Why we built BIK Labs.

We started as an R&D consulting firm in 2017.

BIK was founded in 2017 as an innovation consulting firm. We managed R&D tax deductions, public funding, R&D payroll tax credits. Fifteen years deep in R&D projects for industrial companies across Spain.

But we were never a conventional consultancy. From day one, the question was: how do we do this faster, more precisely, and with less bureaucracy?

2022: we discovered AI and became obsessed.

When GPT-3.5 launched, we were among the first to test it for our work. R&D technical reports. Activity classification. Deductible base calculations. CDTI proposal writing.

It worked. Not perfectly, but it worked. A draft technical report in 20 minutes instead of 10 hours. We became obsessed. Claude, GPT-4, every new version. Every new model we tested the same day it launched. Not because we were trendy early adopters. Because every model improvement meant our consultants could manage more case files at the same quality.

2024: agents, not chatbots.

A chatbot answers questions. You ask, it responds, you copy the answer, paste it into your document. You do the work. The chatbot helps.

An agent executes tasks. You assign a work item: "write the state-of-the-art section for Project X's technical report." The agent reads the project description, objectives, R&D activities, and generates the complete section. Not a generic response — a text based on the real project data.

We built an internal multi-agent system. One agent wrote reports. Another classified R&D vs IT activities. Another calculated deductible bases. Another reviewed. All orchestrated with scripts, Excel, and a lot of duct tape.

It was ugly. But it worked. A consultant who used to manage 8 case files per year started managing 20.

The problem: we had nowhere to manage the agents.

The agents executed work. But where did they live? Where was their task queue? Who controlled what they could do? How did we know how much each one cost in tokens? How did we audit what they had done?

We tried Jira. One agent = one user. Assign issues. It works... sort of. But Jira doesn't know what an AI agent is. No token metrics. No approval gates per agent. It's a 2002 tool trying to manage a 2026 concept.

We tried Linear. Beautiful, fast, elegant. Same problem: it doesn't understand agents.

No tool on the market was designed for a world where AI agents are team members.

We built tools for ourselves. Then for our clients. Then for everyone.

If nobody builds it, you build it yourself. That's what we've always done.

We built a PM from scratch. Projects, cycles, epics, modules, kanban, timeline, table, calendar. Everything you expect from a modern PM. But with one fundamental difference:

AI agents are first-class team members. They have names. They have roles. They have configurable permissions. They have their own task queue. They have performance and cost metrics. They have approval gates. They have audit trails.

It's not an "AI generate" button stapled to a kanban board. It's execution infrastructure for agents. Three layers: BIA (conversational assistant that executes actions), Internal Agents (full LLM with workspace context, auto-provisioned by vertical), and BYOA (bring your own agent — connect your IDE via MCP, PM controls, IDE executes).

We use it every day. It manages itself.

BIK Labs is managed with BIK Labs. Not a slogan. Literally.

Our roadmap lives in BIK Labs. Our weekly cycles. Our CRM with client pipeline. Our wiki with architecture documentation. Our AI agents — Writer Agent for docs, QA Agent for tests, Developer Agent for code — all executing inside the product.

Every feature you see on biklabs.ai was managed as a work item in BIK Labs. If something doesn't work well, we're the first to suffer it.

AI agents aren't a feature. They're the future of work.

For a year we used it only internally. We improved it every week. We added CRM because we needed to manage clients without leaving the PM. We added wiki because we were tired of documentation in Notion and work in another tool. We added MCP so Claude Code and Cursor could talk directly to the PM.

And one day we realized: this isn't an internal tool. This is a product.

If for us — a 10-person innovation consultancy in Etxebarri, Bizkaia — it multiplied productivity 3x, what would it do for a 50-person development team? Or a solo vibecoder running 30 agents?

From Europe, with conviction.

We're not in San Francisco. We're in the Basque Country. And that's not a limitation — it's a decision.

Infrastructure in EU (eu-west-3). GDPR by design. EU AI Act compliance native. PostHog self-hosted — product analytics don't go to Google servers in the US. Our clients' data never leaves the European Union.

Article 50 of the EU AI Act: provenance badge on every AI output. Configurable human oversight via Gates. Audit trail up to 5 years. If you're going to use AI agents in your company, you need to know what they did, when, and who approved it. We give you that from day one.