Sorry this is a bit hastily written because I think EA needs better policy advocacy right away and I don't have much time to spare.
For the past decade, my job has been to advise decision-makers in government: ministers, senior officials, and occasionally the head of government on the choices before them. Proposals reach my desk through a long chain of stakeholders: lobbyists, bureaucrats, other policy analysts, and citizens. My task is to integrate those inputs with the wider landscape - the fiscal position, government priorities, and so on. From that synthesis, I offer advice that helps decision-makers choose.
Once they decide, I make it happen, coordinating across departments, coordinating with others to move billions of dollars. Somewhat influential, but not visible. But I cannot make an impact without the evidence and reason to support it. And there is a disconnect between the evidence and reason Effective Altruism associated organisations produce and what is actually useable to me.
There are Effective Altruists in government who hold higher rank, but I haven’t seen many write up the practical craft of policy advocacy for those trying to do the most good. Power works differently in different countries and these reflections are drawn just from experience.
Commandment One: Know Where Power Resides
Knows where power resides. Every problem has a centre of power. Power may rest with an a minister or a department employee. I don't mean where does the power reside on paper, but where specifically for your problem does it lie. There will often be multiple centres of power, and an incomplete chain between them. For instance, maybe the head of government has a cabinet-style of leadership, and the Minister micromanages her department. Maybe the Secretary doesn't get in the weeds and the Director of the policy area just Okays whatever their underlings say. You need to consider factors like where you are in the budget cycle for your country, the personalities of decision makers and how much they like to delegate, and so on. This is not public information. But if you an insider in government, often they will just tell you if you show you can be trusted not to spill that kind of information. Ask them for a meeting, ask how best to achieve your desired policy change not how do I influence these processes.
Commandment Two: Build Relationships
Good policy advice dies in inboxes. You must follow up. Send the brief, then schedule the meeting. Talk to the desk officer, then the manager, then the senior executives, then the Minister then do it all over again and again. You may be annoying but you will be effective. Government is full of inertia; motion requires persistence. Once someone with power on the inside decides to champion your idea, then stop hassling, and allow them to navigate the corridors of power for you - or you might mess it up.
Emails are necessary, but meetings move things. Face-to-face even if virtual is how trust forms and priorities shift. Do not meet with people for exploratory meetings. Time is precious, have a clear, concrete ask that will shape outcomes.
Commandment Three: Build Consensus
In democracies, consensus is currency. Work to produce cosigned advice - submissions endorsed by all the largest stakeholder and most sensitive groups. It signals coordination and lowers political risk for the minister.
Ask decision makers for funding to do the stakeholder engagement and consult, if you need to.
Commandment Four: Be Credible — Don’t Be Cringe
The presentation of seriousness is how bureaucratic systems decide who to listen to. If you have excellent substance but an embarrassing organisation name, I will be embarrassed to put you forward to a decision maker. Alliance to Feed the Earth in Disasters, Existential Risk Observatory, Global Catastrophic Risk Institute, Future of Humanity Institute. These are over-earnest organisation names.
Your credibility doesn’t come from moral urgency or the quality of your argumentation. The decision-maker doesn’t share your moral lens; they care about whether you can deliver within constraints. Influence depends less on ideals than on institutional weight.
Commandment Five: Stay in Your Lane — Don’t Give Politicians Political Advice
Politicians are experts on politics. They have political advisors, and they have policy adviser. However they are far more comfortable outsourcing policy. Don’t cross that line. Never tell them something “is in your political interest" or how they should play the politics. Maybe it's ego or suspicion, but behind closed doors and smiles politicians generally take it really, really bad when they get political advice from a stakeholder unless they're from their political party or another closely politically aligned group.
Commandment Six: Put the Bottom Line Up Front (BLUF)
Example: “We recommend Option B — it achieves X within Y timeframe at Z cost and is consistent with ABC policy frameworks that already exist that the Government department DFG recently published.” Then backfill the evidence.
Decision makers and their advisors scan; they don’t read. Be visual but don't be all visual, we need to copy and paste into our briefs - recommendations (keep it to just a few), sensitivities, risks, background, timing, consultation. Copyright doesn't apply internally.
Recommendations must be time-bound and costed: what action, when, and for how much. Include clear metrics if possible — “achievable within six months,” “requires $80M in fiscal year 2026-27,” “review in 18 months.".
However, remember decisions are made slowly in Government. Yes there are exceptions, but if you knew how to navigate those you wouldn't be reading this introductory post. If you give a decision make an unreasonable deadline, it's easier to just ignore your advice than give it a platform and make government look bad.
Commandment Seven: Don't make government look bad
This is not a golden rule. More advanced advocates can and do subvert this effectively. However, if you are critical of government in comments to the Media for instance, everyone notices. Unless you're so influential they can't ignore you, they won't trust you inside the tent with the info you need to influence the right things at the right time.
Commandment Eight: Humility
Policy work actually rewards epistemic humility. It’s okay to say “I don’t know,” especially about things like fiscal implications or political timing - things insiders know best. Just don't use terms like epistemic humility (see my point on earnestness earlier).
What matters is follow-through. Ask them what you need to make your policy proposal work and then do it.
Commandment Nine: Engage Across the Aisle
Power moves. The party of government matters now, but the opposition does not need to be your opposition. Not because tomorrow they may govern - frankly political horizons are so short they are unlikely to hold anything against you. Maintain professional relationships across political lines. Talk to everyone who holds real influence, not just those who share your values, positions and not just those in the party of government. Not because they are centres of power, but they have knowledge and advice.
Command Ten: Don't wait
Act now. Your comparative advantage is expertise, that insiders do not have. However, we insiders will not wait for you to catch up. The decisions are being made now, and once locked in it can be very hard to stop, and very opaque to know how far ahead decisions we're locking in and how consequential they will be. But without having the right materials across my desk, I need to justify my advice and decisions with what I have. I simply don't have the time for original research or to translate more theoretical advocacy from Effective Altruist organisations into something actionable. If a non EA source has recommendations that are tightly written enough to copy and paste in as a specific decision the government of the day can make under a piece of legislation, or as part of the development of a new legislation, or to allocate to a new funding proposal in the budget, and an EA source is written at a higher, vaguer level, if I want to get my job done in time and keep my job I will use the non EA source.