Akash Governance Agent: AI-governed DAO based on human signaling

[mehrdadmms]

TLDR;

This post explores a problem and solution hypothesis related to governance and its potential application to Akash. The goal is to start a conversation with governance stakeholders including community members, participants, and validators to better understand pain points and evaluate how an AI agent could help.

The intended outcome is a proposal for community pool funding to develop an Akash Governance Agent that can support delegates, simplify participation, and improve governance processes.

Who Am I

Hi, My name is Mehrdad and I like governance. I’ve been a Technical Product Manager in Web3, with a prior background in Web2.

Relevant Experience

• Led the development of Pocket Network governance upgrade (reference)
• Led the development of Pocket Network RetroPGF platform (reference)
• Fasilitated Pocket RetroPGF program from ideation to execution distributing 400K USD among 83 participating applicants.

Additional Experience

• Founder @ Pattern decentralized devshop (prev known as Microflow) - Developed/Contributed to:
  • Soniva.ai
  • pattern.global
  • HummingBot
  • PocketNetwork
  • RnDAO
  • and more...
• RaidGuild Member
• Web2
  • Tech Director @ CBI Exchange
  • Software Engineer @ Devonso: Dynamics 365 customizations analytics dashboard
  • Founder @ Neolyze: Social Media analytics dashboard
• MSc MMIE - BSc Computer Science

Problem Hypothesis

In general, DAOs have disengaged community, fragmented governance, exponentially growing number of proposals (specially since LLMs came into the scene), and poor reporting; making it hard for community members to engage, while delegates are overburdened, underpaid, and may have self-interests that don't always align with the broader community.

Community Perspective

While DAOs inner circle are usually active, the broader community struggle to engage due to several key challenges:

1. Fragmented Governance: The information around governance is often fragmented among different platforms, additionally to participate in governance community members must follow different discussions and take actions in different places.
2. Lack of Clarity: Proposals are often long and difficult to understand. Participating meaningfully requires time, technical knowledge, and context that many community members simply don’t have.
3. Poor Communication: There is a lack of continuous reporting and active promotion of proposals on social media, making it hard for the broader community to stay informed and engaged.
4. Lack of Sufficient Help: Community members need time and attention from delegates to help and advocate for their needs within the inner circle, but delegates simply can't cater to every community member or provide the support needed.

Delegates Perspective

On the other side, the delegate model has several shortcomings:

1. Low Participation Rate: While the participation rate is significantly higher than non-delegate model, it is still lower than expected and often times delegates must be chased to vote or participate in discussion.
2. Limited Subject Matter Expertise: Not all delegates are experts in every field relevant to governance proposals.
3. Potential Conflicts of Interest: While there's always emphesis over overcommunication, yet still delegates may have hidden personal interests that influence their decisions, which might not always align with the broader community’s guidelines or the long-term vision of the organization.
4. Overburdened and Undercompensated: Delegates represent a collective of community members. It is their responsibility to voice community questions or concerns, provide answers, and promote and report on proposals. However, balancing these responsibilities with other commitments can be overwhelming—especially considering that delegation is typically an unpaid role.
5. Scalability Issue: With AI, we will see exponential growth in the number of proposals as the community expands. Handling this volume of information and responding to every proposal will become impossible, limiting the scalability of governance.

Solution Hypothesis

An AI agent can streamline and enhance the governance process by automating key delegate responsibilities. It can monitor ongoing discussions, interpret sentiment, and form informed opinions based on predefined guidelines and community signals. The agent can engage with the community in real time—answering questions about current and past proposals, and sharing regular updates on governance activity through social media to keep members informed and involved.

Moreover, the AI can simplify the delegation process. Users can interact with it directly to explore delegate options, ask questions, and delegate their tokens—requiring only a simple ask and then transaction signature to complete the process.

Lastly, delegates can replace themselves with an Agent giving it custom public/private guides to act on their behalf based on human signaling.

Prerequisite Key Questions

What is the problem you're trying to solve?

Broadly speaking; reducing governance friction, enhancing governance scalibility, reducing barrier of entry.

How will the work benefit the Akash Network and / or Akash Community?

This work benefits two key groups within the Akash Community:

• Community Members: Agent guides them through Akash governance, providing support and clarity.
• Delegates (Validators): Agent automates governance tasks, transforming information into structured knowledge that they can easily query. It can also act on their behalf based on human singals.

It also benefits the Akash Network by:

• Driving higher engagement in governance proposals.
• Increasing the participation rate of validators (as of my last check, one-third of bonded validators participated in less than 50% of the proposals).

Are there reasons your idea might not work?

1. Risk: This may not be an issue at all. (invalid problem hypothesis)
   Mitigation: This proposal to spatk the canvo.
2. Risk: Using AI agents for DAO governance is pretty complex and can be controversial
   Mitigation: Incremental features based on current painpoints that can be adapted by delegates. Switching over to fully AI-governed DAO can be controvertial but a blend of human delegates and AI-delegates acting on human signal with proper fail-safe mechanism can work.
3. Risk: it might end up causing more issues than it solves
   Mitigation: Creating one AI Agent Delegate as a pilot program to demonstrate the value and identify shortcomings and issues.

