revenue. We were winning awards.

By any traditional management playbook, the recommendation would have been: "keep going, scale further."

But something changed in 2022–2023. LLMs (GPT-4, Claude, Gemini) moved from technical curiosity to operational technology. And we, as specialists in understanding operations, began to see the obvious: agents could perform a significant portion of the work our own team was doing.

The question that stayed in my head for months was: "If we sell Task Mining to clients so they can become more efficient, why are we still operating like a 2018 company?"

The Decision — End of 2023

In 2023, I realized that the true revolution was not merely about seeing processes. It was about acting on them with autonomous intelligence.

The decision was to rebuild Fhinck from scratch as AI First:

• The platform would combine Task Mining (full visibility) + Autonomous AI Agents (that execute, decide, act)
• The internal team would be deliberately lean, multiplied by dozens of agents
• Operations would be redesigned — not automated — for what AI does well
• Legacy systems without a modern API would be replaced
• Culture would shift to rewarding speed over process

I did not sell this to the board with a polished slide. It was a direct conversation about real risk: "We're going this way. It will hurt. There will be friction. But there is no other path that doesn't compromise the company in 3 years."

The Execution — 24 Brutal Months

The reduction from 50 to 6 people did not happen all at once. It happened over 24 months, through natural attrition, retirements, voluntary career changes, and, yes, a few targeted departures when a role and the operation no longer converged.

The rule we applied at every departure was simple and painful: "Every exit became a test: 'Can AI do this? Or are we afraid to admit it can?'"

When the honest answer was "AI can do it," there was no backfill. Instead, we dedicated 2–3 weeks to building the agent that would take over that function.

The 6 Critical Moves

1. Make All Work Visible

We applied our own Task Mining platform to ourselves. We mapped every hour of every Fhincker's time. Where was the time going? Where was the rework? Where were manual processes happening in spreadsheets?

2. Brutal API Audit

Every internal system was evaluated: does it have a modern API? Yes → keep. No → replace. It hurt. We replaced systems that had years of investment behind them. But the alternative was living with agents that were partially blind.

3. Redesign Every Process

We did not automate. We redesigned. The difference is decisive: automating a bad process accelerates the error. Redesigning reconstructs the process assuming the executor is an agent.

4. Specialized Agents, Not Generalist Ones

Every important function became an agent. Customer service Tier 1, document processing, compliance alerts, financial close, prospecting. Agents focused on a defined scope, not a generic chatbot.

5. Sharpening the Axe — Weekly Ritual

Every Friday, we stop working and enter a classroom. We learn a new AI technique. MCP, agents, new models, frameworks. Without this routine, in 6 months we would be obsolete.

6. Culture Redesigned

Fhinck became for adventurers, not for the comfortable. Those who wanted predictable stability, rigid processes, and strict hierarchies sought other companies. Those who stayed embraced speed > perfection, autonomy > hierarchy, impact > activity.

What Happened to the People

This point deserves brutal honesty. The transition was not easy for the team.

"At Fhinck we went through this and it was a frightening journey. Redesigning the company for an AI-oriented future demands radical change (on the part of PEOPLE). From the executive to the intern."

Those who stayed had to reinvent themselves profoundly. We went from a company where each person executed specific tasks to a company where each person orchestrates agents and focuses on strategy, creativity, and relationship management.

For some, it was liberating. For others, it was unsustainable.

Those who left did so for various reasons: some felt the pace of change was incompatible with their life stage; some returned to more traditional companies; some started their own businesses. We stay in touch with former Fhinckers to this day through an active network — and most speak positively of the experience despite how difficult it was.

The Numbers — August 2025 (24 Months Later)

• Team: from 50 to 6 people (-88%)
• Revenue: doubled over the period
• Customer service: 96% without a human operator
• Delivery cadence: feature cycles 3–4x faster
• Geographic coverage: expanded to 15 countries (US, EU, LATAM)
• Active users: more than 800,000 under the platform
• NPS: rose alongside the team reduction (it did not fall, as many predicted)

What We Learned (and What Can Save Time for Those Starting Now)

Lesson 1 — Courage is scarcer than technology. The technology to go AI First is available. What is missing is C-level courage to make difficult decisions. Those who don't make them, stagnate.

Lesson 2 — APIs are life or death. Underestimating the impact of legacy systems without APIs was our biggest initial mistake. It nearly cost us 6 months. Start with the API audit.

Lesson 3 — A smaller team does not mean a heavier load. The common intuition is "if you cut 88% of the team, the 6 who remain will work 8x harder." That is not what happened. Each person works the same (or less) as before — because agents handle what should never have been human tasks in the first place.

Lesson 4 — Culture must be explicit. We published a public manifesto. "For adventurers, not for the comfortable." Without that, the cultural friction becomes diffuse and no one understands what is happening.

Lesson 5 — You cannot survive without continuous learning. Without the Sharpening the Axe ritual, in 6 months our team would have been outdated. The pace of the AI market does not allow for "studying when we get around to it."

The Question That Remains for You

If you are reading this as a CEO or board member, the concrete question is:

"Does your company have the courage to make this transition in 24 months, or will it wait and do it in worse conditions in 36 months?"

The window is open. But not indefinitely. Schedule a conversation if you want to understand how to apply this path in your operation. Fhinck built the platform precisely because we lived through this — and we help other companies live through it with more method and less suffering.