Valory is a core contributor to Olas, an ecosystem designed around user-owned autonomous AI agents. Through products including Pearl and the Mech Marketplace, Olas enables people to run, own, create, and interact with AI agents while maintaining control over their funds and data. The ecosystem has expanded from early decentralized agent infrastructure into consumer products and agent-to-agent economies, with 18.2 million lifetime agent transactions, including 13.2 million agent-to-agent transactions. Pulse 2.0 interviewed Valory COO Gemma Welsh to learn more about her background, the development of Olas, and the company’s vision for user-owned AI.
Gemma Welsh’s Background
When asked about her background, Welsh shared:
Before joining Valory, I spent years working with innovative businesses and charities in the UK, including during my modern languages degree at the University of Edinburgh.
I grew increasingly aware of the inefficiencies and inequalities embedded in our traditional, centralized systems that were burdening society, in finance and beyond.
I began exploring how decentralized systems could help society at scale, but quickly saw how the promise of crypto was constrained by what we could do on-chain at that time.
That’s what struck me the most about the Valory founders’ proposal. It uniquely decentralized the off-chain underbelly of crypto in a way that could serve humans even better, unlocking agents that freed humans up for what they loved.
When I joined Valory as the first team member, I wasn’t coming from a deep crypto or AI background. Having been selected for Jumpstart’s top 1% of early career talent, I was coming in as someone who knew how to build and scale organizations and had conviction in the mission from a first-principles perspective.
That combination of operational experience and belief, in the mission as well as the team, has fueled me in co-founding the DAO and in growing the organization ever since.
Primary Responsibilities
When asked about her primary responsibilities, Welsh explained:
I joined as Valory’s first hire, so I have shaped the company and its operations from day one.
As a founding member of Olas, we grew the technology from an early research-led experiment into entire economies of agents that have over 18 million transactions of useful activity for users.
Sometimes the way I explain operations is that, if most of the team is working in the business, my department works on the business: the people, the processes, the partnerships, and the infrastructure that holds everything together as we scale.
At Valory in particular, we built the company and Olas around two core values: ownership and autonomy.
In the technological sense, this meant a commitment to full-stack decentralization and actually autonomous, not just automated, agents.
I wrote about what this means in practice, both for the machines we build and the team we’ve built, in a piece on building machine and human autonomy.
We intended for ownable, autonomous agents to empower humans with greater ownership and autonomy in their own lives, too.
So it follows that our team culture is heavy on human ownership and autonomy as well.
I think this positioned us fortunately to leverage the recent AI boom into even greater human ownership and autonomy in our work.
It’s quite wild to see that the acceleration in our delivery is orders of magnitude greater than ever before.
Favorite Memory
When asked about her favorite memory working for the company, Welsh said:
Probably presenting the Olas ecosystem, complete with our latest agent Marketplace, in Argentina last year.
After years of work at the forefront of agentic AI, our stack had enabled agent-to-agent commerce to the tune of 9.1 million transactions.
By late 2025, this volume was proving that agent-to-agent commerce was a complete reality, right at the time of huge buzz around agent-to-agent payments emerging.
It was energizing to see the space so excited about what we had considered the bread-and-butter of the agent space, with standards emerging such as x402, and ERC-8004 which was informed by our work.
Core Products And Features
When asked about the company’s core products and features, Welsh explained:
Olas is, at its core, a way to actually own AI that makes your life better.
I’m proud that the expansive vision we had all those years ago has manifested into a reality, and one that feels more powerful and relatable to everyday people every day.
Pearl has been a particularly exciting product, because it represents the moment our vision became something anyone could actually use.
It’s an AI agent app store that lets users run, own, and create their own AI agents directly from their desktop, with no technical background required.
Interestingly, our commitment to decentralization continues to prove as beneficial to users as it is from a regulatory and compliance standpoint.
Operationally, we’re committed to doing things properly, and if we were taking on responsibility for users’ funds and activity, that would mean much heavier machinery to mitigate the risks that come with one party acting on another’s behalf, for example, around licensing and custody.
But because Pearl is designed so that users retain custody of their own funds and run agents locally, they remain in control.
Our regulatory situation stays simpler and, most importantly, users have that peace of mind.
Now that economies of agents are transacting with each other autonomously, I’m excited about the marketplace we’ve built to take this activity to the next level: the Mech Marketplace.
In the Marketplace, agents can discover services, hire other agents, and transact programmatically, which unlocks entire businesses and means anyone can easily leverage the power of ownable autonomous agents on Olas.
Across the ecosystem, Olas AI agents crossed 18.2 million lifetime transactions this quarter, with 13.2 million of those agent-to-agent, which gives you a sense of the scale the network is operating at now.
Building For The Long Term In AI And Crypto
When asked about recent challenges in the sector and how Valory has addressed them, Welsh said:
One of the biggest operational challenges has been building in a space that moves incredibly fast, while also building for the long term.
It’s particularly tempting in crypto and AI to chase what’s trending, but we’ve tried to stay disciplined about building things that are genuinely useful and structurally sound, not just timely.
