AI CONTROL investor deck15 slides · Human authority over what AI agents do
Download PDF PowerPoint 12-slide summary
The ask

€750,000

The raise takes AI Control from a tested build to a product we can sell. It funds 18 months and gets us to 3 paying pilots.

18 months · 3 paying pilots · See the slide
Planned shares. No pricing has been tested with a customer yet. Nothing here is an offer, or a forecast of returns.
Slide 1: I N V E S TO R P I T C H · N O V E M B E R 2 0 2 6 AI Control Human authority over what AI agents do. Every action asks first, a person stays in charge, and there is a signed record of who allowed what. Built, pre-launch
1 / 15
Slide 2: T H E P R O B L E M AI agents now act, and nobody can prove who allowed it They do, not just say Agents place orders, change systems, send messages and steer machines. One mistaken step can move money or stop a line. One
2 / 15
Slide 3: W H Y N O W Three things changed at the same time 01 Agents got hands Models can now call tools, run code and operate software on their own. The gap between an answer and an action has closed. 02 They reached real system
3 / 15
Slide 4: T H E P R O D U C T One question before every action: is this allowed? The agent asks AI Control before it acts. The answer comes back in one of three forms. Nothing is allowed unless a rule says so. Every answer is writ
4 / 15
Slide 5: H O W I T W O R K S From the agent’s request to a record anyone can verify AI agent Wants to do something SDK guard Asks first. Stops if there is no answer Decision Rules, risk score, operating mode Human Approves the ri
5 / 15
Slide 6: W H AT A C U S TO M E R G E T S Six controls that work today Default-deny rules Each agent gets scoped grants. Anything outside them is refused. Human approvals Risky actions wait for a named person. The requester cannot
6 / 15
Slide 7: A D E C I S I O N I N P R A C T I C E The same agent, three requests, three outcomes Read last month’s invoices Covered by a read-only grant for the finance agent. ALLOWED Refund a customer above the limit Inside its rol
7 / 15
Slide 8: W H AT I S B U I LT The control plane exists and is tested 81% of the core control plane is built: sign-in, decisions, approvals, audit, console, SDKs 1,580 automated test cases across the backend, the console and both S
8 / 15
Slide 9: W H E R E W E S TA N D Built, not yet launched WORKING TODAY Organisations, roles, single sign-on, passkeys Allow, deny or ask-a-human decisions Approvals with separation of duties Signed, hash-chained audit evidence Con
9 / 15
Slide 10: W H AT M A K E S I T D I F F E R E N T It decides before the action, and the proof is not ours to edit A gate, not a camera Monitoring tools report what an agent did. AI Control answers before the action happens, and the
10 / 15
Slide 11: W H O B U Y S I T Teams putting AI agents into work that matters Operations and IT Agents that change systems, tickets and infrastructure. They need approvals and an audit trail. Strongest fit today. Finance Agents near
11 / 15
Slide 12: H O W I T E A R N S A subscription per organisation, sized by agents under control PILOT €4,500 for 3 months One team The working core Our help to set it up TEAM €1,500 a month Up to 25 agents Single sign-on and approval
12 / 15
Slide 13: T H E N E X T S I X S T E P S From a tested build to a product we can sell STEP 1 Deploy Put it live and tag the first release STEP 2 Prove the tests Run the full suite on that build and keep the result STEP 3 First anch
13 / 15
Slide 14: R I S K S W E N A M E O U R S E LV E S Four open questions, and what we do about each The chain proof is unproven No anchor has been broadcast yet. Step 3: one real proof on Solieum, verified by an independent script. Co
14 / 15
Slide 15: T H E A S K €750,000 to take AI Control from built to sold 25% Launch Hosting, monitoring, backups and the first release. 30% Trust Independent security review, legal texts, the name. 45% First customers Pilots, real con
15 / 15