TankTank — Opportunity Analysis
Last updated: March 9, 2026 (discovery call — Alex, Mike, Alex Butaud)
Sources: Call transcript (Mar 9) · Original taste infrastructure one-pager
Purpose: Evaluate TankTank as a technical partnership opportunity and plan next steps
Who they are
TankTank (also referred to as Pimentank) — Amsterdam-based creative agency. Alex Butaud is the partner responsible for all execution (production) and self-appointed IT lead. Originally a film producer.
Small team: creatives, a client director named Lara (builds the Miro brand boards), project managers, and producers. Alex B handles everything from million-euro film productions to banners, websites, and installations.
Alex B insists on the distinction: "I say it always because people come to me and say that you are a production company. I say, no, I'm a creative agency. Our main competence is creativity. It's not production."
AI investment: ~1,800 EUR/month on AI subscriptions (ChatGPT, Claude, Gemini, Midjourney, others). Every creative has access to every tool. Conscious decision to build AI culture internally.
What they do
Creative campaigns and production — full spectrum. Film production that used to cost 500K EUR is now done in 5 days with one person using AI for ~60K. Clients are still pushing back on price.
Banner format adaptation — major bottleneck. Clients like Allianz Direct need 500+ banner formats. Done manually in Photoshop. Media agencies demand PSD files. Built a tool called AdBrief AI for browser-based format generation, but it only exports PNG, not PSD.
Brand guideline development — currently building one for Reform House. Lara creates visual Miro boards per client with products, personas, brand positioning, CI guides.
AI image generation — commercial-grade. Exploring style training/embeddings (Nano Banana, X Fields) for consistent client brand imagery.
Tools and infrastructure
Tool Purpose ────────────────────────────────────────────── Google Workspace Everything (Drive, Sheets, Slides, Gmail) Miro Brand boards, client context visualization AdBrief AI Banner format generation (browser-based) Photoshop Primary creative tool, banner production ChatGPT/Claude/Gemini AI assistants (all creatives have access) Midjourney + others Image generation Previously: Wrike Project management (dropped for cost) Explored: Vertex AI Google Cloud, but Gemini processes in Stockholm Explored: Cursor Alex B tried to code tools himself, hit a wall
Key clients
Reform House — New client. Currently approving final style guide (large PDF with full brand CI). Alex B recommends starting here because they're defining everything from scratch — cleanest data. This is client #1 for the pilot.
Allianz Direct — Insurance company. Present in Spain, Italy, Netherlands, Germany. Major banner production client (500+ formats). Previously asked TankTank for a branded AI image generation tool — TankTank sold it on paper but couldn't deliver the technical quality. Compliance-conscious. Media agency requires Photoshop file delivery.
itHappy — Sushi company. Client wants everything in AI. Asked Alex B: "Can you promise me all my products will always be the same in AI?" Alex B exploring style training for product consistency. Wants to start the data structuring project with this client too.
Foll Food — Oldest client, most data but least current activity. Wrong pick for starting — too much legacy, not enough momentum.
BLM/VLAN — Alex B scraped their 3,000-page website to JSON and built a custom GPT. Creatives tested it and found hallucination: "the seven pillars of our client was this, this and this... and that's not true." This is the pain point that demonstrates why structured, verified data matters.
What they want from GG
A technical partner. Alex B has the vision — he designed a detailed architecture diagram of a "TankTank AI Tool" with orchestration agent, per-client sub-agents (company info, CI, campaign history, briefings, legal, competitor analysis), chat interface, and usage logging.
He can't build it. He tried Cursor, followed ChatGPT's instructions, and hit a wall: "I don't know what I'm doing. I'm just doing what ChatGPT tells me to do."
He explicitly wants to be involved in the technical side — not just receive deliverables, but learn and participate.