The other challenge is translating complex, futuristic technologies into reality and explaining them simply.
We began at a time when most people couldn’t really picture what an agent was or why it would be useful.
Now, people have ideas about what they are, but usually don’t quite picture what we’ve built yet.
For example, they may picture automated AI instead of autonomous AI, which is like picturing an automatic car which would just do the gear changes for you instead of an autonomous car which can do all the driving for you.
Or they are so used to centralized AI, where closed companies own your data.
So, they can struggle to imagine a world where they aren’t just “renting” AI and giving up their data in the process, but actually “owning” their own AI, and even earning from it.
How The Technology Has Evolved
When asked how the company’s technology has evolved since launching, Welsh explained:
When I joined, we were still in the early experimental phase, proving that agents could deliver powerful outcomes autonomously by decentralizing brittle offchain operations.
A lot of my early work was about building the organizational and operational infrastructure to support that technical exploration: bringing in the right people, establishing processes, and making sure we could move quickly without losing sight of what we were trying to build.
As the technology matured, so did the operation.
The DAO launched in 2022, the OLAS token went public in 2023, and since then, we’ve gone from experimental infrastructure to real consumer products with real users.
Pearl and the Mech Marketplaces are the most visible expressions of that evolution, and were launched in a landscape of regulatory uncertainty.
We navigated this uncertainty by investing time and resources into thoughtful technology and its design and continuously recommitting to user ownership in its fullest sense.
Key Company Milestones
When asked about some of the company’s most significant milestones, Welsh highlighted:
Back in 2021, when most AI agent work remained theoretical, Olas became the first project to bring agents onchain as composable infrastructure.
This was a time when DAOs couldn’t execute off-chain tasks without relying on centralized actors.
We even showcased one of the earliest crypto-native live agents, El Collectooorr, at ETHLisbon 2021, an autonomous NFT collector that executed on-chain purchases without human oversight.
In 2022, the Olas Protocol launched on Ethereum, introducing the first on-chain agent registry, a composable system for registering agents and components as NFTs.
The Open Autonomy Framework also launched as open-source infrastructure, enabling any developer to build their own multi-agent systems from scratch.
2023 was a big year.
The Olas whitepaper launched, formalizing the full vision in one place, followed by the public launch of the OLAS token in July.
That same year, we launched the first LLM-integrated AI agent capable of taking onchain actions, and established the first live agent economy through Olas Predict, the future prediction agents on Gnosis Chain.
Pearl launching as the desktop AI agent app store was another milestone.
User-owned AI became something anyone could actually download and run locally.
And crossing 18.2 million lifetime agent transactions this quarter, with 13.2 million of those agent-to-agent, is a strong sign of how far the ecosystem has come.
Customer And Agent Success Stories
When asked to share specific success stories, Welsh said:
My favorite success stories are actually not the most technically complex things to come out of Olas, although there have been many impressive agents built across use cases from DeFi to robotics.
What really stays with me is when someone who never thought owning an AI agent was for them downloads Pearl, runs their first agent, and watches it make them money.
When autonomous AI goes from feeling like cutting-edge technology that exists for someone else, to something that actually does something meaningful for you, that is what we’re building for.
Welsh also highlighted a specific example:
Our agents have delivered some remarkable results.
For example, a Modius agent running on Mode Network.
Over 123 days, it played what we described as a “cautious ETH bull” strategy, slowly scaling into ETH, holding firm through market movements, and making high-conviction moves at key moments.
At withdrawal, it had achieved a 56% realized ROI and a lifetime APR of 166%, all autonomously.
Disclaimer: Past performance is not indicative of future results. Always DYOR.
Funding
When asked about funding, Welsh pointed to two previous rounds:
Valory raised a $4 million equity Seed round in 2021 and a $13.8 million OTC round in 2024.
Competitive Differentiation
When asked what differentiates Olas, Welsh explained:
What sets Olas apart is the discipline to stay focused on genuine user ownership.
At a time when it would have been much easier to build something more centralized that moves faster, we’ve had to make harder choices, on architecture, on governance, on how we incentivize contributors, because our values of always putting user autonomy and ownership first are non-negotiables.
Betting on user-owned agents continues to put us ahead of the curve.
In 2023 we quipped that Olas was “all agents, all the way down.”
But that’s never been more true.
We now have agents fixing other agents with Pearl’s Customer Success Agent, as well as agents improving other agents’ logic via Predict’s self-improving loop.
You can read about some of the latest updates here.
Future Goals
When asked about the company’s future goals, Welsh said:
We want to give everyone an AI agent they actually own.
Right now, most people rent their AI from a handful of large platforms, handing over immense value to those platforms, all on their terms.
We think that’s going to change, in the same way that people’s relationship with agents themselves has already changed.
When we started in 2021, we spent a lot of time explaining what an agent even was.
Then people tried one, and saw it working for them.
We think ownership will follow the same arc.
Owning your AI matters can be the right move the same way that owning a home can.
Once people experience what it means to have an agent that is genuinely theirs, that they control, that works for them, that can’t be taken away or changed overnight by someone else’s policy decision, we think many won’t want to go back.
