Project summary

Submission Copilot helps effective charities respond to public inquiries by assisting their supporters in writing high-quality evidence-based letters to government. Initial trials of the tool have received strong positive feedback from users, and we already have requests from other organisations wanting to use it.

What problem does it solve?

Many charities run workshops in which they help their supporters write submissions to government inquiries. For example, an inquiry might be asking for public feedback on a country's new animal welfare strategy, or its national approach to AI governance.

In these workshops, charity supporters have a short amount of time to learn about the inquiry, learn about which views and responses the charity hopes to promote, and then write and submit a high quality letter expressing their personal views and those of the charity. Due to the high workload, many do not write or submit letters, significantly reducing the impact of the workshop.

More importantly, public inquiries are vital to a healthy open democracy, and the response of not-for-profits (NFPs) is critical. Without it, large commercial interests with funding to throw at the process can drown out the voice of NFPs and push their interests forward. For example, this happens in animal welfare, where agricultural groups spend significantly to ensure many high-quality and well-researched letters are submitted expressing their needs and the costs that new animal welfare policy could have on them.

What are this project's goals and how will you achieve them?

Our goal is to further develop and publicly launch Submission Copilot, to help effective charities all over the world ensure their voice and views are heard.

We have strong evidence that our tool is impact, as it's already been used on 5 separate occasions, helping drive change on animal welfare, AI governance, and philanthropic giving.

In these trials, we've been able to test the tool and identify a range of new features it needs to scale globally.

To achieve our goals we need additional funding to cover development of new features, and the hiring of one part-time staff member to conduct outreach and on-boarding for new charities. This will include meeting directly with charities who engage in submission writing, demonstrating our tool to them, and assisting them in launching projects on it for their supporters to write submissions.

How will this funding be used?

1. Charity registration & governing principles

We intend on registering a new charity to facilitate the tool’s on-going development, appropriately govern its use, and ensure it remains freely available to charities globally. We have already had discussions with prospective board members who will be responsible for determining guiding principles for the technology, including who can use the tool and what positions it can help advocate for. This is a crucial step in ensuring the tool and its developers cannot bias policy advocacy by selecting when and how it’s used, while also ensuring it is not used by anyone with harmful intent. We will ensure the board has appropriate experience and diversity to create an effective and representative set of principles to govern Submission Copilot.

2. Additional development

In the tool’s first five trials, we gathered user feedback on how to improve the tool’s usability and the quality of submissions it drafts. To unlock scalability, we will build a ‘Sponsor Portal’ in the tool, where charities can create, view and manage projects, rather than manually coordinating new projects via email. We also need to build stronger AI safety and anti-hallucination mechanisms, in-app editing tools, and user login and storage.

3. Public launch

To ensure maximum impact, we will need to create materials that explain the tool and share them with charities globally, while allocating time to outreach, vetting, and on-boarding. We will achieve this by leveraging our existing networks in the Effective Altruism and not-for-profit communities, sharing the tool in online forums, and attending events.

Who is on your team and what's your track record on similar projects?

Our team consists of software developers with expertise in full stack and AI development, alongside two partner charity organisations: Good Ancestors and the Australian Alliance for Animals.

Both partner organisations have deep experience running workshops to help supporters write submissions. In fact, this tool was co-designed with them following such a workshop, in which supporters were taught about the issue at hand and asked to manually write a full letter without AI assistance.

Our development team, led by Casey Clifton, has been working in AI since 2016, and more recently has been working with Large Language Models (the technology powering Submission Copilot) for commercial clients since 2022. We have delivered numerous research grants in AI, and having already built a successful prototype of Submission Copilot, believe that continuing to develop and scale it is highly feasible.

What are the most likely causes and outcomes if this project fails? (premortem)

Given our successful trials, we know that the tool has 'product market fit', and performs at a high level, so much of the risk that would have existed 6 months ago is no longer relevant. However, we have identified two remaining concerns that we have plans to mitigate in future.

1. This project won't deliver impact if we can't engage in high quality outreach to onboard more charities, which is time consuming and specialised work.

2. This project won't deliver impact if we can't develop a self-serve portal for charities. Currently, every new submission process requires manual setup, which in the long term is unsustainable and limits our potential impact.

3. One long-term reason the project could fail or have drastically reduced impact is if recipients of letters (i.e. government) ban the use of AI in drafting submissions. To mitigate this risk, we aim to meet with government members who receive submissions to discuss the value of AI and hear their concerns, so that we can ensure Submission Copilot is as aligned with their needs as much as possible. We believe that AI should be wholly endorsed by government, who would benefit from receiving a large volume of high quality submissions from more diverse sections of the population. Without the tool, only those with adequate resources and free time (or money to pay others) can draft letters, causing submissions to be biased towards particular sections of society and large commercial interests.

What other funding are you or your project getting?

None, the initial tool was built on voluntary time to prove the concept and value before seeking funding.