Bring us a hard problem.
Fixed scope, fixed fee, and the names of everyone who will do the work, all agreed before we start. Published research is contractually firewalled from every engagement. Conversations begin in confidence.
Four doors. One standard.
Ministries and state programmes
National AI strategy, sovereign deployment and regulatory design. The founding fellowship is being appointed on one test: that the people advising on a state AI programme have run one, not studied one. A half-day Sovereign AI Briefing for ministers, or a fellow-led Deployment Review that audits a live programme and reports what it finds, including when the finding is unwelcome.
Boards and Chief AI Officers
From pilot to production: strategy, governance and the deployment disciplines that decide whether a system survives contact with your own institution. A Board Briefing on deployment risk, or a Vendor Review before a contract is signed. We sell no software and resell nothing.
Building or backing AI infrastructure
We advise technology companies and their investors on entering regulated markets across the United States, the United Kingdom, the European Union, the Gulf and Singapore: which rules actually bind, how public procurement really decides, and where deployment is happening rather than where it was announced.
Funders of public-interest capability
Fund independent deployment capacity for the governments and regions AI-safety philanthropy overlooks: outside the US and EU, where the distance between AI ambition and AI governance is widest. Every funding relationship is disclosed and none carries editorial influence.
Four formats. Scope fixed on day one.
Briefing
A session for a board, cabinet or leadership team on strategy, deployment risk and governance, led by a fellow who has made the same decision at state or enterprise scale. Fixed fee. No follow-on obligation, and no proposal at the end of it unless you ask for one.
Advisory Sprint
Two or three fellows, one deliverable: a deployment review, a procurement audit, a governance architecture. Assembled for the specific problem and dissolved when it is answered. Not a standing retainer, because a retainer rewards presence rather than resolution.
Market-Entry Counsel
For AI and infrastructure companies entering a market they do not yet know: regulatory navigation, procurement reality, and introductions calibrated to where deployment is happening. We tell you which of your assumptions about the region are wrong before you spend against them.
Sponsored Research
Underwrite the Institute’s recurring published research. Sponsorship is named in the publication, carries zero editorial control, and no sponsor sees a draft before the field does. No exceptions, including for the sponsor who asks.
From enquiry to engagement.
You write, we read
A person reads it, not a queue. You will have a reply within two working days, including if the answer is no.
A call, in confidence
Thirty minutes to establish what the problem actually is. Often it is not the one in the brief. Nothing is billed and nothing is committed.
Scope and fee, in writing
What we will deliver, who will deliver it, by when, and for how much. One page. If we cannot staff it properly, we say so here.
The work
Delivered by the named fellows, to the scope agreed. You get findings, not activity, and you get them whether or not they are what you hoped for.
What we will not do.
An institute is worth what its refusals are worth. These are ours, and they are contractual rather than aspirational.
We do not sell software
No product, no resale, no referral fee, no revenue share with any vendor we might assess. Our advice on a system cannot be worth money to us in any direction.
No client sees research first
Advisory work is firewalled from publication. No funder, client or sponsor reviews, approves or previews anything we publish, and the firewall is written into the engagement rather than promised in a brochure.
We decline what we cannot staff
If no fellow has actually done the thing you need, we will tell you and, where we can, name someone who has. A think tank that never turns work away is a consultancy with a mission statement.
No junior leverage
The people who scope the work do the work. You are not sold a partner and delivered an analyst, and the fee does not assume a pyramid underneath it.
Tell us the problem.
The more specific you are, the more useful the first reply will be. If you are not sure the Institute is the right home for it, say that too: we would rather point you somewhere better than take work we cannot do well.
On fees: briefings are a fixed sum agreed in advance, sprints are priced per engagement against a written scope, and sponsored research is annual. Nothing is billed hourly, because an hourly rate rewards the wrong thing. We will give you a number before you commit any time to us.
Prefer email? engage@complexai.org. Press enquiries go to the press page.
The models became general. The deployments did not.
Every engagement on this page points at one gap: the distance between a model that demos well and a system that governs well. That distance is where AI actually fails, and it is closed by engineering and accountability rather than by a better model. Multi-objective, high-stakes and auditable was the definition of Complex AI in 2019, and it is still the standard every mandate here is held to.