What we showed them
Alex F demoed Swerk:
- Customer management with all documents indexed via RAG
- Chat interface using Claude API
- Agent and skill system (task-based playbooks, not role-based agents)
- Document versioning and collaboration
- MCP server integration — use from GPT, Gemini, Claude Desktop natively
- Pipeline concept: transcribe calls → extract personas/JTBD → generate documents
Mike explained the agent philosophy: the models are good enough that you don't need specialist agents (copywriter, art director). You need good culture-level instructions and task-based skills. "Instead of giving them a task, you want it to know more generically... convey more culture."
The bigger opportunity
Joint go-to-market. Mike proposed approaching TankTank's clients with AI-powered services they aren't buying today. The internal tools are table stakes — the real revenue is in selling AI solutions through TankTank.
Allianz Direct image tool: TankTank already sold them on the concept of a branded AI image generation tool (anyone in the company can prompt a picture, always in the same brand look). They couldn't deliver technically. With GG's capability, this becomes buildable.
Training and workshop revenue: AI training sessions for client marketing teams. Creative pipeline tools. Internal marketing tools licensed to clients.
Mike: "I want to think about the things that we can do for your customers that they're not buying from you today."
IP discussion
Alex B raised a fair concern: "If I develop tools with you guys and we optimize them... basically, you have knowledge or an experience that you can sell anywhere else. And I don't say it's wrong. I was just thinking, okay, what's the value for me?"
Alex F's response: "Anything we build is yours." And: "The building is actually really quick. The durable thing here, the thing that lasts, is your actual business."
Mike's reframe: the real money isn't in the retainer for internal tools — it's in jointly going to TankTank's clients with new AI services. The internal tools are a prerequisite, not the product.
Agency economics (context)
Alex B on the compression: "We went from 500,000 to 60,000 and now you're still telling me it's way too expensive? I'm a bit scared for agencies because we use integrated production and we used to make our margin with production... creative is not rentable anymore."
The margin squeeze is real. AI production is collapsing prices but clients expect the savings. The agencies that survive will be the ones that find new revenue streams (AI services, licensed tools, data-driven offerings) — not the ones that keep selling time cheaper.
Revenue potential
Revenue stream Estimate ─────────────────────────────────────────────────── Monthly retainer (tech partner) TBD — needs pricing discussion Allianz image tool (if rebuilt) Project-based, potentially significant Client AI training/workshops Per-session or retainer add-on Joint GTM licensing deals Revenue share on new AI services
For context: Alex B already spends ~1,800 EUR/month on AI subscriptions alone. The retainer needs to be justified against that baseline — we need to deliver more value than another tool subscription.
What we don't know
Will Lara and the creatives actually adopt? Alex B is enthusiastic but he's the partner, not the daily user. Lara builds the Miro boards. The creatives do the work. We need their buy-in.
How much data cleanup is needed? Alex B described their server as "like the bin from home." Mike pushed back: let us ingest it raw and use AI to sort it. But the reality may be messier.
Is the Allianz Direct image tool opportunity still live? They sold it on paper and couldn't deliver. Has the client moved on, or is there still appetite?
MOU structure. Alex F raised the need for a memorandum of understanding. Nothing defined yet — needs to cover scope, IP, confidentiality, pricing, duration.
Retainer pricing. Alex F: "We haven't completely figured out what to charge for this kind of thing, just to be honest with you." Needs internal discussion between Alex and Mike.
Next steps
1. Finalize NDA. Already drafted and sent back. Alex B confirmed it's in process.
2. Get Reform House Google Drive access. Alex B will share once NDA is signed. This is the pilot client.
3. Call with Lara. Understand how the team uses client context today. Get the Miro board walkthrough. Essential for adoption.
4. Call with one creative. Test what we build with an actual user. Get feedback on what's useful vs. what's noise.
5. Alex + Mike: pricing discussion. Define the retainer structure and come back with a proposal.
6. Share proposal / MOU. Cover scope, timeline, IP ownership, pricing, and what happens after 3-6 months.
Call transcript (Mar 9, Alex + Mike + Alex B) →