This discussion was created to spark conversations with Community members, Governance participants and delegates to assess their needs and current governance frictions to identify additional risk and mitigation methods.

Is work already being done somewhere in the Akash Community?

If a governance AI Agent existed, I’d answer this question with 100% confidence. But for now, I’ll just go with "I believe not."

Is the project you're working on open-source or closed?

It will be open-source.

Beyond Akash

In one sentence: AI-governed DAO based on human signaling.

Related Work

As a PoC, RaidGuild and I have created TOAD. A Trustless Onchain Autonomous Delegate. It indexes proposals and discussions, forms an opinion and votes on behalf of delegators.
You can check it out at:

• Website: http://toadn.com/
• Repo: https://github.com/raid-guild/toad-monorepo

Ask

This discussion was created to kick off market research on the problem/solution hypothesis.

I’d like to speak with interested community members, governance participants, and delegates (validators) to better understand their pain points and explore the validity of the hypothesis.

If validated, The insights from these interviews will be used to create a proposal for community pool funding to support the development of the Akash Governance Agent. These interviews will help define:

• Core features to prioritize
• Timeline, budget and key deliverables
• Impact metrics and success criteria

[piber-dev]

hi @mehrdadmms thanks for writing this! here's what i think:

In general, DAOs have disengaged community, fragmented governance, exponentially growing number of proposals (specially since LLMs came into the scene), and poor reporting; making it hard for community members to engage, while delegates are overburdened, underpaid, and may have self-interests that don't always align with the broader community.

i don't think that akash has this problem yet, so i don't think the problem hypothesis is valid at this moment. there is a pretty long process from idea to proposal detailed in the "Best Practices : Akash Governance Proposal's" doc which is practically a requirement if you want your proposal to pass. what this means is that it's very easy for validators to find the serious proposals, as there will be discussion links and mentions of the projects in multiple SIG/WG calls before a serious proposal hits the chain.

i do like that the idea does not force validators or other users to use this tool - it's just a tool that can be used. do you know of any examples of this being done in other communities where we can see how it could look like? the akash report group is currently starting to write down and publish summaries of everything that is happening on akash, which will be published to the open-source website repo and a newsletter - maybe this is enough for validators to get their "lazy information" from? at the end of the day, if they do not understand a topic, i don't think they should vote yes or no blindly.

[mehrdadmms]

That makes sense. It's clear Akash already has strong processes in place. Because of that, and given the current scale, governance overload doesn't seem like a pressing issue.

That said, it makes me curious about amplifying existing processes and filling in community blind spots.

What do you think of an agent that could augment efforts like the Akash Report group. For example, by indexing past SIG/WG mentions of a proposal or auto-tracking dormant groups?

As for other examples, Optimism recently funded a series of projects through Foundation Mission Request: AI Delegate Development. The key focuses of that mission are:

• Autonomous Agent: Makes governance decisions without human input.
• Analytical Depth: Understands complex proposals and their implications.
• Transparent Reasoning: Publishes clear logic behind every vote.
• Information Integration: Learns continuously from forums and research.
• Auditability: Open-source and fully verifiable decision trail.
• Self-Assessment: Rates its confidence and knows its limits.
• Milestone-Based Progress: Measured by proposal, prototype, testing, and delivery.
• Real-World Deployment Ready: Eligible for live adoption if successful.

[CaptainSplince]

Hey @mehrdadmms - Thanks for the detailed overview of your idea. Two things pop out for me. One is Akash is not technically a DAO so in order to make this work, we would need to create or formulate an Akash DAO for this to become a reality, correct?

Point two is, If you're passionate about this, you could technically start work on this today. If your request is, "kick off market research on the problem/solution hypothesis," you're more then welcome to. The best practices Piber mentioned above is very focused on funding ideas, but it looks like your request isn't asking for funding of anything at this time. So I believe your best next steps is to continue to source community feedback and support for your idea.

[mehrdadmms]

Hey @CaptainSplince really appreciate the feedback. Could you clarify what you meant by “Akash not being a DAO”? From my perspective, while Akash isn’t explicitly branded as a DAO, it operates as a decentralized organization.

My core interest is in blending governance with AI to reduce friction and support decision-making based on community signals. That said, this idea has evolved. After several conversations on Discord, I realized the current governance setup already meets the community’s needs. There’s no burning pain point to solve right now, so I’ve shifted my focus toward amplifying existing processes and helping fill in blind spots with AI. Based on your experience, do you see any gaps I should be focusing on?

I'm not seeking funding at this stage, just exploring whether there’s enough potential here to justify a community pool proposal and eventually a product. I’d really appreciate it if you could share this with anyone who might be interested in the topic.